Public support for the legalization of recreational marijuana has increased substantially over the last decade and a half, rising from 25 percent in 1995 to 68 percent in 2020 (Gallup 2020). Between January 2012 and December 2022, 21 states and the District of Columbia have enacted recreational marijuana laws (RMLs), which legalize the possession, sale, and, in most cases, home cultivation of small quantities of marijuana (e.g., one or two ounces) for those ages 21 and older for any purpose, including recreational use. The first state to enact an RML was Washington on December 6, 2012, followed by Colorado four days later. Each state that has adopted an RML had previously adopted a medical marijuana law (MML) on average,13 years prior. Unlike most MMLs, RMLs do not require registration as part of a state registry nor do they require a doctor's recommendation to treat an “allowable medical condition” (National Council of State Legislatures 2022). To legally obtain marijuana at an open recreational dispensary, individuals must simply show proof that they are age 21. In addition, nearly all states that legalize recreational marijuana permit home-growing of marijuana plants (Anderson and Rees 2021).

While RMLs have garnered substantial support among policymakers and the public, the American Medical Association (AMA) and the American Public Health Association (APHA) have questioned their efficacy and withheld their support for recreational marijuana legalization (AMA 2021; APHA 2020). Public health experts warn that the legalization of recreational marijuana could normalize smoking among adults, leading to an increase in tobacco use (Ong 2016). One piece of evidence used to support this hypothesis is that co-use of marijuana and tobacco, often as “blunts,” has increased in recent years among U.S. adults (Goodwin et al. 2018; Schauer et al. 2015; Coley 2021).

Opponents of RMLs argue that frequent and heavy marijuana smoking has been linked to lung disease (American Lung Association 2021) and a host of respiratory problems, including chronic cough, bronchial episodes, and increased phlegm productivity (National Academies of Sciences, Engineering and Medicine 2017).1 Many of the adverse respiratory health effects of marijuana smoking are shared with tobacco smoking (Tashkin 2013). Nonsmokers’ rights organizations argue that secondhand marijuana smoke is as important to curb as secondhand tobacco smoke (American Nonsmokers’ Rights Foundation 2021). In addition, there is evidence that frequent marijuana use (especially at younger ages) is linked to increased risk of depression (Buckner et al. 2010; Dumas et al. 2002), psychiatric problems (Guttmannova et al. 2017; Keith et al. 2015; Zvolensky et al. 2006, Zvolensky et al. 2008), and impaired cognitive function (Jager and Ramsey 2008). There is also evidence that when combined with pharmaceutical drugs or alcohol, marijuana use may generate important health risks to users (Ashton 2018; Sokolovsky et al. 2020).2

However, adverse health effects of marijuana use are, in part, mitigated by smoke-avoidant methods of marijuana consumption (i.e., baked goods, edibles, beverages). Moreover, relative to tobacco use, far fewer studies have linked marijuana consumption to cancers, heart disease, or stroke (National Academies of Sciences, Engineering, and Medicine 2017). To the contrary, moderate marijuana use is associated with some important health benefits, including reduced risks of mortality and mobility, and alleviating pain, anxiety, and many side effects of cancer and HIV treatments (Hall et al. 2005; Fiz et al. 2011; Ware et al. 2010; Anderson et al. 2014; Sabia et al. 2017; Choi et al. 2019; Anderson and Rees Forthcoming).

In contrast, tobacco use is the leading cause of preventable death in the United States (U.S.) and has been linked to nearly one-half million deaths per year (Centers for Disease Control and Prevention 2021a). In addition, its consumption leads to increased risk of emphysema, cancers of the colon, liver, head, and lung, and stroke (U.S. Department of Health and Human Services 2020). If marijuana and tobacco are complements for adults, the direct medical costs — including “internality” costs resulting from non-rational addiction (Gruber and Köszegi 2001) — and health-related externality costs of RMLs could be substantial.3

Recent research shows that the legalization of marijuana for medicinal purposes is associated with a small reduction in tobacco cigarette use, suggesting that medical marijuana and tobacco are substitutes for adults (Choi, Dave, and Sabia 2019). However, the effects of RMLs could differ from MMLs for a number of reasons. First, to the extent that a non-trivial share of those who were induced to consume marijuana from MMLs were treating allowable medical conditions (i.e., pain, fibromyalgia, nausea, and side effects of cancers and HIV treatment), the marginal individual induced to consume marijuana from an RML may differ substantially on characteristics related to health production (i.e., age, health stock, depreciation rate). In this circumstance, the local average treatment effect (LATE) of RMLs on adult tobacco use may differ as well.

Second, the market for tobacco products changed dramatically from the mid-1990s, when the first MML was enacted (California in 1996). For instance, approximately one-quarter of the U.S. population lived in states with an MML before electronic cigarettes (e-cigarettes) were introduced to the U.S. market in 2006-2007 (Office of the Surgeon General 2016). While initial e-cigarette sales occurred largely via internet sales, it was during the 2009-2012 period that retail sales of e-cigarettes widely expanded (Huang et al. 2021). Moreover, the introduction of JUUL (Juice USB Lighting) in June 2015 — accompanied by a wide assortment of flavors — also greatly expanded the set of electronic nicotine delivery systems (ENDS) products (Truth Initiative 2019).

The wide availability of e-cigarette products at the time of RML-adoption may result in a very different tobacco use response. For example, vaping pens and JUUL devices used for e-cigarette consumption may be adapted to permit marijuana smoking (Miech et al., 2020), which could result in a complementary relationship between e-cigarettes and marijuana. On the other hand, if e-cigarettes and marijuana both serve to produce euphoria (“high”), generate utility from a smoking experience, or serve the ends of quitting cigarette smoking, then marijuana and e-cigarettes may be substitutes.

Third, the tobacco policy environment continued to evolve during the period when RMLs were adopted. Between 2012 and mid-2019, state excise taxes on cigarettes increased 36 times with an average increase of $0.473 per pack, with 37 states attaining excise taxes of over $1 per pack (Orzechowski and Walker 2018). In addition, over this same period, 10 states and two large counties enacted taxes on ENDS (Abouk et al. 2023), clean indoor air laws were expanded in eight states (Centers for Disease Control and Prevention 2021b), and 16 states and the District of Columbia enacted a minimum legal purchasing age of 21 for all tobacco products (Hansen et al. 2021).4 These tobacco control policies increase the shadow price of tobacco at the same time RMLs were enacted, which could dampen a complementary relationship and potentially reinforce a substitutional relationship between recreational marijuana and tobacco. Alternatively, those who continue to consume tobacco following the enactment of a wide set of tobacco control policies (i.e., taxes, informational campaigns, clean indoor air laws, minimum legal purchasing ages) may have a relatively more inelastic demand for tobacco than prior smokers who were impacted by an MML. In this case, we might expect adult tobacco consumption to be relatively unaffected by the enactment of RMLs.

Finally, in part due to the success of past tobacco control efforts, a substantial share (55.1 percent) of current smokers expresses a desire to quit cigarette consumption (Centers for Disease Control and Prevention 2021c). This may reflect time-inconsistent preferences, social pressure to respond in this manner, or a desire for the costs of rational addiction to be lower. If the former reason dominates, the availability of a new consumer product, recreational marijuana, may aid these individuals in cessation efforts.

This study is the first to comprehensively examine the impact of the legalization of recreational marijuana on adult tobacco use. In so doing, we make several important contributions. First, we make use of a newly available nationally representative panel dataset, the Population Assessment of Tobacco and Health (PATH). These longitudinal data permit the estimation of (1) dynamic difference-in-differences estimates (event-study analysis), and (2) discrete-time hazard models that partial out individual (time-invariant) unmeasured heterogeneity and model dynamic transitions across consumption margins for both marijuana and tobacco. In addition, we draw auxiliary data from the National Survey on Drug Use and Health (NSDUH) and employ difference-in-differences models using novel estimators developed by Callaway and Sant'Anna (2021) to expunge bias in two-way fixed effects (TWFE) models caused by heterogeneous and dynamic effects of RMLs on marijuana and tobacco use.

We document three key findings. First, “first-stage” results from the NSDUH and PATH show consistent evidence that RML adoption increases adult marijuana use by 2- to 5-percentage-points, including through vaping. Discrete-time hazard fixed effects estimates suggest that this increase in consumption is driven by an RML-induced increase in initiation of marijuana use among prior non-users. Event-study analyses, including those obtained using Callaway and Sant'Anna (2021) estimates, produce findings consistent with parallel pre-treatment trend. Second, we find no evidence that legalization of recreational marijuana increases adult tobacco use. The preponderance of the evidence suggests that RML adoption is associated with a small, lagged decline in tobacco use. NSDUH-based estimates suggests that two or more years after the adoption of an RML, legalization is associated with an approximately 0.5 to 2 percentage-point decline in tobacco use, including from cigarette consumption. Individual fixed effects estimates from the PATH show that RMLs are associated with a significant 1-to-2 percentage-point lagged decline in ENDS use. Third, we find that RML adoption accompanied by the opening of recreational marijuana dispensaries is associated with larger increases in ENDS use than RML adoption without open dispensaries. We conclude that the availability of recreational dispensaries is important supply channel to explain substitution between marijuana and tobacco among adults.