Notes on long-term effects of minimum legal drinking age laws on marijuana and other illicit drug use in adulthood (Krauss et al., 2015)
Abstract
Topic: Long-term effects of permissive minimum legal drinking age (MLDA) laws on marijuana and other illicit drug use in adulthood.
Data source: National Survey of Drug Use and Health (NSDUH), 2004–2012 linked with historical state MLDA laws.
Sample: Participants born 1949–1972 (age 31–63 at observation), n = 110,300.
Outcome measures: Past month and past year marijuana use; marijuana abuse/dependence (DSM‑IV); past month and past year use of illicit drugs other than marijuana; and illicit drug abuse/dependence (DSM‑IV).
Key findings (full sample): Exposure to permissive MLDA laws during adolescence was not significantly associated with marijuana use or marijuana use disorder in adulthood, nor with illicit drug use or abuse/dependence in adulthood.
Notable exception (men): Exposure to permissive MLDA laws associated with a 20% higher odds of past-year illicit drug use other than marijuana (aOR = 1.20, 95% CI 1.09–1.32).
Implication: Restricting alcohol access during adolescence did not increase long-term marijuana or overall illicit drug use; allowing alcohol purchase before age 21 could increase risk of some illicit drug use later in life, particularly among men.
Supplementary notes: Supplementary material available online; authors describe methods to assess misclassification from using state of residence at survey as a proxy for MLDA exposure.
Introduction
Background on MLDA history:
Pre-1970s: Many states had MLDA < 21; some 18–20.
1970s: Many states lowered MLDA to < 21 (linked to other age-of-majority changes).
Late 1970s–1980s: States raised MLDA back to 21; 1984 National Minimum Drinking Age Act mandated 21 by 1986 to receive federal highway funds.
Prior evidence: Higher MLDA (21) linked to reductions in alcohol use, crashes, and fatalities (DeJong & Blanchette 2014; Lovenheim & Slemrod 2010; Wagenaar & Toomey 2002; others).
Prior findings on MLDA and drugs:
Mixed evidence: Some studies suggest alcohol policies reduce drug use (short-term), others suggest substitution or complementarity.
Examples: Higher beer prices linked to reduced marijuana use (Farrelly et al. 1999; Pacula 1998); increased alcohol price associated with decreased marijuana use in college samples (Williams et al. 2004).
Early long-term work: Norberg et al. (2009) found adults with permissive MLDA more likely to have illicit drug use disorder other than marijuana; marijuana use disorder increase not statistically significant.
Rationale for current study:
Replicate Norberg et al. (2009) in a different national dataset and extend to an older cohort to examine longer-term effects.
Examine both marijuana and other illicit drugs, and both usage and disorders in adulthood.
Consider whether adolescents’ alcohol policy environment shapes later substance use patterns via durable behavioral paths.
Objectives and hypotheses:
Primary: Whether permissive MLDA exposure during adolescence increases the likelihood of marijuana use/use disorders and illicit drug use/disorders in adulthood.
Secondary: Examine whether effects differ by birth cohort (two groups: 1949–1960 vs 1961–1972) and by sex; investigate potential substitution or complementarity effects.
Hypothesized that restrictive MLDA would not increase drug use in adulthood; policy permissiveness could have heterogeneous long-term effects.
Materials and Methods
Subjects
Data: NSDUH 2004–2012, cross-sectional survey of about 70,000 non-institutionalized civilians aged 12+ annually.
Sampling: Independent, multistage area probability samples for all states and DC; computer-assisted interviewing with ACASI to improve reporting of illicit drug use.
Data linkage: NSDUH participant data merged with state MLDA policy data and other state covariates.
Geographic data access: Through CBHSQ Data Portal (UMich) to protect confidentiality.
Inclusion criteria: Participants born 1949–1972 (n = 111,600) who could have been exposed to MLDA changes during 18–21 years of age.
Exclusions: District of Columbia (n = 1,300) due to missing state covariates data, leaving n = 110,300.
Weighting and disclosure control: Reported unweighted Ns rounded to the nearest 100; analyses used sample weights.
Outcome measures
Marijuana:
Past-month use; past-year use; DSM‑IV marijuana abuse/dependence.
Illicit drugs other than marijuana:
Past-month use; past-year use; DSM‑IV abuse/dependence.
Drugs included: cocaine, heroin, hallucinogens, inhalants, non-medical use of pain relievers, sedatives, stimulants, tranquilizers.
Independent variable of interest: MLDA exposure
Definition: Permissive MLDA exposure equals the ability to legally purchase alcohol before age 21.
Data coding: MLDA policy data coded as in Grucza et al. (2012); state of residence at survey used as proxy for state of exposure at ages 18–21.
Addressing misclassification: Supplementary analysis using Panel Study of Income Dynamics (PSID) to estimate cross-state migration; reported misclassification rates and direction (see Supplementary Material 11).
Exposure specification: Binary indicator MLDA{s,k} where s = state of residence at survey, k = birth year; MLDA{s,k} = 1 if state s allowed purchasing before age 21 for someone born in year k; 0 otherwise.
Proxy considerations: Approximately 89% of respondents had MLDA exposure correctly matched under the proxy; 11% misclassified (47% toward permissive, 54% toward 21 MLDA). Misclassification considered to bias toward zero (attenuate effects) rather than create false positives.
Covariates
Individual-level (
X1_{ist}): Sex, race/ethnicity, age (quartiles), education.State-level (
X2_{st}): Unemployment rate, per-capita income (inflation-adjusted to 2012 dollars), citizen political ideology.Fixed effects:
State-by-year fixed effects δ_{st} to control for time-invariant state characteristics and time-varying state characteristics.
Birth-year fixed effects γ_k to control for cohort-invariant, birth-year related factors.
Additional considerations:
Other state covariates (e.g., alcohol taxes) were considered but dropped due to collinearity with state fixed effects (variance inflation factor checks).
Statistical analysis
Primary method: Logistic regression to model dichotomous outcomes as a function of adolescent MLDA exposure, with adjustments for covariates and fixed effects.
Analytic framework: Regression extension of differences-in-differences to multiple time points (Angrist & Pischke, 2008).
Model structure (conceptual):
For outcome Y{i,t} (e.g., past-month marijuana use) for individual i in year t in state s, born in year k,
Where MLDA{s,k} is the exposure indicator; δ{st} are state-by-year fixed effects; γ_k are birth-year fixed effects; X^{(1)} and X^{(2)} are covariates.
Stratified analyses:
By birth cohort: 1949–1960 vs 1961–1972.
By sex: male vs female.
Analytical tools:
SAS (Version 9.2) with the surveylogistic procedure, clustering on state to obtain robust standard errors, and applying sample weights.
Sensitivity analyses:
Continuous MLDA exposure (instead of a dichotomous indicator).
Restriction to participants with high school education or less to reduce misclassification due to migration (see Supplementary Material 12).
Measures of effect and interpretation
Effect size: Adjusted odds ratios (aOR) with 95% confidence intervals (CI) for the association between permissive MLDA exposure and each outcome.
Key interpretation notes:
aOR ≈ 1 with CI including 1 suggests no significant long-term association.
An aOR > 1 indicates higher odds of the outcome with exposure; <1 indicates lower odds.
Reporting conventions: All main results are drawn from models that include state and birth-year fixed effects, and are weighted with NSDUH sample weights.
Results
3.1 Description of the sample
Demographics (weighted):
Women: 51.2%; Men: 48.8%
White (non-Hispanic): 69.0%
Other race/ethnicity distributions provided in Table 1.
Marijuana use prevalence: Past month 4.7%; Past year 7.8%; Abuse/Dependence 0.7%.
Illicit drugs other than marijuana: Past month 2.9%; Past year 6.2%; Abuse/Dependence 1.2%.
Among illicit-drug users (non-marijuana), most common drugs were opioid pain relievers used non-medically (53.4% past month; 57.6% past year).
Cohort differences (birth years):
Younger cohort (1949–1960) vs older cohort (1961–1972) showed higher marijuana and illicit drug use in younger cohort:
Marijuana past month: 5.1% vs 4.2% (p < .001)
Marijuana past year: 8.8% vs 6.7% (p < .001)
Marijuana abuse/dependence: 0.9% vs 0.5% (p < .001)
Illicit drugs past month: 3.4% vs 2.4% (p < .001)
Illicit drugs past year: 7.5% vs 4.9% (p < .001)
Illicit drugs abuse/dependence: 1.5% vs 0.8% (p < .001)
3.2 MLDA laws over time
Time frame: 1967–1992 (participants aged 18–20 during changes).
Trend: Number of states with permissive MLDA (purchase before 21) rose from 14 in 1968 to a peak of 39 in 1976–1978, then declined to 1 state (Louisiana) by 1988–1992.
3.3 Associations between permissive MLDA laws and drug use
Marijuana outcomes (Table 2): Under-21 MLDA exposure and outcomes adjusted for covariates and fixed effects showed no significant associations:
Past month marijuana use: aOR = 0.97 (95% CI 0.87–1.08)
Past year marijuana use: aOR = 1.04 (95% CI 0.95–1.14)
Marijuana abuse/dependence: aOR = 0.99 (95% CI 0.77–1.27)
Illicit drugs other than marijuana (Table 3): No significant associations overall:
Past month use: aOR = 1.05 (95% CI 0.89–1.23)
Past year use: aOR = 1.08 (95% CI 0.97–1.20)
Abuse/dependence: aOR = 1.13 (95% CI 0.88–1.45)
Stratified by birth cohort and sex (Table 4): Generally no significant associations for marijuana outcomes in any cohort/sex.
Notable exception: Men exposed to permissive MLDA had higher odds of past-year illicit drug use (aOR = 1.20, 95% CI 1.09–1.32, p = .023).
Continuous MLDA exposure: Similar results to dichotomous exposure (see Supplementary Material 2; Table S1).
3.4 Addressing proxy exposure misclassification (state of residence at survey)
Restriction to participants with high school education or less to reduce migration-related misclassification did not reveal significant marijuana associations for any outcome.
Illicit drugs other than marijuana among those exposed to permissive MLDA showed:
Past-year use: aOR = 1.15 (95% CI 1.01–1.30) overall among HS- or-less group.
By sex: men aOR = 1.26 (95% CI 1.07–1.49).
By cohort: older cohort (1949–1960) aOR = 1.56 (95% CI 1.09–2.22).
Acknowledgment: Proxy exposure introduces random error rather than systematic bias; cross-state migration analyses (PSID) support limited bias toward Type I error (see Supplementary Material 12).
3.5 Summary of main quantitative findings
Across the full sample:
No significant long-term association between permissive MLDA and marijuana use or marijuana use disorders in adulthood.
No significant association between permissive MLDA and illicit drug use or disorders in adulthood.
Subgroup findings of note:
Men exposed to permissive MLDA: higher odds of past-year illicit drug use other than marijuana (aOR ~ 1.20).
Among those with lower education, exposure associated with higher past-year illicit drug use (aOR ~ 1.15 overall; stronger in men, and stronger in older birth cohort).
Overall pattern: Evidence does not support long-term elevation of marijuana use due to permissive MLDA; possible long-term increase in non-marijuana illicit drug use among certain subgroups, especially men with lower education.
Discussion
Interpretation of primary findings:
Restricting access to alcohol during adolescence did not increase long-term marijuana use or marijuana use disorders.
Allowing alcohol purchases before age 21 could be associated with higher risk of some illicit drug use in adulthood, particularly among men.
Comparison with Norberg et al. (2009):
Norberg reported increased odds of drug use disorder other than marijuana with permissive MLDA; marijuana use disorder not significantly elevated.
The current study largely aligns in showing no robust effect on marijuana outcomes; some evidence for increased past-year illicit drug use among men, which can be viewed as partial replication of Norberg’s illicit-drug finding under different data (NSDUH 2004–2012).
Potential mechanisms:
Alcohol problems may contribute to broader psychopathology (mood/anxiety disorders) which could drive misuse of prescription drugs (common among this cohort) and subsequent illicit drug use.
Substitution vs complementarity: prior short-term studies showed opposition (substitution) or complementarity; long-term results here suggest possible complementarity for non-marijuana illicit drugs, especially in men, albeit modest.
Strengths of the study:
Large, nationally representative sample with longitudinally relevant exposure data across two decades.
Use of fixed effects for state and birth year to control for unobserved, time-invariant confounders and cohort effects.
Consideration of both substance use and disorders; inclusion of both marijuana and non-marijuana illicit drugs.
Sensitivity analyses addressing exposure misclassification via a proxy (state of residence) and migration data (PSID).
Limitations:
Exposure misclassification due to using state of residence at survey as a proxy for state during ages 18–21; though misclassification expected to be largely random, it could attenuate true associations.
Self-report bias in illicit drug use; social desirability could affect reporting; NSDUH uses ACASI to mitigate but biases remain possible.
Inability to capture cross-border access to alcohol (interstate variation in MLDA and local ordinances) or non-purchasing policies for minors.
Possible confounding from contemporaneous age-of-majority policy changes and other policy reforms (e.g., contraception access, education policy) during the 1970s–1980s; though stratified analyses by birth cohort aimed to address this.
Generalizability limited to non-institutionalized adults; results may not apply to currently incarcerated or institutionalized populations.
Conclusions:
A 21-year MLDA did not increase long-term marijuana use or marijuana use disorders.
Allowing under-21 alcohol purchase was associated with increased past-year illicit drug use (non-marijuana) in adulthood in some subgroups (notably men and those with lower education), suggesting possible long-term complementarity for certain illicit drugs.
Policy relevance: restricting alcohol access during adolescence appears not to elevate long-term marijuana risk; any potential long-term illicit-drug risk should be weighed in policy discussions about underage drinking laws.
Supplementary materials and references
Supplementary materials available online (see article’s note on online version and DOI).
Key references cited in discussion include: DiNardo & Lemieux (2001); Cameron & Williams (2001); Farrelly et al. (1999); Norberg, Bierut, & Grucza (2009); Kaestner & Yarnoff (2011); Grucza et al. (2012); Norberg et al. (2009); DeJong & Blanchette (2014); Wagenaar & Toomey (2002); Wechsler & Nelson (2010).
Appendices and tables (summary references)
Table 1: Demographics and prevalence by cannabis and other illicit drug use; cohort comparisons.
Table 2: Association between Under-21 MLDA exposure and marijuana use outcomes (aORs and CIs).
Table 3: Association between Under-21 MLDA exposure and illicit drugs other than marijuana (aORs and CIs).
Table 4: Stratified by cohort and sex for marijuana and illicit drug use outcomes.
Table 5: Analysis restricted to participants with high school education or less.
Figure 1: Timeline of MLDA policy changes across states (1967–1992).
Acknowledgments: Funding and contributions; notes on potential conflicts of interest (e.g., Dr. Bierut’s patent).