Further Consideration of the Impact of Tobacco Control Policies on Young Adult Smoking in Light of the Liberalization of Cannabis Policies

Abstract

Introduction

Changing patterns of cannabis consumption related to the liberalization of cannabis policies may have a countervailing effect on tobacco use. We analyzed whether cannabis policies have tempered the effects of tobacco control policies as well as the extent to which they were associated with young adult cigarette smoking.

Aims and Methods

Combining data on tobacco and cannabis policies at the state, county, and city levels with the nationally-representative geocoded National Longitudinal Survey of Youth 1997 and Census data, we use multilevel regression and fixed effect analyses to examine the impact of cannabis policies on any past 30-day cigarette smoking, frequency of smoking, and past 30-day near-daily smoking among young adults while accounting for community and individual covariates.

Results

Tobacco control policies, including significant effects of comprehensive smoking bans, total vending machine restrictions, single cigarette sale restrictions, and advertising restrictions, remain robust in reducing young adult smoking, net of cannabis policy liberalization, including the legal status of possession, penalties for sale, and medical cannabis. Cannabis policies do not directly affect young adult smoking patterns in an adverse way.

Conclusions

This paper provides evidence that the liberalization of cannabis laws has not adversely affected the efficacy of tobacco control efforts.

Implications

While the effects of tobacco control policies on smoking are well-established, little research has considered how the liberalization of cannabis policies may affect these relationships, which is important given the co-use of these substances. This paper provides evidence that the liberalization of cannabis laws has not adversely affected tobacco control efforts.

Introduction

Tobacco policies such as smoke-free air laws, increased excise taxes, and sales restrictions, among others, have proliferated over the past three decades. Declines in tobacco consumption, particularly among adolescents and young adults, have followed tobacco control policy implementation.^1–21 In 1990, 31.5% of young adults (18–25) reported past-month smoking, a prevalence that declined to 17.5% by 2019.^22,23 Although the increase of tobacco control policies during recent decades is a positive development for population health, during the same period several countries and U.S. states have liberalized laws related to cannabis, increasingly moving towards decriminalization, legalization of cannabis for medical use, or legalized recreational use. In doing so, they have enacted a range of liberalized policy positions that further move away from policies of criminalization that may inhibit use. For instance, among states that have legalized recreational cannabis, all first passed laws legalizing medical cannabis. However, other forms of liberalization, such as decriminalization that removes criminal penalties for personal possession, are also important, as decriminalization often represented an intermediate step between criminalization and legal recreational cannabis (eg, California, Colorado, Massachusetts). Further, decriminalization remains the policy in many other states today following prior periods of criminalization (eg, Maryland, Missouri). Numerous cities also have decriminalized before the state in which they are located did (eg, Detroit, Bloomington, IL) or are located in states that have yet to decriminalize (eg, Miami, Orlando, Tampa, Atlanta, Philadelphia, Pittsburgh, Madison), emphasizing the importance of the local level in analyses. Although use of cannabis and tobacco are interrelated, the impact of these changing cannabis policies on tobacco control efforts has been less well-considered.

The interrelationship between tobacco and cannabis consumption has been well-established.²⁴ Although concerns have traditionally centered on tobacco as a “gateway” to cannabis use,^25–27 cannabis use also shapes patterns of tobacco consumption.^27–36 The co-use of both substances is common among young people. In the U.S. during 2019, among past month combustible cigarette smokers, 46% of those aged 18–25 and 60% of adolescents aged 12–17 also used cannabis in the past month.²³ Further, once cannabis use begins, frequency of cigarette use increases, such that cannabis use is associated with more rapid transitions into daily smoking and nicotine dependence.^27–30 Coadministration is common through cannabis–tobacco mixtures or “chasing” cannabis with cigarettes and can produce nicotine withdrawal.^30–36 Tobacco producers have taken advantage of this relationship through comarketing of cannabis and tobacco products, such as cigars and vaping devices.^37,38 Even with the rise of e-cigarettes, the most prevalent method of co-use remains combustible tobacco and cannabis products.³⁴ Regarding tobacco control efforts, if access to cannabis increases and tobacco control efforts are undermined by changes in the legal status of cannabis, we anticipate this might affect cigarette use and potentially adversely impact gains made through implementing tobacco control policies.

Changing patterns of cannabis consumption associated with the liberalization of cannabis policies may have a countervailing effect on trends in cigarette use driven by tobacco control efforts. Within a context of the co-use relationship between tobacco and cannabis, the influence of policies affecting one substance may have a spillover effect on the other. For instance, Miech and colleagues attributed stability in the rate of cannabis use among high school students in the wake of cannabis policy changes to substantial reductions in youth tobacco use.³⁹ This suggests that tobacco control policies may have had an impact on patterns of cannabis use among young people, by tempering cigarette consumption and ultimately transitions to cannabis use within a changing cannabis policy context. However, alternative patterns of use may manifest from the relationships identified above, wherein mechanisms such as co-use and coadministration made increasingly possible through liberalized cannabis policy increases tobacco use and dependence. Thus, we may expect that the liberalization of cannabis policies might reshape individual-level cigarette consumption patterns in a manner that disrupts the gains made through tobacco control policy. As such, the reassessment of the effects of tobacco control policies in light of the shifting cannabis policy context is important. In this manuscript, we examine whether longstanding associations between cigarette use and tobacco control policies hold in models that account for cannabis policy changes, and secondarily whether such changes in cannabis policy affected cigarette use among a cohort of U.S. youth as they aged into adulthood.

Methods

Individual-level Data

Individual-level data came from the National Longitudinal Survey of Youth 1997 (NLSY97), a large, U.S. nationally-representative, geocoded sample (N = 8984). Adolescents ages 12–16 were randomly sampled in 1997 and surveyed annually to 2011 and biennially thereafter. The retention rate was 80% by 2015, the most recent available year substance use was queried. Cannabis use was not queried in 2013, and thus, we exclude that survey year in our analysis because cannabis use is an important control variable. That is, the analysis includes NLSY97 data from the years 1997 to 2011 and 2015. The restricted-access, geocoded NLSY97 identified the respondents' metropolitan statistical area (MSA), county, and state. We analyzed a subset of respondents whose city of residence could be identified by combining MSA and county information with a variable assessing whether the respondent lived in the MSA's principal city. We restricted our analyses to those living in the largest principal city of an MSA, given the importance of local-level tobacco control^{20,21,40–42} and increasing implementation of local-level cannabis policy. This subset amounts to 39 014 observations among 5373 individuals within 269 cities. We retain about 80.5% of the data in multilevel models after excluding missing data. Table 1 provides descriptive statistics and detail on variable coding.

Table 1.

Descriptive Statistics for the NLSY97 Pooled across all Observations (N = 39 014)

	Mean or %	SD	Range
Individual-level outcomes
Any cigarette use	30.79%
Frequency of cigarette use	6.56	11.61	0–30
Near daily cigarette use	18.51%
Cannabis policy
Cannabis possession
Felony	4.23%
Misdemeanor	75.63%
Decriminalized	19.62%
Legal	0.53%
Cannabis sale felony	49.99%
Medical cannabis
Illegal	74.93%
CBD only	1.99%
Legal with restrictions	0.93%
Legal	22.15%
Tobacco control policy
Comprehensive Smoking Ban	25.13%
Young Adult Possession Restriction	72.87%
Single Cigarette Sale Restriction	35.52%
Vending Machine Restriction	10.22%
Any Advertisement Restriction	67.55%
City variables
Excise tobacco tax	1.02	1.01	0.03–6.16
Population size
less than 50 000	4.11%
50 000–100 000	12.91%
100 000–250 000	18.75%
250 000–500 000	15.41%
500 000–1 million	19.12%
Greater than 1 million	29.69%
Population density (log)	8.20	1.02	4.99–10.23
% Owner occupied	50.15	10.33	21.20–88.78
% Under 18	24.29	3.56	5.10–34.95
% Female-headed households	15.75	4.85	2.30–31.62
% Non-Hispanic White	47.66	19.46	3.24–94.92
Respondent variables
Cannabis use	15.72%
Age	22.12	5.14	12–36
Married	14.04%
Both parents in household	51.49%
Moved between counties	12.20%
Children	29.63%
Work schedule
Daytime	31.74%
Evening	42.26%
Nighttime	4.33%
No single time of day	21.67%
Employment status
Did not work	40.14%
Part-time	23.78%
Full-time	36.07%
Enrollment status
Less than high school	11.60%
High school or GED	21.27%
Some college, not enrolled	15.71%
2 Year degree	2.64%
4 Year degree or higher	13.30%
Enrolled in high school	20.01%
Enrolled in college	15.46%
Received mostly A's in high school	10.90%
Age at initial interview
12	19.98%
13	20.35%
14	20.29%
15	20.21%
16	19.16%
Female	50.43%
Race/ethnicity
Non-Hispanic White	34.00%
Non-Hispanic Black	37.79%
Hispanic or Latino	24.15%
American Indian, Eskimo, Aleut	0.78%
Asian or Pacific Islander	1.84%
Other	1.44%
% of Peer smokers
less than 10%	27.06%
About 25%	22.18%
About 50%	24.45%
About 75%	18.26%
More than 90%	8.06%
Parent's educational attainment
Less than high school	18.38%
High school diploma	31.89%
Some college	23.78%
4 Years of college or more	25.94%
Parent's Health
Good-excellent	75.12%
Fair-poor	13.63%
No parent info	11.25%
U.S. Native	94.98%

For continuous variables, mean, SD, and range are displayed. For categorical variables, the percentage in each category is displayed.

Dependent Variables

In each survey, respondents who indicated they ever smoked an entire cigarette were asked the number of days they smoked during the 30 days before the interview. We created three outcome variables: any past month cigarette use, frequency of use (the number of days of use during the prior 30 days, logged to account for the count structure), and a variable for smoking near-daily (25 or more days in the past 30). The latter is coded as such to assess more regular patterns of smoking.

Independent Variables

Age was entered as a series of fixed-effects to account for time. Age in 1997 was included to control for cohort effects. Regarding family, we included indicator variables for whether the respondent lived with a parent, was married, or had children. We also accounted for a past year move across at least one county. We included categorical variables for job status and schedule, percentage of peers who smoked in 1997 (the only year it was measured), and a dummy variable for receiving “mostly A’s” in high school for academic performance. We also included parents' self-reported health as a proxy for intergenerational health influences and parents' education level to capture household socioeconomic status (SES) during adolescence. The respondent's SES was assessed using a time-varying measure that combined school enrollment status and degree attainment. We included controls for race–ethnicity, U.S. nativity, and gender. Finally, we included an indicator for whether the respondent used cannabis in the past 30 days.

Tobacco Policy Data

Tobacco policy data come from the Americans for Nonsmokers' Rights Foundation (ANRF). ANRF collected a complete repository of tobacco-related ordinances and regulations across the country by date enacted (ie, the date that the policy goes into effect, as there is often a delay from passage). From the ANRF repository, we created a location-year dataset at the state, county, and city levels for each data year. Since higher geographic level policies are not independent of city policies (eg, a state policy automatically means a city smoking ban, and therefore the variables must match), we recoded cities within counties or states with policies reflecting this status. Thus, all policy information is statistically at the city-level. We used Federal Information Processing Standards (FIPS) codes available in both datasets to link the geocoded NLSY97 to ANRF data at the city level. We used these procedures to include several tobacco control policies,⁴³ including comprehensive smoke-free air laws (defined as policies mandating that workplaces, bars, and restaurants are 100% smoke-free without exceptions), youth possession restrictions indicating that it is illegal for minors to possess tobacco in that location, single cigarette sale restrictions (including nationwide coverage beginning in 2011 as a result of The Family Smoking and Tobacco Control Act [TCA] of 2009), complete vending machine prohibitions, any advertising restrictions (above and beyond that of the Tobacco Master Settlement Agreement and TCA), and excise taxes per pack (total amount across geographic levels).

Cannabis Policy

Similar procedures were used to link NLSY97 data to cannabis policies. As no publicly available dataset existed, we obtained cannabis policy data from the National Organization for the Reform of Cannabis Laws, Nexis Uni, HeinOnline, Municode, and American Legal Publishing Corporation. Together, these sources contain all state laws and most county and local ordinances since their inception. Starting with the year 2018, we coded current cannabis policies for all states and the counties and cities represented among the MSAs in the NLSY97. Then, we used legal citations of 2018 cannabis laws to search these databases, working backward in time to identify changes in cannabis policies for each state, county, and city for a complete annual dataset of cannabis policies from 1997 to 2018. This policy data was then merged with the available years in the NLSY97 (1997–2011; 2015). The Ohio State University's law librarians assisted and confirmed the reliability of the data. We include several indicators for cannabis policy (baseline category listed first), including possession status (felony, misdemeanor, decriminalized, legal), criminal sale status (felony, misdemeanor; in locations with legalized recreational cannabis, this represents the punishment if sold without a license), and medicalization (illegal, CBD only (not THC), legal with restrictions (eg, no flower), and legal).

U.S. Census Data

City-level measures from census data are included to account for local-level factors. We created a categorical measure of population, while population density is continuous (logged because of skewness). We also included the percentage of female-headed households, owner-occupied housing, non-Hispanic whites, and minors.

Analysis

We used two complementary approaches analogous to our prior work on tobacco control, which contains more detailed information on both modeling procedures.^19–21,44 First, we used multilevel regression models.⁴⁵ Observations were nested within individuals, which were nested within cities. Our three-level model thus includes random intercepts for both the individual level (Level 2) and the city level (Level 3). These models adjust for person- and city-level averages through a variance parameter defining a normal distribution for each of those averages. At the lowest level denoting time (Level 1), the predictors represented time-varying measures. At the individual level, we had the time-invariant characteristics of the respondent. Since policy and city characteristics were time-varying, they were Level 1 variables. A limitation of this approach is that it cannot assess within-person changes in cigarette use, which is the advantage of our other modeling strategy.

Second, we use individual-level panel fixed effects (FE) models to predict smoking, which have several advantages.⁴⁶ First, FE models account for each individual's change in behavior over time as policy changes, allowing us to interpret the effect of enacting cannabis policies and tobacco control on a given individual's smoking behavior. Second, the models net out the effect of stable individual-level characteristics by person-centering the variables, effectively removing sources of unobserved heterogeneity for all but time-varying predictors. However, because it is necessary to center on the individual's average in the FE framework, an analytic level for the city is not formally included. Although, if an individual does not move, any effect of static city-level propensities for smoking that would affect that individual would be netted out of the model. To be clear, the variables for cities, including policy, are time-varying, which locates them at the lowest level of the model with the time-varying individual-level measures. However, there is no formal adjustment for average differences in smoking between cities. Finally, in the binary models, only those who experience change in the outcome are included (and hence the smaller N in our table). While each modeling approach has limitations, the combination of the two approaches addresses the limitations of each method, with consistent results between them bolstering confidence in the conclusions.

For both modeling approaches, we used the logit link for the two binary outcomes of any past 30-day cigarette use and near-daily cigarette use. Thus, results are presented as adjusted odds ratios (AOR). For frequency, while typical to use count models for days used in the past 30 days, there is no FE model equivalent of a Poisson or negative binomial regression. We took the log link for frequency of use to approximate a count model using linear regression. The exponentiated betas [Exp(B)] from these linear regressions provide the percentage change in the outcome as with count models. The mean Variance Inflation Factor of 3.64 indicated reasonable collinearity among the predictors.

Results

Descriptive Statistics

Table 1 displays descriptive statistics pooled over the entire observation period for all variables included in our models. Thus, while categories and values are mutually exclusive at each time point, they are subject to change over time for time-varying predictors. Here, we focus on the outcomes and policy predictors. Across all observations for our three tobacco outcomes, 30.8% reported past 30-day smoking, there was an average of 6.56 smoking days during the past month, and 18.5% reported smoking near daily. These figures compare well with other nationally representative data of similar age groups from the National Survey on Drug Use and Health.⁴⁷

For cannabis policy, we find that respondents lived in a location where cannabis possession was a felony in 4.2% of observations, a misdemeanor in 75.6%, decriminalized in 19.6%, and legal in 0.5% of observations. For whether cannabis sale is a misdemeanor or felony, half of the observations are in each category. The two most common categories for medical cannabis are illegal (74.9%) and legal (22.1%). For tobacco control, respondents resided in a location with a comprehensive smoking ban in 25.1% of all observations. They resided in a location with single cigarette sales and vending machine sale restrictions in 35.5% and 10.2% of observations, respectively. Minor possession restrictions applied to 72.9% of observations, while advertising restrictions applied to 67.6%. Finally, average excise taxes across all observations were $1.02.

Only considering the mean across all observations, however, obscures the considerable changes occurring in both tobacco and cannabis policies as this cohort ages from 1997 to 2015, shown in Figure 1. The top panel shows cannabis policy. Fully legal medical cannabis climbed steadily over this period, increasing from 12.3% of locations where respondents in our analytic sample resided in 1997 to 21.4% in 2006 to 39.2% by 2015. Decriminalization was steady between 16% and 18% until 2010, then increased rapidly to 31.1% in 2011 and 39.2% by 2015. The criminal status of cannabis sales outside legal venues was relatively stable. Although legalized possession was only 0.5% of all observations, 8.8% of respondents lived in a location with legalized possession in 2015, the only year in our analytic sample with such a coding.

Figure 1.

Percentage of NLSY97 respondents residing in a location with cannabis (top) and tobacco (bottom) policies, 1997–2015.

Changes can also be observed in the bottom panel for several tobacco control policies. Youth possession bans, total vending machine restrictions, and advertising restrictions are relatively stable. However, comprehensive smoking bans and excise taxes steadily increase. Few respondents lived in a location with a comprehensive smoking ban through 2002. From 2002 to 2006, this percentage increased from 3.0% to 20.7%. The percentage then experienced a more rapid increase to 60.6% by 2011. Over the entire period, average excise taxes (on the secondary axis) increased from $0.42 in 1997 to $1.94 in 2015. Single cigarette sale restrictions were stable around 25% of locations in which our respondents resided until the total ban on such sales effective 2011 through the TCA.

This considerable change in both policy domains provides a particular advantage in the FE models that assess whether individual-level cigarette use outcomes change as policy becomes enacted. We now turn to our regression results.

Regression Results

Table 2 shows a condensed version of the results of our regression models that only depicts the effects of cannabis and tobacco control policies; although all covariates are included in these full models, the table with a full listing of all covariates appears in Supplementary Appendix A. Before considering our primary analytic goal regarding whether significant effects hold for tobacco control policies when including changes in cannabis policy, we first discuss the effects of the latter. The results regarding cannabis policy are clear: in no case did we find a significant effect of cannabis liberalization on increased past month use, frequency of use, or near-daily use of cigarettes. In fact, the only significant findings show a negative association in the multilevel models between legal medical cannabis (relative to illegal) and any smoking (AOR = 0.789, p < .05) and near-daily cigarette use (AOR = 0.721, p < .05). Notably, there were no significant findings for the FE models that assess individual-level change in cigarette use as cannabis policy liberalized. We further note that we find similar results when removing individual-level cannabis use from the model (not shown), implying that the effect of cannabis policy on cigarette use is not operating through its effect on individual-level cannabis use, and thus, cannabis use does not obscure any potential significant policy effects.

Table 2.

Regressions for Tobacco Cigarette Outcomes in the NLSY97

	Any cigarette use AOR [95% CI]		Freq. of cigarette use (log) Exp(B) [95% CI]		Near daily cigarette use AOR [95% CI]
	Multilevel	Fixed effects	Multilevel	Fixed effects	Multilevel	Fixed effects
Cannabis policy
Cannabis possession (vs. felony)
Misdemeanor	0.960	0.952	1.026	0.983	1.412	1.176
	[0.595,1.549]	[0.467,1.943]	[0.879,1.197]	[0.816,1.184]	[0.777,2.563]	[0.470,2.941]
Decriminalized	0.993	1.004	1.076	1.058	1.748	2.257
	[0.594,1.659]	[0.482,2.089]	[0.915,1.265]	[0.873,1.281]	[0.922,3.314]	[0.862,5.909]
Legal	1.276	1.103	0.997	0.935	0.697	0.616
	[0.551,2.956]	[0.408,2.983]	[0.788,1.261]	[0.723,1.209]	[0.244,1.987]	[0.173,2.195]
Cannabis sale misdemeanor (vs. felony)	1.167	1.367	1.057	1.067	1.116	1.381
	[0.921,1.479]	[0.933,2.004]	[0.982,1.138]	[0.972,1.170]	[0.838,1.487]	[0.833,2.290]
Medical cannabis (vs. illegal)
CBD only	0.958	0.869	0.964	0.995	0.810	1.038
	[0.638,1.437]	[0.578,1.306]	[0.870,1.069]	[0.898,1.102]	[0.492,1.333]	[0.630,1.711]
Legal with restrictions	0.684	0.802	0.877	0.960	0.589	0.995
	[0.384,1.218]	[0.448,1.435]	[0.762,1.009]	[0.836,1.102]	[0.294,1.180]	[0.498,1.986]
Legal	0.789*	0.812	0.943	0.973	0.721*	0.757
	[0.643,0.969]	[0.646,1.022]	[0.890,1.000]	[0.915,1.035]	[0.550,0.946]	[0.546,1.051]
Tobacco control policy
Comprehensive Smoking Ban	0.852*	0.845*	0.934***	0.939**	0.689***	0.714***
	[0.734,0.988]	[0.729,0.980]	[0.899,0.971]	[0.904,0.975]	[0.572,0.829]	[0.592,0.862]
Minor Possession Restriction	0.965	1.111	0.993	1.044	0.837	1.024
	[0.794,1.172]	[0.879,1.404]	[0.935,1.056]	[0.976,1.117]	[0.640,1.095]	[0.697,1.503]
Single Cigarette Sale Restriction	0.883	0.848	0.963	0.956	0.748**	0.697**
	[0.746,1.046]	[0.706,1.019]	[0.919,1.009]	[0.912,1.002]	[0.607,0.922]	[0.553,0.879]
Vending Machine Restriction	0.562***	0.755	0.830***	0.917*	0.488**	1.154
	[0.411,0.769]	[0.538,1.057]	[0.759,0.907]	[0.841,1.000]	[0.311,0.768]	[0.638,2.088]
Any Advertisement Restriction	0.786*	0.632**	0.926*	0.855***	0.884	0.586**
	[0.636,0.971]	[0.475,0.843]	[0.866,0.990]	[0.791,0.924]	[0.679,1.151]	[0.395,0.872]
Excise Tobacco Tax	1.060	1.077	1.013	1.017	0.976	0.949
	[0.981,1.145]	[0.995,1.165]	[0.993,1.034]	[0.997,1.037]	[0.886,1.075]	[0.861,1.046]
Observations	31,395	16,703	29,056	31,501	29,056	9,920

*p < .05, **p < .01, ***p < .001. Full models with all covariates shown in Supplementary Appendix A.

Importantly, many of the tobacco control measures demonstrate significant negative associations, supporting past research and confirming that tobacco control policies remain effective in the face of changing cannabis policy. Comprehensive smoking bans are negatively associated with all outcomes using both modeling strategies. We use smoking bans to provide an extended example of the interpretation of the models. For the multilevel model, the presence of a smoking ban was associated with 14.8% lower odds of any past 30 days smoking (p < .05). The FE model shows that for a given individual, years living in a location with a smoking ban was associated with 15.5% lower odds of any past 30 days smoking relative to years living in a location without a ban (p < .05). Smoking bans were also associated with a 6.4% decrease in the number of days smoking in the past 30 days (p < .001). In years specifically residing in a location with a ban, the number of days smoking was 6.1% lower (p < .001). Finally, smoking bans were associated with a 31.1% lower odds of near-daily cigarette use (p < .001), and the odds were 30.3% lower specifically in years residing in a location with a ban (p < .001).

Advertising restrictions and vending machine restrictions are also associated with all three outcomes, although not always via both modeling approaches. Advertising restrictions were associated with lower odds of any cigarette use (multilevel: AOR = 0.786, p < .05; FE: AOR = 0.632, p < .05), fewer days smoking (multilevel: Exp(B) = 0.926, p < .05; FE: Exp(B) = 0.855, p < .001), and lower odds of daily use (FE: AOR = 0.586, p < .01). Vending machine restrictions were also associated with lower odds of any cigarette use (multilevel: AOR = 0.562, p < .001), fewer days smoking (multilevel: Exp(B) = 0.830, p < .001; FE: Exp(B) = 0.917, p < .05), and lower odds of daily use (multilevel: AOR = 0.488, p < .01). Finally, single cigarette sale restrictions were associated with lower odds of daily use in both modeling procedures (multilevel: AOR = 0.748, p < .01; FE: AOR = 0.697, p < .01). Thus, the significant effects shown here, including individual-level change as tobacco control policies are enacted, are robust to the changing cannabis policy environment. Removing individual-level cannabis use from the model (not shown) also did not affect the conclusions regarding the tobacco control measures.

Discussion

Cigarette smoking has declined considerably among adolescents and young adults over the past three decades. In 1990, almost 1 out of 3 young adults ages 18–25 (31.5%) reported cigarette smoking within the past month, but by 2019 that prevalence had dropped to 17.5%.^22,23 As described in Figure 1, both tobacco control and cannabis policies have also experienced considerable changes since the 1990s. The primary findings from these analyses indicate that the implementation of liberalized cannabis policies has not disrupted the efficacy of tobacco control policies. Even controlling for cannabis policies, we still found negative associations between the varied cigarette use outcomes and comprehensive smoking bans, advertising restrictions, vending machine restrictions, and single cigarette sale restrictions. The results for smoking bans²⁰ and vending machines¹⁹ via both modeling approaches are consistent with our prior tobacco control research using this dataset before the inclusion of cannabis policies. Thus, cannabis liberalization has not offset gains made in tobacco control. Tobacco control policies continue to be associated with reduced cigarette use outcomes overall, as well as with within-individual reductions in cigarette smoking behaviors when policies are enacted.

Our results also indicate that cannabis liberalization has not altered recent cigarette use, frequency of use, or near-daily cigarette use among a cohort of U.S. adolescents aging into young adulthood. Despite the well-documented relationship between the consumption of cannabis and tobacco,²⁴ cannabis policy liberalization is not associated with individual-level patterns of cigarette use. Cannabis policy could affect individual-level cigarette use through cannabis use, given the high rate of co-use in adolescence and young adulthood and the reinforcing effects between these substances.²⁴ However, our conclusions were the same with or without individual-level cannabis use in our models, suggesting that cannabis policy is not operating in this manner.

We emphasize that our findings were quite robust to two separate but complementary modeling procedures. The multilevel model has the advantage of explicitly controlling for city-level clustering but is ultimately a weighted coefficient of between- and within-unit effects. While the FE models can only cluster at one level and hence the city adjustment cannot be included (and also because some individuals will move between cities), it has the advantage of being able to model within-individual change in cigarette use as cannabis and tobacco policies are enacted while also controlling for any time-invariant unobserved heterogeneity. Together, the similar results from two complementary modeling approaches provide additional confidence in our findings.

Limitations

We must note some limitations. First, we included only young adults whose city could be identified, restricting analyses to subjects residing in central cities of MSAs within the U.S. As such, we limit our generalizability to such young people. Second, the measurement of city-level contextual factors was strategic given that this is the lowest geographic level at which tobacco and cannabis policies are enacted. We recognize, however, that some community-level socioeconomic and other factors function at lower levels akin to the neighborhood. Third, the policy measures are coded as the presence of such a policy. We cannot speak to the enforcement of policies or the relative availability of cannabis via legalized means. While we urge research that incorporates enforcement and availability, we find significant effects of tobacco control regardless, as is consistent with the literature due to the shifts in normalization that occur with new policies.⁴⁰ Fourth, we only examine combustible cigarette use because of data availability. However, given increases in the use of vaping devices among young people for both nicotine and cannabis in recent years,^48–50 research is warranted in this area. Finally, we recognize that this is a single cohort of U.S. adolescents in 1997 who largely experienced cannabis liberalization in late adolescence and early adulthood. Further, the policy landscape for both cannabis and tobacco continues to evolve, with an increasing number of states liberalizing their medical and recreational cannabis laws and new national tobacco policy such as Tobacco 21 and the advent of federal, state, and local policies to combat the rise in e-cigarette use. Thus, we highly encourage additional research on this topic with other age cohorts and in other countries.

Conclusions

Importantly, the liberalization of cannabis laws does not disrupt gains made through the implementation of tobacco control policies. Also, we see no evidence that liberalized cannabis policies are directly associated with increased smoking behaviors among young adults. Within a context of rapidly changing cannabis policies throughout the U.S. and several countries, these results provide positive news that newly implemented cannabis laws may not adversely affect tobacco control efforts that have reduced cigarette use among young people.

Supplementary Material

A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.

Funding

This work was supported by the National Institute on Drug Abuse (Grant #R21DA044447; PI: Vuolo).

Declaration of Interests

The authors have no conflicts of interest to report.

Acknowledgments

This work was supported by the National Institute on Drug Abuse (Grant #R21DA044447; PI: Vuolo). This research was conducted with restricted access to Bureau of Labor Statistics (BLS) data. The authors would like to thank the staff at the American Nonsmokers’ Rights Foundation (ANRF), particularly Maggie Hopkins and Laura Walpert. The views expressed here do not necessarily reflect the views of NIDA, the BLS, or ANRF. We also thank Laura Frizzell, Joy Kadowaki, Emily Harris, Alexandra Marin, Jake Brosius, and Emily Ekl for research assistance, as well as Emma MacGuidwin for legal research assistance.