2011- Notes on Alcohol Policy, Social Context, and Infant Health: The Impact of Minimum Legal Drinking Age

Abstract

  • Objective: Examine distal (secondary) effects of the minimum legal drinking age (MLDA) on infant health indicators and whether shifts in drinking behaviors or birth composition contributed to these effects.
  • Context: MLDA was raised in the U.S. in the late 1980s to reduce intoxication-related injuries, especially motor vehicle accidents.
  • Data & design: State- and year-fixed-effects models using two population-based datasets merged for 1985–2002: Natality Detailed Files (birth outcomes) and BRFSS (drinking behaviors). Analysis focused on potential mothers who were 14 years old when MLDA varied (i.e., MLDA at age 14).
  • Key findings: A MLDA of 18 years (at the time of the mother’s 14th birthday) increased adverse infant health outcomes (LBW, low Apgar, prematurity), with stronger effects for infants of black mothers. A younger MLDA was also associated with higher rates of black female school drop-outs and binge drinking patterns; conversely, older MLDA laws at age 14 reduced binge drinking among black women. Overall, elevating MLDA had longer-term positive effects on infant health and social/behavioral indicators, particularly for black families.
  • Implications: Stricter early-life alcohol access environments may improve subsequent infant health across generations by shaping drinking behaviors and broader social contexts.

Introduction

  • Policy context and rationale
    • Public health policies can have unintended consequences beyond their direct targets.
    • Prohibition (18th Amendment) showed policy can backfire; similarly, changing MLDA creates natural experiments to study downstream effects on infant health and social outcomes.
    • In the U.S., MLDA was raised to 21 by the Uniform Drinking Age Act (1984) to reduce motor vehicle accidents; by 1990 all states complied.
  • Relevance to infant health and behavior
    • Prior literature shows older MLDA reduces youth alcohol use, drunk driving, and some adverse social outcomes (e.g., high school dropouts, teen births, STDs).
    • A study noted that MLDA of 18 at conception year was linked to worse birth outcomes; this study extends to long-run effects via the environment experienced by girls entering childbearing years.
  • Key research questions
    1) Is the MLDA environment at age 14 (early teenage years) related to infant health outcomes later when those women give birth?
    2) Do shifts in birth composition or in drinking behaviors during adolescence explain any observed birth outcomes improvements?
    3) Are there racial differences in these associations?
  • Conceptual framework
    • MLDA is used as an indicator of alcohol availability in the environment in which young women pass through their teen years.
    • Hypothesis: Stricter MLDA (older age) → stricter alcohol environment → potential long-run improvements in birth outcomes and related behaviors.

Data

  • Natality Detailed Files (Vital Statistics Natality), 1970–1992
    • Contains all birth certificates; provides infant health indicators (birth weight, 5-minute Apgar, gestation length) and mothers’ characteristics.
    • Sample restriction: mothers ≤21 at delivery; 14-year-olds’ MLDA exposure identified through year of birth and residence.
    • Outcome limitations: some infants’ health indicators missing; dropping rules applied (drop if health data missing; otherwise keep if 1–2 indicators missing).
    • Descriptive findings (Table 1, Panel A):
    • LBW (<2500 g): 10.7%
    • ELBW (<1500 g): 2%
    • Apgar <7: 3%
    • Premature (<37 weeks): 15%
    • Black infants had higher health burdens than white infants (LBW 13.5% vs 7.9%; ELBW 2.7% vs 1.5%; Apgar <7 3.5% vs 2.4%; prematurity 18% vs 11%).
    • Marital status and educational differences: higher marriage rate among white mothers (35%) than black mothers (12%).
    • Drinking during pregnancy in Natality data: <2% reported drinking; this is lower than national estimates (10–12%), likely due to data capture method and stigma.
    • Limitation acknowledged: Natality data inadequately captures drinking behavior; BRFSS used for drinking behavior data.
  • Behavioral Risk Factor Surveillance System (BRFSS), 1985–2002
    • Cross-sectional, representative of non-institutionalized population.
    • Drinking behaviors queried: drinking prevalence in the previous month and binge drinking (≥5 drinks per occasion) in the previous month.
    • Summary (Table 1, Panel B):
    • 51% of young women drank in the previous month; average 2.7 drinks per occasion.
    • White women more likely to drink (53%) than black women (46%).
    • Among drinkers, white women: 21% binge; black women: 25% binge.
  • MLDA policy data
    • Source: Distilled Spirits Council of the U.S. (DISCUS).
    • MLDA is defined as the age in effect for the estimated year of a woman’s 14th birthday, allowing for year-to-year variation within states.
    • Coding rule when policy differs by beverage type (beer vs liquor): if beer allowed at 18 but liquor at 21, MLDA coded as 18.
  • Sample construction and key variables
    • Unit of analysis: birth events linked to MLDA environment at mother’s age 14.
    • Main exposure variable: state-specific MLDA when the mother was 14 (18, 19, 20, or 21).
    • Covariates: race, age, marital status, education; smoking during pregnancy; real per capita income (state-level macro condition); real beer taxes (federal + state).
    • Additional controls: other alcohol policies to reduce omitted variable bias.

Specification

  • Econometric model
    • For each birth outcome Yit (e.g., LBW, low Apgar, premature) and for drinking outcomes (drinking and binge drinking among mothers), estimate: Y{it} = eta0 + eta1 MLDA^{14}{s,t} + oldsymbol{eta}X' oldsymbol{X}{i} + us + aut + oldsymbol{ u}{s,t} + ext{error}_{it}
    • Where:
    • MLDA^{14}_{s,t} is the MLDA policy in effect for the mother at age 14 in state s and year t (values 18, 19, 20, or 21).

    • us are state fixed effects; aut are year fixed effects.
    • oldsymbol{X}_{i} includes race, age, marital status, education, smoking during pregnancy.
    • Real per capita income and real beer taxes are included as additional controls to capture macroeconomic and tax policy variation.
    • The model is estimated with OLS for continuous outcomes and Probit for dichotomous outcomes; marginal effects are reported for probit models.
    • The authors note potential endogeneity concerns and acknowledge that causality cannot be fully established due to data limitations; they discuss potential two-stage approaches but do not implement them here.
  • Key notes on interpretation
    • The coefficient on MLDA^{14}_{s,t} captures the long-run association between the 14-year-old environment and later birth outcomes, net of fixed effects and observed covariates.
    • Marginal effects in probit models are reported for dichotomous outcomes (e.g., LBW, Apgar <7, prematurity; drinking, binge drinking).

Summary Statistics (Table 1)

  • Natality Data (All Mothers; White vs Black)
    • % Born < 2500 g: overall 10.68%; White 7.54%; Black 13.54%.
    • % Born < 1500 g: overall 2.07%; White 3.87%; Black 1.456? (text shows 1.456 but likely 1.52?); Black higher than White in ELBW? Text variance noted.
    • % Apgar <7: overall 2.96%; White 6.14%; Black 3.51%? (text shows 3.51 for Black; 7.32% for prematurity in White? The table text is slightly garbled; use described gaps: Black higher LBW/ELBW and prematurity; Apgar <7 higher in Black.)
    • % Premature: overall 14.87% (White 9, Black 18.77% in text discrepancy; interpret as Black higher).
    • % Mothers married: White 23.93%; Black 12.10% (in table, White higher marriage rate).
    • % Mothers high school dropouts: White 59.30%; Black 60.66% (similar); % with some college or higher higher among Whites.
    • % Mothers smoke during pregnancy: White 34.53%; Black 27.77% (text shows White higher smoking; Black lower in pregnancy? The BRFSS smoking capture notes issues—summary indicates higher smoking among groups in Natality data but BRFSS used for drinking behavior.)
    • % Mothers drank during pregnancy: White ~1.77%; Black ~1.92%; small proportions; data limitations discussed.
    • N (Natality): 11,051 White mothers; 7,448 White? 4,603 Black? (Table 1 shows N across groups; rely on text).
  • BRFSS Data (Panel B)
    • % drink in past month: 51.11% overall; White 52.65%; Black 45.72%.
    • # drink per time: 2.71 overall; White 2.73; Black 2.67.
    • % binge among drinkers: White 21.12%; Black 25.63%.
    • % high school dropouts / % high school graduates / % some college / % college or more: various (text provides percentages by race), showing educational composition differences between White and Black respondents in BRFSS.
    • N BRFSS: ~284,382 (All); ~245,718 White; ~38,619 Black; (numbers per panel shown in Table 1).

Table 2. The effects of MLDA on birth outcomes

  • Outcomes analyzed: Low Birth Weight (LBW), Low Apgar Scores, Premature Birth
  • Models: Columns 1–6 include different specifications; all models include state and year fixed effects and covariates.
  • All Mothers (Panels A)
    • MLDA18: LBW +0.0014; Low Apgar +0.0112; Premature +0.0002*; p-values: LBW<0.0001; Apgar<0.0001; Premature=0.0514.
    • MLDA19: LBW -0.0016; Apgar -0.0180; Premature -0.0002*; p-values: LBW=0.0020; Apgar<0.0001; Premature=0.0511.
    • MLDA20: LBW -0.0005; Apgar -0.0103***; Premature 0.0001; p-values: LBW=0.2174; Apgar<0.0001; Premature=0.2145.
    • MLDA21: LBW -0.0024; Apgar -0.0182; Premature -0.0004***; p-values: LBW<0.0001; Apgar<0.0001; Premature=0.0002.
    • N (Across models): 11,051; 11,051; 5,792; 5,792; 9,900; 9,900.
  • White Mothers (Panels B)
    • MLDA18, MLDA19, MLDA20, MLDA21 effects on LBW are not significant; Apgar and Premature show varying small effects; overall no robust LBW impact for Whites; some small negative Apgar effects appear under certain MLDA categories with small p-values.
    • Key note: White mothers show no LBW changes with MLDA18; some small negative Apgar effects under MLDA19; similar patterns for Premature with small magnitudes.
  • Black Mothers (Panels C)
    • MLDA18: LBW +0.0181; Apgar +0.0372; Premature +0.0061** (significance two asterisks).
    • MLDA19: LBW -0.0202; Apgar -0.0257***; Premature -0.0056 (some non-significant p-values).
    • MLDA20: LBW -0.0073; Apgar -0.0368***; Premature -0.0027 (some p-values significant for Apgar).
    • MLDA21: LBW -0.0283; Apgar -0.0498; Premature -0.0095***; robust significance across outcomes.
    • N: 4,603 (overall); 4,603 White; 2,321 Black; etc. (per panel).
  • Notes
    • Marginal effects are presented; models include maternal education, age, marital status, smoking during pregnancy, real income per capita, real beer taxes; state and year fixed effects.
    • Sample restricted to mothers younger than 21 years old.
    • Significance: * p<0.10, ** p<0.05, *** p<0.01.

Table 3. The effects of MLDA on compositional change of births

  • Focus: Compositional shifts in birth characteristics by age 15–17 and % High School Dropouts
  • All Mothers (Panel A)
    • MLDA18: 0.0069* for % age 15–17; p=0.1682 (non-robust); 0.0004 for % high school dropouts; p=0.8501.
    • MLDA19: -0.0028 (age 15–17); 0.0016 for high school dropouts; p-values 0.6404 and 0.5252 respectively.
    • MLDA20: -0.0253 for age 15–17; 0.0005 for high school dropouts; p=0.0046 and 0.8892.
    • MLDA21: -0.0053 for age 15–17; -0.0028 for high school dropouts; p=0.9286 and 0.2477.
    • N = 11,051.
  • White Mothers (Panel B)
    • MLDA18: -0.0015 age 15–17; -0.0049 dropouts; p=0.8191 and 0.0471.
    • MLDA19: 0.0078 age 15–17; 0.0048 dropouts; p=0.3128 and 0.0937.
    • MLDA20: -0.0102 age 15–17; 0.0057 dropouts; p=0.3519 and 0.1606.
    • MLDA21: -0.0012 age 15–17; 0.0048 dropouts; p=0.8736 and 0.0937.
    • N = 7,448.
  • Black Mothers (Panel C)
    • MLDA18: 0.0160** age 15–17; 0.0056* dropouts; p=0.0336 and 0.0958.
    • MLDA19: -0.0139 age 15–17; -0.0014 dropouts; p=0.1161 and 0.6405.
    • MLDA20: -0.0047 age 15–17; -0.0042 dropouts; p=0.713 and 0.4629.
    • MLDA21: -0.0093 age 15–17; -0.0106*** dropouts; p=0.2856 and 0.0108.
    • N = 4,603 (across panels).
  • Notes
    • Marginal effects are reported; regressions control for education, age, marital status, smoking during pregnancy, real income per capita, real beer taxes; state and year fixed effects.
    • Interpretation: Compositional shifts (e.g., more very young mothers, more/less HS dropouts) occur, particularly among Black mothers, but explain only a modest portion of birth outcome changes.
    • Found minimal compositional shifts among White mothers.

Table 4. The MLDA and the drinking behaviors

  • Outcomes: Drinking (0/1) and Binge Drinking (0/1)
  • All Mothers (Panel A)
    • MLDA18: Drinking 0.0015; Binge -0.0095; p-values 0.6048 and 0.0102.
    • MLDA19: Drinking -0.0057*; Binge -0.0003; p-values 0.0880 and 0.9364.
    • MLDA20: Drinking -0.0042; Binge 0.0404; p-values 0.4206 and 0.6541.
    • MLDA21: Drinking -0.00056; Binge 0.0149**; p-values 0.1543 and 0.0039.
    • N = 284,382 (drinking); 154,080 (binge).
  • White Mothers (Panel B)
    • MLDA18: Drinking 0; Binge -0.007; p-values >0.1 and 0.3846.
    • MLDA19: Drinking -0.0046; Binge -0.0021; p-values 0.1945 and 0.6486.
    • MLDA20: Drinking -0.0057; Binge 0.056; p-values 0.2928 and 0.4037.
    • MLDA21: Drinking -0.0085***; Binge 0.0257; p-values 0.0421 and 0.0974.
    • N = 245,718 (drinking); 140,328 (binge).
  • Black Mothers (Panel C)
    • MLDA18: Drinking 0.0015; Binge 0.0219***; p-values 0.8421 and 0.0200.
    • MLDA19: Drinking -0.0282; Binge -0.0046; p-values 0.0482 and 0.9603.
    • MLDA20: Drinking -0.0823; Binge -0.0659; p-values 0.0006 and <0.0001.
    • MLDA21: Drinking -0.0234; Binge -0.0656*; p-values 0.1256 and <0.0001.
    • N = 38,619 (drinking); 13,752 (binge).
  • Notes
    • The results indicate differential effects of MLDA on drinking and binge drinking by race.
    • Black women show larger reductions in drinking and binge drinking with older MLDA (especially MLDA=20 or 21) compared to White women, while White women show smaller or more modest shifts.
    • Overall interpretation: Stricter early-life alcohol availability environments appear to influence later drinking patterns, with racialized differences in the magnitude of effects.

Primary results and interpretation (Textual synthesis)

  • Primary results (Table 2, main findings)
    • For all mothers, an MLDA of 18 at age 14 increases LBW by 0.14 percentage points, while MLDA-19 lowers LBW by 0.16 points and MLDA-21 lowers LBW by 0.24 points (LR-χ² significant for each: p<0.001 for 18 and 21; 0.001 for 19).
    • Race-specific effects: No LBW change for White infants; Black infants show substantial sensitivity: MLDA-18 raises LBW by ~1.8 percentage points; MLDA-21 reduces LBW by ~2.8 percentage points (both p<0.001).
    • Low Apgar scores: A more lenient MLDA (18) raises risk of Apgar <7 by ~1.1 percentage points for all, with larger effects among Black infants (~3.7 percentage points). Older MLDAs (19, 20, 21) reduce Apgar <7 by about 2 percentage points, with stronger racial disparities (White ~0.36 percentage points, Black ~3.7 percentage points).
    • Premature births: MLDA-18 tends to increase prematurity slightly (0.02 percentage points) for the overall sample; Black infants show larger effects (0.6 percentage points). MLDA-21 predicts the largest decrease in prematurity for Black mothers (~1 percentage point).
    • Overall takeaway: Stricter MLDA during adolescence is associated with better birth outcomes later, particularly for infants of Black mothers.
  • Pathways (Tables 3 and 4; secondary results)
    • Pathway 1: Compositional changes in births (Table 3)
    • Exposure to MLDA-18 at age 14 increases the likelihood of very young mothers (ages 15–17) among Black mothers by about 1.6 percentage points (p<0.001).
    • Lenient MLDA environments (lower MLDA) increase the proportion of high school dropouts among Black mothers by about 0.5 percentage points; MLDA-21 reduces HS education attainment by about 1.1 percentage points in Black mothers.
    • White mothers show little to no compositional shifts attributable to MLDA variations.
    • Conclusion: Compositional shifts contribute modestly to observed birth outcome improvements, mainly among Black mothers.
    • Pathway 2: Behavioral changes (BRFSS-based; Table 4)
    • Across all mothers, younger MLDA (18) increases the probability of drinking later (0.15 percentage points) with drinking patterns influenced by MLDA-19 and MLDA-21; stepwise changes in MLDA have limited impact on White women’s drinking, but MLDA-21 reduces drinking probability for Black women by 2–8 percentage points and reduces binge drinking more substantially for Blacks.
    • For Black women, MLDA20/21 reduces drinking by ~8% points and binge drinking substantially; for White women, MLDA21 reduces drinking modestly and has no strong effect on binge drinking.
    • Interpretation: Changes in alcohol availability during adolescence appear to influence formation of habitual drinking, with distinct racial patterns; older MLDA reduces binge drinking more for Black women than for White women.

Conclusions

  • Main conclusions
    • An older MLDA at age 14 (i.e., stricter alcohol access during adolescence) is associated with improved infant health outcomes later, particularly for infants born to Black mothers (lower LBW, lower prevalence of low Apgar scores, and fewer prematures).
    • The observed improvements appear to be mediated through two pathways: (i) compositional shifts in birth cohorts (especially among Black mothers, with modest effects on very young mothers and HS dropouts), and (ii) behavioral shifts in drinking patterns (notably, reduced binge drinking among Black women and reduced overall drinking among Black women with higher MLDA; weaker effects among White women).
    • The study provides evidence that alcohol policy can have cross-generational health benefits by shaping the social context of adolescence and early childbearing years.
  • Policy implications
    • Strengthening MLDA during adolescence may yield long-run benefits for infant health and may reduce alcohol-related risk behaviors in subsequent generations, especially for Black communities.
    • The findings support public health arguments against lowering the MLDA and align with broader evidence that alcohol policy can influence social contexts (e.g., education trends, teen pregnancy, and risky sexual behavior).
  • Limitations and caveats
    • Causality cannot be firmly established due to data limitations; a two-stage least squares approach could help if suitable instruments were available.
    • Potential omitted variable bias remains (e.g., Medicaid expansion for pregnant women, insurance status, prenatal care utilization) because some variables were unavailable in Natality data.
    • Alcohol consumption during pregnancy in Natality data underreports use; BRFSS provides better measurement of drinking behaviors but is not linked at the individual level to Natality outcomes.
    • The study uses MLDA at age 14 as a proxy for early alcohol environment; actual individual-level alcohol exposure during pregnancy is not directly observed.
  • Real-world relevance and ethical considerations
    • The research underscores the broader social and ethical implications of policy changes that affect adolescents’ environments, not just immediate outcomes (e.g., crime, drinking in adolescence, and risky sexual behavior).
    • It highlights potential racial disparities in policy effects and the need for equity-focused evaluation when implementing public health policies.
  • Final synthesis
    • Policy-induced changes in alcohol availability during adolescence can generate lasting health benefits for the next generation, particularly for Black families, through both changes in birth cohort composition and shifts in drinking behaviors.
    • This study contributes to the broader understanding of how public health policy can create cascading effects across generations and how race intersects with these effects.

Equations and key formulas

  • Long-run effect specification (conceptual form)
    • For a birth outcome Y (e.g., LBW, Apgar, Premature) or a drinking behavior D (0/1 indicators):
      Y{it} ext{ or } D{it} = eta0 + eta1 ext{MLDA}^{14}{s,t} + oldsymbol{eta}X' oldsymbol{X}{i} + us + aut + oldsymbol{ u}{s,t} + ext{error}_{it}
  • Model types
    n- OLS used for continuous outcomes
    n- Probit used for dichotomous outcomes; marginal effects reported in probit models

Connections to broader themes

  • Natural experiments and fixed effects
    • The study leverages variation in MLDA across states and over time as a natural experiment to infer associations with later outcomes, controlling for fixed state and year characteristics.
  • Life-course and intergenerational health effects
    • The analysis aligns with life-course epidemiology, examining how environmental exposures during adolescence can shape health trajectories of the next generation.
  • Health equity and policy design
    • The differential effects by race highlight the importance of considering equity in policy evaluation and the need to tailor interventions to address disparate social contexts.

Key terms to remember

  • MLDA: Minimum Legal Drinking Age
  • LBW: Low Birth Weight (< 2500 g)
  • ELBW: Extremely Low Birth Weight (< 1500 g)
  • Apgar score: 5-minute assessment of newborn health; <7 considered low
  • BRFSS: Behavioral Risk Factor Surveillance System
  • Probit model and marginal effects: used for binary outcomes
  • Fixed effects: state and year controls to account for unobserved heterogeneity
  • Compositional shift vs. behavioral shift: changes in birth cohort composition vs. changes in individual drinking behavior as mechanisms

Sources and data limitations (at a glance)

  • Data limitations acknowledged in the study:
    • Natality data underreports drinking during pregnancy; BRFSS used for drinking behavior indicators.
    • Causality difficult to establish; potential omitted variable bias (e.g., Medicaid expansion, prenatal care utilization).
    • Incomplete alignment between infant health outcomes and drinking behavior at the individual level.
  • Data sources used in the analysis:
    • Natality Detailed Files (1970–1992)
    • BRFSS (1985–2002)
    • DISCUS MLDA policy data
    • Real per capita income and real beer taxes as controls