Associations between state-level general population alcohol policies and drinking outcomes among women of reproductive age: Results from 1984–2020 National Alcohol Surveys

Abstract

Background:

Policies specific to alcohol use during pregnancy have not been found to reduce risks related to alcohol use during pregnancy. In contrast, general population alcohol policies are protective for the general population. Here we assessed whether US state-level general population alcohol policies are related to drinking outcomes among women of reproductive age.

Methods:

Secondary analyses of 1984–2020 National Alcohol Survey data (N=13,555 women ≤44 years old). State-level policy exposures were government control of liquor retail sales, heavy beer at gas stations, heavy beer at grocery stores, liquor at grocery stores, Sunday off-premise liquor sales, and Blood Alcohol Concentration (BAC) driving limits (no law, .10 limit, .05-.08 limit). Outcomes were past 12-month number of drinks, ≥5 drink days, ≥8 drink days, and any DSM-IV alcohol abuse/dependence symptoms. Regressions adjusted for individual and state-level controls, clustering by state, and included fixed effects for survey month and year.

Results:

Allowing Sunday off-premise liquor sales vs. not was related to having 1.20 times as many drinks (95% CI: 1.01, 1.42), 1.41 times as many ≥5 drink days (95% CI: 1.08, 1.85), and 1.91 times as many ≥8 drink days (95% CI: 1.28, 2.83). BAC limits of 0.05–0.08 for driving vs. no BAC limit was related to 0.51 times fewer drinks (95% CI: 0.27, 0.96), 0.28 times fewer days with ≥5 drinks (95% CI: 0.10, 0.75), and 0.20 times fewer days with ≥8 drinks (95% CI: 0.08, 0.47).

Conclusions:

US state-level policies prohibiting Sunday off-premise liquor sales and BAC limits of 0.05–0.08 for driving are related to less past 12-month overall and heavy drinking among women 18–44 years old.

INTRODUCTION

Policies regarding alcohol use during pregnancy

The prevalence of alcohol use during pregnancy in the US has remained at 10–15% since the 1990s (Centers for Disease Control and Prevention, 1997; Denny et al., 2019) despite the dramatic increase in the number of state-level policies targeting alcohol use during pregnancy over the same period (Roberts et al., 2017). Policies regarding alcohol use during pregnancy are intended to reduce harms to infants related to pregnant people’s drinking. However, research has found that while state-level policies targeting alcohol use during pregnancy may be related to drinking among pregnant people (Roberts et al., 2019), they are also related to increased risks of low birthweight and premature births, infant maltreatment, and infant morbidities related to maternal alcohol consumption (Subbaraman et al., 2018; Subbaraman and Roberts, 2019; Roberts et al., 2023).

A legal epidemiology study using 1985–2016 Behavioral Risk Factor Surveillance System (BRFSS) data found that two pregnancy-specific alcohol policies (1. requiring notices warning of harm from consumption during pregnancy be posted in licensed settings where alcohol is sold and 2. legally defining alcohol use during pregnancy as child abuse or neglect) were associated with less binge and/or heavy drinking during pregnancy (Roberts et al., 2019). However, one policy (mandating priority access to substance use treatment for pregnant people and women with children) was associated with higher odds of any drinking during pregnancy (Roberts et al., 2019). Furthermore, each of these policies is associated with higher odds of low birthweight and preterm births, less prenatal care utilization, and significant costs related to low birthweight and preterm births, suggesting that pregnancy-specific policies ultimately do not have their intended effects (Subbaraman et al., 2018; Subbaraman and Roberts, 2019). Thus, recommendations to reduce alcohol use during pregnancy and improve birth outcomes through state-level policies targeting alcohol use include examining broader general population alcohol policies rather than continuing with the pregnancy-specific policies currently in effect (Subbaraman et al., 2018).

General population alcohol policy associations with birth outcomes

Indeed, general population alcohol policies that are previous research has found to reduce population-level consumption may be associated with better birth outcomes, e.g., fewer preterm and premature births. Studies of the minimum legal drinking age (MLDA) have found that a lower MLDA is related to higher prevalence of low birthweight and premature births overall (Fertig and Watson, 2009; Zhang and Caine, 2011) as well as more binge drinking among Black women in particular (Zhang and Caine, 2011). Another study using birth certificate data from >150,000,000 US births reported that both government control over wine retail sales and reduced general-population per capita alcohol consumption were associated with improved birth outcomes (Subbaraman et al., 2018). However, whether general population alcohol policies relate to drinking among pregnant people or those of reproductive age who may become pregnant remains an open question.

Differential associations between general population alcohol policies and drinking by gender

Stricter general population alcohol policies like higher alcohol taxes, fewer store hours, limits on the number and types of alcohol outlets, and government control of alcohol sales reduce drinking in the general population (Babor, 2010; Gruenewald, 2011; Nelson et al., 2013), but associations across demographic subgroups may vary (Subbaraman et al., 2020). For example, results from both econometric finite mixture models (Ayyagari et al., 2009) and causal, deterministic, epidemiological models (Meier et al., 2010) support that women are more sensitive to alcohol prices than men. Women also show stronger preference for off-premise drinking (e.g., drinking at home) and wine, while men show stronger preference for on-premise drinking (e.g., at bars) and beer; these differential preferences imply that women might react differently to specific alcohol policies than men (Meier et al., 2010). Supporting this, a study using survey data found that decreased density of off-premise outlets was related to less drinking among white and Hispanic women while government control of liquor sales and decreased bar density was related to reduced drinking among white men specifically (Subbaraman et al., 2020). However, the same study found that, contrary to expectation, higher beer taxes were related to reduced drinking among women of all race/ethnicities but not among men (Subbaraman et al., 2020). Notably, no prior study has examined relationships between general population alcohol policies and drinking among pregnant people or women of reproductive age in particular. Examining women of reproductive age makes sense because people who drink during pregnancy continue drinking patterns from before they become pregnant (Ethen et al., 2009), and historically, most people who become pregnant are cisgender women.

Though Sunday alcohol sales have not been studied as much alcohol taxes or government control of sales, the evidence supports that removing restrictions on Sunday sales is related to worse public health outcomes, e.g., more drinking (Carpenter and Eisenberg, 2009; Yörük, 2014a), crime (Yörük and Lee, 2018), and traffic accidents (McMillan et al., 2007; McMillan and Lapham, 2006). In the US, removing state-level Sunday sales restrictions was associated with increased state-level per capita consumption in Delaware, Pennsylvania and New Mexico but not in Massachusetts and Rhode Island (Yörük, 2014a), and in Canada, province-level Sunday sales are associated with a significant increase in drinking on Sundays (Carpenter and Eisenberg, 2009). Since Sunday sales laws generally affect off-premise sales, restricting Sunday sales might be particularly relevant for women (Meier et al., 2010). However, Sunday sales laws have not been studied within demographic subgroups to our knowledge. Removing restrictions on Sunday sales continues to be a hot political topic in the US, with ballot measures expanding local Sunday sales options having passed in at least four states (Arkansas, Georgia, Ohio, South Carolina) in the November 2022 election.

Laws regarding blood alcohol concentration (BAC) limits while driving are another hot political topic in the US. Most recently, the state of Utah lowered its BAC limit for driving from .08 to .05 in 2018, with similar legislation proposed but not passed in Delaware, Washington, and Hawaii (Morain and Largent, 2019). In the US, lower BAC limits are unequivocally related to fewer alcohol-related traffic accidents (Fell and Voas, 2006; Wagenaar et al., 2001), but no study that we know of has examined whether laws regarding BAC limits for driving relate to drinking in the general population or among women specifically. A study using comparative interrupted time series, one of the strongest study designs, found that lowering the drink-drive limit from .08 to .05 BAC in Scotland had no impact on traffic accidents and a small reduction in per capita consumption from on-premise sales specifically (Haghpanahan et al., 2019). Supporting this, Scottish bar owners and managers in a qualitative study reported fewer people drinking after work and more people leaving premises earlier on weekdays after the drink-drive BAC limit was lowered (Sumpter et al., 2020). Given recent legislative changes and public health advocacy efforts to lower BAC limits in other states (Morain and Largent, 2019), data on how BAC limit laws affect drinking in the US are needed. Focusing specifically on women of reproductive age can help provide evidence as to whether these general population alcohol policies may be relevant for another key public health outcome of interest, harms related to pregnant people’s alcohol consumption.

Study aim and rationale

The aim of this study is to examine how state-level policies regarding government control of liquor sales, types of alcohol outlets, Sunday sales, and BAC limits for driving are related to drinking outcomes among women of reproductive age between 1984–2020. We focus on these particular policies because 1) prior results suggest that general population alcohol policies may be more protective than pregnancy-specific policies for both birth outcomes and drinking outcomes; 2) no study has examined whether any of these general population alcohol policies relate to drinking among women of reproductive age specifically; 3) these particular general population alcohol policies vary substantially over time and across states, and have the most complete state/year-level data available for the time period under study (e.g., tax data are not available for all states; store hours data are not available for the whole time period); 4) no study has examined whether BAC laws relate to drinking in the US; and 5) BAC laws and Sunday sales laws are policies that are currently being debated in the US.

MATERIALS AND METHODS

Sample

Outcome data came from the 1984–2020 National Alcohol Surveys (NAS, N=51,433 total respondents pooled across years), a US representative survey of the US population age 18 and older. The NAS has been conducted approximately every five years since the 1960s with considerable standardization of measurement methods since 1979; measures for the outcomes of interest are available for 1984–2020. The length of the study period is necessary for sufficient variation in all policies of interest.

Sampling methods for the NAS have changed over time from a multi-stage cluster design (1984–1995) to random digit dialed (2000–2020). An address-based sample (ABS) and nonprobability sample from a pre-recruited web panel were added in 2020. Although not unique in conducting repeated surveys including alcohol questions, the NAS has the longest time frame with commensurate, detailed alcohol use pattern and problem measures in the US. Interviews were conducted in-person (1984–1995) via telephone (2000–2020) and on-line (for the ABS and panel samples in 2020) in English or Spanish. Black and Hispanic populations were oversampled for all years except 1990. For the 2020 NAS, the cooperation rate for the combined telephone and ABS samples was 42.2%. Our analyses focus on N=13,555 women of reproductive age, defined as 44 years old or less, following US Centers for Disease Control and Prevention practices (Centers for Disease Control and Prevention, 1997; Denny et al., 2019).

Outcomes

We examined past 12-month number of drinks, number of ≥5 drink days, number of ≥8 drink days, and any DSM-IV alcohol “abuse” or “dependence” symptoms as outcomes. Past 12-month number of drinks were measured using the graduated frequency series, which assesses frequencies of drinking in a graduated series of quantity intervals (Greenfield et al., 2009). The graduated frequency measures provide the number of days using 5–7, 8–11 and ≥12 drinks; these were used to calculate the number of ≥5 drink days and ≥8 drink days for 1990–2020. A single measure of the frequency of days having ≥8 drinks was used for 1984. Having any DSM-IV symptoms (yes/no) was assessed from whether the respondent endorsed having any DSM-IV symptoms out of the 11 DSM-IV “abuse” or “dependence” criteria in the past 12 months. We used any DSM-IV abuse/dependence items as a proxy for alcohol problems as the number of those meeting abuse/dependence criteria was too low for outcome analyses. Data for past 12-month number of drinks and number of ≥5 drink days were available for 1990–2020; data for number of ≥8 drink days, and DSM-IV symptoms were available for 1984–2020.

State/year-level alcohol policies

The main exposure variables were time-varying state-level indicators regarding whether the respondent’s state had the following general population policies in effect at the time of the respondent’s interview year: government control of liquor wholesale or retail, heavy beer sales permitted at gas stations, heavy beer sales permitted at grocery stores, and liquor sales allowed at grocery stores; Sunday off-premise liquor sales; Blood Alcohol Concentration (BAC) driving laws (three categories: no law, .10 limit, .05-.08 limit). Besides BAC laws, all policies were coded dichotomously as yes/no.

Data for state government control of liquor sales at the wholesale or retail level came from the Liquor Handbooks for 1984–1997 (Jobson’s Liquor Handbook, 1992; Adams Liquor Handbook, 1997) and from NIAAA’s Alcohol Policy Information System (APIS) for 1998–2020 (“APIS Home | APIS - Alcohol Policy Information System,” n.d.; Bloss, 2011). Data on whether heavy beer (>3.2% alcohol by volume) or liquor sales are permitted at gas stations or grocery stores came from the Liquor Handbooks for 1984–2014 (Jobson’s Liquor Handbook, 1992; Institutes and LaVilla, 2009) and the National Alcoholic Beverage Control Association (NABCA) for 2015–2020 (“Survey Database | National Alcohol Beverage Control Association,” n.d.). Data for off-premise Sunday liquor sales came from the Distilled Spirits Council of the United States (Council, 2004) for 1984–2003 and from NABCA for 2004–2020. BAC law data came from the National Highway Traffic Safety Administration (“National Highway Traffic Safety Administration (NHTSA),” 2000), APIS, and original legal research. For all policies, states with local variability (e.g., local options) were coded based on the policy for the most populous county in that state.

Statistical analyses

First, we used bivariate statistics to describe trends in demographics and in the prevalence of respondents living in states with each policy of interest across survey years. We then used negative binomial regression to test whether policies are related to count outcomes (past 12-month number of drinks, number of ≥5 drink days, number of ≥8 drink days) and logistic regression to test whether policies are related to any past 12-month DSM-IV alcohol abuse or dependence symptoms. For each outcome, the first set of regression models included all policies simultaneously with a fixed effect for survey year and adjustment for clustering by state. The fixed effect for survey year adjusts for secular trends in alcohol use behaviors. The second set of regressions for each outcome additionally adjusted for respondent-level age, race/ethnicity, marital status, education, employment, income, interview month; state/year-level poverty, unemployment (census.gov); and state-level consumption patterns (categorized as high, moderate, low (Kerr, 2010)). We did not adjust for per capita consumption in our main analyses because we conceptualize it as on the causal pathway between policy exposures and drinking outcomes of interest; therefore, adjustment for per capita consumption may attenuate estimates. Still, we conducted sensitivity analyses including it because, conceptually, per capita consumption could be a reason for policy adoption. All bivariate and multivariable analyses were survey-weighted to account for probability of selection into the sample and performed in Stata v17 (StataCorp, 2021).

RESULTS

Descriptives

Figure 1 illustrates the number of US states with each general population alcohol policy for the study period, 1984–2020. Table 1 describes demographics for the 13,555 women ≤ 44 years old by NAS survey year. The distributions of race/ethnicity, marital status, and education all changed significantly over time, with decreases in the prevalence of respondents who are Non-Hispanic White, married/cohabitating, and/or with less than a high school education. The distribution of the sample living in states with each policy also changed significantly over time, with a higher prevalence of respondents living in states with looser policies over time. Table 1 also shows past 12-month average number of drinks, number of days with ≥5 drinks, number of days with ≥8 drinks, and prevalence of having any DSM-IV alcohol abuse or dependence symptoms by survey year. Only the average number of drinks consumed in the past 12 months changed significantly over the study period, increasing from 134.9 in 1990 to 232.6 in 2020 (P<0.05 in bivariate test for difference between years).

Figure 1.

Number of US states with each general population alcohol policy during study period, 1984–2020

Table 1.

Survey-weighted demographic statistics for women ≤ 44 years old, 1984–2020 National Alcohol Surveys (N = 13,555)

Year
	1984	1990	1995	2000	2005	2010	2015	2020
% unless noted	n=1,896	n=656	n=1,609	n=2,116	n=1,586	n=1,483	n=1,308	n=2,901	P a
Age M(SE)	30.0 (.31)	30.7 (.36)	31.9 (.32)	31.7 (.20)	31.6 (.26)	31.4 (.29)	31.5 (.32)	31.3 (.18)	0.01
Race
Non-Hispanic White	78.2	69.4	69.9	67.6	63.3	60.9	56.5	56.5	0.01
Non-Hispanic Black	12.3	15.0	13.9	14.3	12.7	13.7	14.5	15.6
Hispanic	6.7	11.0	11.2	13.4	16.5	17.1	19.9	19.5
Other/missing	2.8	4.6	5.0	4.8	7.5	8.3	9.1	8.4
Marital status
Married/cohabitating	65.8	60.6	63.5	62.2	64.5	60.1	59.5	58.8	0.01
Not married/cohabitating	34.2	39.5	36.5	37.4	35.5	39.9	40.5	41.2
Education
Less than high school	16.6	14.1	14.0	11.7	11.3	11.3	10.0	6.0	0.01
High school	40.6	42.5	40.0	29.1	27.4	26.2	23.2	28.2
Some college	26.2	24.5	26.5	32.5	30.3	31.6	34.5	35.7
College or more	16.6	19.0	19.5	26.7	31.0	30.9	32.3	30.2
Employment
Employed full- or part-time	64.7	69.3	71.1	70.7	65.2	60.0	68.3	66.5	0.41
Not employed full- or part-time	35.3	30.7	29.0	29.3	34.8	40.0	31.7	33.5
% Respondents living in state with
Government control of liquor retailc	30.8	26.0	29.1	31.8	28.8	27.8	27.4	27.4	0.10
Heavy beer sold at gas stations	89.4	87.1	90.4	89.5	90.4	90.5	88.8	92.0	0.01
Heavy beer sold at grocery stores	93.5	94.2	91.7	93.0	94.4	94.1	94.5	93.1	0.05
Liquor sold at grocery stores	47.4	48.2	47.0	37.1	38.6	39.3	40.7	40.1	0.76
Sunday off-premise liquor sales	56.3	55.7	58.1	60.7	75.6	78.5	71.9	88.8	0.01
No BACb limit for driving	20.5	14.4	8.7	3.7	0	0	0	0	0.01
BAC limit for driving < 0.10	77.7	70.3	60.1	50.4	0	0	0	0
BAC limit for driving 0.05–0.08	1.8	15.3	31.2	45.9	100	100	100	100
Past 12-month outcomes M(SE)
Number of drinks	N/Ae	134.9 (19.1)	100.9 (11.1)	120.2 (10.3)	136.9 (10.6)	135.7 (13.7)	201.2 (21.5)	232.6 (15.3)	0.01
Number of days ≥5 drinks	N/A	8.5 (1.9)	5.2 (1.0)	7.9 (1.0)	6.3 (1.0)	7.0 (1.2)	11.7 (2.1)	12.6 (1.3)	0.12
Number of days ≥8 drinks	4.0 (0.9)	3.8 (1.0)	1.9 (0.4)	3.8 (0.7)	2.1 (0.4)	3.0 (1.0)	6.0 (1.4)	6.6 (1.1)	0.08
% Any DSM-IV symptomsd	N/A	19.4	15.4	12.8	14.5	19.1	18.5	20.6	0.56

^aP-value from bivariate test comparing 2020 vs. 1984

^bBAC = Blood Alcohol Concentration

^cAny government control of wine or spirits sales at the wholesale or retail level

^dFor alcohol abuse or dependence

^eN/A: Not available for 1984

Associations between state-level policies and drinking outcomes

Table 2 displays incidence rate ratios (IRRs) with 95% confidence intervals (CIs) from survey-weighted negative binomial regression models predicting past 12-month number of drinks, number of days with ≥5 drinks, and number of days with ≥8 drinks. Results are not shown for DSM-IV alcohol abuse or dependence symptoms because it was not significantly related to any of the policies examined. For each outcome in Table 2, Model 1 adjusts for all policies simultaneously and includes an indicator for survey year; Model 2 further adjusts for individual- and state-level covariates. First, government control of liquor retail sales was not significantly (P<0.05) related to any outcomes except fewer ≥8 drink days in fully adjusted Model 2 (IRR = 0.64, 95% CI: 0.45, 0.90). Next, heavy beer sales at gas stations was related to significantly higher volume, more ≥5 drink days, and more ≥8 drink days in initial Models 1, but only the association for volume remained significant in Model 2 (IRR = 1.21, 95% CI: 1.03, 1.41). Neither grocery store (i.e., heavy beer allowed vs. not; liquor allowed vs. not) variable was related to any outcome in either Models 1 or 2.

Table 2.

Incident rate ratios with 95% confidence intervals from survey-weighted negative binomial regression models predicting alcohol outcomes among women ≤ 44 years, 1984–2020 National Alcohol Surveys (N=13,555)

	Individual-level past 12-month Outcomes (Years of outcome data)
	Number of drinks (1990–2020)		Number of days ≥ 5 drinks (1990–2020)		Number of days ≥ 8 drinks (1984–2020)
	N=11,609		N=11,622		N=13,512
State-level policies	Model 1a	Model 2	Model 1	Model 2	Model 1	Model 2
Government control of liquor retail sales vs. not	1.04 (0.90, 1.21)	1.00 (0.85, 1.17)	1.14 (0.90, 1.44)	0.99 (0.76, 1.30)	0.85 (0.59, 1.23)	0.64 (0.45, 0.90) *
Gas stations, heavy beer allowed vs. not	1.27 (1.10, 1.50) ***	1.21 (1.03, 1.41) *	1.81 (1.20, 2.71) **	1.07 (0.70, 1.64)	2.05 (1.20, 3.50) ***	0.90 (0.46, 1.72)
Grocery stores, heavy beer allowed vs. not	0.75 (0.51, 1.08)	0.78 (0.55, 1.11)	0.50 (0.22, 1.16)	0.86 (0.50, 1.48)	0.54 (0.17, 1.67)	1.00 (0.48, 2.01)
Grocery stores, liquor allowed vs. not	1.01 (0.87, 1.16)	1.06 (0.92, 1.22)	0.96 (0.76, 1.22)	1.02 (0.79, 1.33)	1.00 (0.74, 1.35)	0.96 (0.72, 1.30)
Sunday off-premise liquor sales allowed vs. not	1.12 (0.96, 1.31)	1.20 (1.01, 1.42) *	1.10 (0.82, 1.49)	1.41 (1.08, 1.85) *	1.40 (0.94, 2.07)	1.91 (1.28, 2.83) ***
Blood Alcohol Concentration (BAC) law
0.10 BAC limit for driving vs. no limit	0.62 (0.35, 1.10)	0.63 (0.34, 1.15)	0.54 (0.23, 1.23)	0.48 (0.18, 1.33)	0.44 (0.24, 0.81) ***	0.45 (0.20, 1.00) *
0.05–0.08 BAC limit for driving vs. no limit	0.52 (0.28, 0.98) *	0.51 (0.27, 0.96) *	0.36 (0.14, 0.90) *	0.28 (0.10, 0.75) *	0.26 (0.12, 0.58) ***	0.20 (0.08, 0.47) ***

^aFor each outcome, Model 1 adjusts for all policies simultaneously and includes an indicator for survey year; Model 2 further adjusts for respondent-level age, race/ethnicity, marital status, education, employment, interview month; state/year-level poverty, unemployment; and state consumption patterns.

^***P<.001,

^**P<.01,

^*P<.05 also bolded

^†P<.10

The next policy, whether off-premise Sunday liquor sales are permitted, was not significantly (P<0.05) related to any outcome in Models 1 but did appear significantly related to all outcomes after adjustment for covariates in Models 2. Specifically, living in a state with Sunday off-premise liquor sales allowed was related to having 1.20 times as many drinks (95% CI: 1.01, 1.42), 1.41 times as many ≥5 drink days (95% CI: 1.08, 1.85), and 1.91 times as many ≥8 drink days (95% CI: 1.28, 2.83) over the past 12 months than living in a state that does not allow off-premise liquor sales on Sundays.

Finally, 0.05–0.08 BAC laws were significantly related to all outcomes in Table 2, with the magnitude of effect estimates remaining similar in Model 2 vs. Model 1 for each outcome. Specifically, having a BAC limit of 0.05–0.08 for driving vs. no BAC limit was related to 0.51 times fewer drinks (95% CI: 0.27, 0.96), 0.28 times fewer days with ≥5 drinks (95% CI: 0.10, 0.75), and 0.20 times fewer days with ≥8 drinks (95% CI: 0.08, 0.47) in the past 12 months. The magnitudes of effect estimates for having a BAC limit of 0.10 vs. no BAC limit were also similar between Models, though only the estimate for number of days with ≥8 drinks was significant at P<0.05 in both Models 1 and 2. Post hoc analyses in Table 3 replicated models from Table 2 using the BAC limit of 0.10 as the referent. Table 3, Models 2 show that a 0.05–0.08 BAC limit for driving vs. a 0.10 BAC limit is related to significantly fewer days with ≥5 drinks (IRR=0.57, 95% CI: 0.38, 0.86) and significantly fewer days with days with ≥8 drinks in the past 12 months (IRR=0.44, 95% CI: 0.26, 0.74).

Table 3.

Incident rate ratios with 95% confidence intervals for BAC driving limit laws, using 0.10 BAC limit for driving as referent: post hoc survey-weighted negative binomial regression models predicting alcohol outcomes among women ≤ 44 years, 1984–2020 National Alcohol Surveys (N=13,555)

	Individual-level past 12-month Outcomes (Years of outcome data)
	Number of drinks (1990–2020)		Number of days ≥ 5 drinks (1990–2020)		Number of days ≥ 8 drinks (1984–2020)
	N=11,609		N=11,622		N=13,512
State-level BACb policy	Model 1a	Model 2	Model 1	Model 2	Model 1	Model 2
BAC law vs. 0.10 BAC limit
No limit for driving	1.62 (0.91, 2.90)	1.59 (0.87, 2.93)	1.87 (0.81, 4.29)	2.07 (0.75, 5.69)	2.29 (1.24, 4.21) **	2.21 (1.00, 4.89) *
0.05–0.08 limit	0.85 (0.63, 1.15)	0.81 (0.62, 1.07)	0.67 (0.42, 1.05)†	0.57 (0.38, 0.86) **	0.61 (0.37, 0.99) *	0.44 (0.26, 0.74) **

^aFor each outcome, Model 1 adjusts for government control of liquor retail sales, heavy beer sales permitted at gas stations, heavy beer sales permitted at grocery stores, liquor sales allowed at grocery stores, Sunday off-premise liquor sales, and includes an indicator for survey year; Model 2 further adjusts for respondent-level age, race/ethnicity, marital status, education, employment, interview month; state/year-level poverty, unemployment; and state consumption patterns.

^bBAC = Blood Alcohol Concentration

^***P<.001,

^**P<.01,

^*P<.05 also bolded

^†P<.10

Sensitivity analyses

Post hoc sensitivity analyses examined government control of liquor wholesale as an alternate to government control of liquor retail sales. Effect estimates for government control, Sunday sales, and BAC laws were all robust (Supplemental Table 1). We also ran post hoc sensitivity analyses adjusting for state-level per capita alcohol consumption (measured in gallons of ethanol consumed in population age 15 and older) and separately restricting to pre-2005 survey years, which is when all variation in BAC laws occurred. Results were robust when adjusting for per capita consumption, supporting inferences made (Supplemental Table 2). When restricting to pre-2005 survey years, associations between BAC laws and days with ≥5 drinks and days with ≥8 drinks were also robust while associations between BAC laws and volume were no longer significant at the P<0.05 level.

DISCUSSION

Summary of results

This is the first study to assess whether state-level general population alcohol policies are related to drinking outcomes among women of reproductive age. Overall, findings support that Sunday off-premise liquor sales restrictions and 0.05–0.08 BAC limits for driving are related to less overall and heavy drinking (number of ≥5 drink days and ≥8 drink days) in this population. Prior studies corroborate that looser restrictions on Sunday sales are associated with worse drinking and other health outcomes in both the US and Canada (Carpenter and Eisenberg, 2009; McMillan et al., 2007; McMillan and Lapham, 2006; Yörük, 2014b; Yörük and Lee, 2018). Since at least four US states (Arkansas, Georgia, Ohio, South Carolina) expanded local Sunday sales in November 2022, alcohol consumption might rise overall and among women of reproductive age in these states in the coming years. Researchers and healthcare practitioners should continue tracking how these policy changes affect drinking and other health outcomes.

Our literature search revealed many studies reporting protective associations between stricter BAC laws and alcohol-related crashes, fatalities, and injuries (e.g., as reviewed in Fell and Voas, 2006), but no studies examining how BAC laws relate to alcohol consumption in the US. Thus, this might be the first study to show that stricter BAC limits for driving are associated with drinking outcomes. In their review, Fell and Voas (2006) propose that stricter BAC limits for driving “serve as a general deterrent to drinking and driving.” Our findings suggest that 0.05–0.08 BAC limits for driving might further serve as a deterrent for overall drinking and heavy drinking among women 18–44 years old; this finding is particularly novel and exciting as it supports that stricter BAC limits for driving may have benefits beyond impacting the decision to not drink and drive—stricter BAC driving limit laws may also relate to the decision to not drink heavily in this population. Thus, there may be unintended public health benefits of stricter BAC limits for driving beyond their protective effects on driving outcomes. As this study focused on women of reproductive age, future studies should examine how BAC laws impact per capita consumption as well as individual-level alcohol-related outcomes in the general population.

While our results suggest that allowing heavy beer in gas stations is associated with increased volume, we did not find robust significant associations between drinking outcomes and government control of liquor sales or types of alcohol outlets. The lack of robust results for government control are similar to our prior analyses of the 2000–2015 NAS, which showed that government control of liquor sales was mainly protective for white men (Subbaraman et al., 2020). Although we found some evidence of a protective relationship between government control and days with ≥8 drinks, the lack of robust associations between government control and all drinking outcomes among women is surprising given that government control of alcohol sales has been linked to reduced odds of low birthweight and premature births (e.g., Subbaraman et al., 2019). Thus, additional research to identify empirical mechanisms through which government control of alcohol sales may affect birth outcomes – e.g., reduced drinking from before to during pregnancy or drinking among pregnant people’s partners – is needed.

We also found inconsistent or null results for types of alcohol outlets, contrary to prior results which showed that higher alcohol outlet density was related to increased odds of any drinking (beer outlets) and increased liquor volume (beer and liquor outlets) among white women specifically (Subbaraman et al., 2020). The differences in the null results found here and significant results reported in Subbaraman et al., 2020 could be due to 1) differences in measurement of types of alcohol outlets, i.e., here we used dichotomous indicators for heavy beer and liquor sales at grocery stores and Subbaraman et al., 2020 used continuous alcohol outlet density measures, which might have more power but are not available for the longer time period under study here; 2) here we examined overall drinking volume and heavy drinking outcomes whereas Subbaraman et al., 2020 examined any drinking and beverage-specific volume; and/or 3) here we focused on all women aged 18–44 while Subbaraman et al., 2020 examined women within racial/ethnic subgroups and did not restrict on age.

Implications for policy approaches to reduce harms related to alcohol use during pregnancy

Results indicate that in order to improve drinking outcomes among women of reproductive age, which includes pregnant people, a shift from policies focused on stigmatizing and punishing pregnant people who drink to policies focused on alcohol use more generally is sorely needed. Given that most policies focused on alcohol use among pregnant people do not reduce drinking during pregnancy (Roberts et al., 2019), the current findings support that broader alcohol policies may be relevant for reducing drinking during pregnancy and related harms more effectively than policies that target pregnant people specifically. Historically, state-level policies related to alcohol use during pregnancy have been categorized as punitive (e.g., punishing pregnant people who drink with civil commitment) or supportive (e.g., providing early intervention, treatment, and services to pregnant people by mandating priority substance treatment for pregnant people). While both kinds of policies have increased dramatically across US states since the 1980s, state alcohol and pregnancy policy environments have generally become more punitive in recent years (Roberts et al., 2017). This is alarming because punitive alcohol and pregnancy policies are associated with policies that restrict women’s reproductive rights and not with effective state-level alcohol policies (Roberts et al., 2017). Since most policies targeting alcohol use during pregnancy do not have their intended effects on drinking (Roberts et al., 2019) and are actually related to worse birth outcomes and less prenatal care utilization, the outcomes they ultimately aim to improve (Subbaraman et al., 2018; Subbaraman and Roberts, 2019), current results underscore that a shift from laws focused on singling pregnant people out via stigmatizing and punishing policies to laws focused on alcohol use in the general population could yield better outcomes.

Strengths and Limitations

This is the first study to examine how general population alcohol policies relate to drinking outcomes among women of reproductive age and the first study to examine how BAC driving limits are related to consumption in the US. Additional strengths include the use of repeated cross-sectional datasets covering all US states and a long enough period to examine key policy changes over time; the ability to generalize results to the US population of women ≤ 44 years old; and the robustness of significant findings across models and in sensitivity analyses. Still, these results come with caveats. First, results cannot be interpreted causally. Next, analyses did not adjust for state-level store hours as these data are not available for the entire time period. State-level tax data are also not available for all states and will be examined separately in planned analyses. Planned analyses will also examine the MLDA as this policy has nuances (e.g., MLDA for possession vs. consumption) beyond the scope of the current analyses.

Current data lack statistical power to examine potential relationships between policies and drinking among pregnant people specifically. To better understand whether general alcohol policies are more relevant than pregnancy-specific policies for reducing alcohol use and related harms among pregnant people, future studies should examine relationships between general population policies and outcomes among pregnant people specifically. Finally, drinking outcome data are self-report, and it is likely that past 12-month number of drinks, number of ≥5 drink days, and number of ≥8 drink days are all under-reported; any bias resulting from this should be non-differential across state policies and therefore towards the null.

Conclusion

Using data from all 50 US states for 1984–2020, we find that state-level policies prohibiting Sunday liquor sales and BAC limits of 0.05–0.08 for driving are related to lower overall drink volume and frequency of ≥5 drink days and ≥8 drink days among women of reproductive age, 18–44 years old. Policymakers and public health professionals who wish to reduce harms related to alcohol use during pregnancy through state-level policies should consider broader general population policies. Future research using stronger study designs, e.g., interrupted time series, state fixed effect, or difference-in-difference models, should examine impacts of BAC driving limit laws in the general population as stricter laws may have unintended benefits beyond reducing alcohol-related crashes.