Effects of Restricting Alcohol Sales on Fatal Violence: Evidence from Sunday Sales Bans

Abstract

Introduction.

Homicides and suicides are the second- and third-leading causes of death among young people (aged 10–24) in the US. While a substantial share of these deaths involve alcohol, evidence is needed on whether specific alcohol policies, such as day-based sales restrictions, help prevent these deaths.

Methods.

We constructed total and firearm-related homicide and suicide counts by state, year, and day-of-week from the Multiple Cause of Death Micro-data 1990–2019. Repeals of Sunday bans were taken from the Alcohol Policy Information System. Two-way fixed effects Poisson models with standard errors clustered at state-level and population offset control for state, year and day-of-the-week fixed effects and state time-varying covariates.

Results.

Repealing Sunday bans is associated with an increase in homicides (IRR=1.125; 95% confidence interval [CI]:1.02–1.24) and firearm homicides (IRR=1.17; 95% CI:1.03–1.33). Analyses by day-of-the-week show significant associations with homicides not only on Sundays, but also other days, consistent with delays in death. There was no significant relationship for suicides.

Conclusion.

Restricting alcohol availability may prove a useful policy tool to reduce homicides, given that day-based restrictions are associated with changes in deaths rather than only shifting across days-of-the-week.

Introduction

In 2020, homicides and suicides accounted for more than 70,500 deaths in the United States, with more than 60% of these violent deaths involving a firearm.¹ Through its effects on cognitive impairment, aggression, and impulsivity,^2,3 heavy alcohol use is thought to play a role in many of these deaths. An estimated 30% of homicides and suicides are attributable to alcohol,⁴ and alcohol is involved in more than 30% of firearm homicide and firearm suicide deaths.⁵ While these findings suggest alcohol may be an important factor in violence overall, and specifically among violent deaths involving firearms, more evidence is needed on whether specific policies aimed at restricting alcohol availability can effectively reduce violent deaths.⁵

Several systematic reviews have concluded that efforts to regulate days or hours of alcohol sales are effective at reducing alcohol consumption and related harms, although the strength of evidence varies across policies and outcomes.^6–9 Despite being one of the most common restrictions on alcohol availability in the U.S., Sunday sales laws, which regulate whether individuals can purchase alcohol for off-premise consumption on Sundays, have received relatively little attention as a potential policy lever for reducing violence. While empirical evidence indicates that repealing Sunday bans increases alcohol consumption,^10,11 evidence for effects on crime or violence are generally mixed or based on evidence from a single state or local jurisdiction.^12–14 The only study with data from multiple (seven) states, documented significant increases in assaults, larceny, and vandalism following the repeal of Sunday bans.¹⁵ Interestingly, this study, and the literature as a whole, support effects specific to “Sunday” (often defined through the early Monday hours to capture events that occur soon after midnight). International studies for Germany and Sweden also demonstrated positive associations of greater availability by day or hours with hospitalizations and crime.

To our knowledge, no study has assessed whether the effects of repealing Sunday bans on violence extend to homicides; nor has there been any research examining how policies regulating off-premise Sunday sales influence suicide, despite the strong association between acute alcohol use and suicide¹⁶ and the oftentimes impulsive nature of self-directed harm¹⁷ that may be modified by proximate alcohol availability. Nevertheless, the literature thus far suggest that repealing Sunday bans affects risky behaviors, which ultimately may affect violent deaths not only on Sundays, but also more broadly. Furthermore, the effects on violent deaths may not be limited to Sundays if increased availability on Sundays changes behaviors on other days (e.g., Sunday purchases change the likelihood of drinking throughout the week), or because violent deaths may not be an immediate consequence of intoxication due to delays between consumption, injury, and death. This paper provides the first longitudinal quasi-experimental evidence of the total and day-specific effects of Sunday sales policy changes on homicides and suicides, overall and firearm-involved, using data from all 50 states.

Methods

Data and measures

Using data from the National Vital Statistics System (NVSS) Multiple Cause of Death Micro-data files from 1990 to 2019, deaths were classified based on the ICD-9 or ICD-10 codes as homicides, firearm homicides, suicides, or firearm suicides (eAppendix A for ICD codes). Counts of each homicide and suicide category were constructed by state*year*day-of-week (and by state*year) over the period 1990 to 2019.

The main explanatory variable of interest is the repeal of state laws banning Sunday sales of alcohol beverages for off-premises consumption. Thus, our estimated policy effects represent the effect of transitioning from a more to less restrictive policy regime regulating alcohol sales on Sundays. Information on repeal dates is drawn from NIAAA’s Alcohol Policy Information System (APIS) over the period from 1998–2019 (eAppendix B).¹⁸ To extend the series to the entire NVSS period (1990–2019), we add New Mexico’s policy change in 1995.¹⁹

Our models control for several state-level time-varying characteristics, including economic, demographic, and political measures common in studies of violence (eAppendix C for details on measures, data sources, and pre-processing). Given the role of firearms in violent deaths, we also control for the household gun ownership rate using data from Schell et al.²⁰ Finally, to address potential confounding from changes in religious adherence,²¹ we control for measures of religious affiliation using data from the Association of Religion Data Archives and the U.S. Religion Census.

Study procedures were approved by RAND’s Institutional Review Board. Analyses were conducted using Stata17/MP (College Station, TX).

Approach

In light of prior literature showing that the repeal of Sunday sales bans increases alcohol use and alcohol-related crime, the repeal Sunday sales bans is anticipated to increase violent deaths. However, contrary to the existing literature, which tends to focus on the effects on crime and traffic crashes on Sundays only, we consider that the associations with violent deaths may not be confined to Sundays. There are two primary reasons we may anticipate spillover effects to other days-of-the-week. First, deaths occurring in the early hours after Sunday midnight are categorized as Monday deaths in the NVSS, even when they are likely attributable to Sunday drinking. Unlike other datasets in the Sunday alcohol sales literature, which record the exact hour of the events, the NVSS includes only the date of death. Therefore, we cannot recategorize deaths in the early hours on Monday as occurring on “Sunday,” as other papers have done with crime and traffic crash outcomes.^14,15 Second, prior studies have focused on consequences that are more immediately related to drinking or intoxication; for example, an individual must still be intoxicated (or have detectable levels of alcohol in their bloodstream) in order to be recorded as an alcohol-involved traffic crash or to be arrested for an alcohol-related offense. In the case of mortality, however, the timing of death may be relatively distant. That is, while a heavy drinking episode that precipitated an ultimately fatal injury may occur on Sunday, the victim may survive until the next day or even a few days later. Our analysis of the 2010 Nationwide Inpatient Sample shows that the typical length of stay for those admitted to a hospital who later died of assault was 1 day with an interquartile (IQ) range of 0 to 5 days. The median length of stay was also 1 day for firearm assaults, but the IQ range was substantially shorter at 0 to 2 days. The lengths of stay were similar for suicide (2 days; IQ range: 0–4) and firearm suicides (1 day; IQ range: 0–2). An assault victim admitted after midnight Sunday night and therefore, coded as Monday, survives for the median length of stay of one day, this implies mortality on Tuesday. At the upper IQ range (75^th percentile) of length of stay, it can also imply mortality as late as Saturday.

Furthermore, there are conceptual reasons why providing evidence for the overall effects of changes in day-based alcohol sales restrictions has greater relevance than day-specific effects. Namely, if restricting alcohol sales on a given day reduces harms on that day, but produces offsetting effects in the form of increased harms on other days, a day-specific analysis will overstate the welfare benefits. For these reasons, our main models estimate the average effect of repealing Sunday bans across all days-of-the-week; however, we present day-specific results to better understand the pattern of effects, as well as several sensitivity analyses.

Analysis for Common Effect Across Days

Our main approach involves a two-way fixed effects model, which estimates a common effect across all weekdays between repealing Sunday bans and each mortality category over the 30-year period from 1990–2019. We model death counts for each category by state*year*day-of-week using a Poisson model with standard errors clustered at the state-level and a population offset, so the results are interpreted as effects on death rates. The Poisson is our preferred model as it is known to provide consistent estimates for a range of data-generating processes. These models control for state fixed effects to remove average state characteristics as well as year fixed effects to remove year-specific factors common to all states. In addition, the models include an intercept for each day of the week to account for day-to-day variability in homicides and suicides within the week. We initially estimate an unadjusted model without covariates and then add the following time-varying covariates: household firearm rate (3-year lag), income in 2000 dollars, GOP control (3-year lag), unemployment rate, percent of population aged 15 to 29, percent of population Black, percent of population Hispanic, percent of population with BA degree or higher, percent of population Catholic, and percent of population Evangelical Protestant.

We assess the robustness of our primary results to alternative specifications and samples including: 1) Poisson model with analysis period 1998–2019 based solely on repeals of Sunday bans as reported in the NIAAA APIS, 2) Poisson model adding of state-specific trends; and 3) Negative Binomial model. In the eAppendix, we also show estimates from models with data aggregated across all weekdays to the state*year level. Estimating model with aggregated data provides additional insight into whether the repeal of Sunday bans produces an increase in homicides and suicides, but with a smaller sample that may have implications for precision.

Analysis of Day-Specific Effects

Although we anticipate that repealing Sunday bans may affect total violent deaths, it is possible that the magnitudes of the associations vary over the course of the week. For example, the magnitude of the association could be larger on Sundays simply because off-premise alcohol was not previously available on Sundays and a consequential share of deaths may occur instantaneously. Therefore, we next estimate the Poisson model with interactions between day-of-the-week and the repeal of Sunday bans. In the eAppendix, we also include models estimating only an association on Sunday for comparability to the literature, although these models are not preferred for mortality outcomes because of the potential delay between behaviors resulting in potentially fatal injury and death. The eAppendix also shows results from models estimated separately (i.e., stratified) for each day-of-the-week, which again has implications for sample size and hence precision.

Additional Robustness Checks and Falsification Tests

The main analyses rely on some form of two-way fixed effects models. Recently, concerns about the standard two-way fixed effects models have spurred the development of alternative approaches designed to address concerns about the robustness of two-way fixed effects in contexts of heterogeneous or dynamic treatment effects and staggered policy adoption.²² However, our analyses are not well-suited to the alternatives developed thus far. These alternatives often consider linear estimation of continuous outcomes (rather than count outcomes) and are better-suited for situations where all (or at least most) of the sampling units are eventually treated during the analysis period. In our situation, 29 states were already treated prior to 1990 (hereafter, “always treated”), 13 were treated during the analysis period 1990–2019 (hereafter, “treated”), and 8 remained untreated throughout (hereafter, “never treated”). While we cannot directly apply these new approaches to our data, we acknowledge that the fundamental concern underlying two-way fixed effects models that do not adequately model important aspects of policy heterogeneity still apply. If so, differences in findings may emerge when comparing those units treated (i.e., allowing Sunday sales by removing bans) during the analysis period versus those who were always treated and/or those who were never treated. Therefore, we estimate models on various subsamples with the goal of demonstrating that the relative associations are similar.

Finally, to support that our results for violent deaths are causally attributable to Sunday sales bans, and do not reflect broader trends in mortality in states that repealed the bans, we assess whether repealing Sunday bans influences all-cause mortality. Although alcohol is a leading cause of preventable mortality, it is a relatively small share of all mortality. If we consider only acute causes, alcohol contributes approximately 60,000 of the 2.8 million deaths each year. Therefore, we do not anticipate an association with all-cause mortality. If the models demonstrate an association, it may be the case that other confounders are driving the finding for all-cause mortality, which would then have concerning implications for our findings with respect to homicide and suicide.

Results

Table 1 presents descriptive statistics for homicides and suicides as well as all-cause mortality (Table 1). While all-cause mortality varies little by day of the week, that is not the case for violent deaths: Homicides peak on weekends rather than weekdays, while suicides peak on weekdays (eTable 1).

Table 1:

Summary Statistics for State-level Mortality (1990–2019)

Panel A: Number	Mean	Std Dev	Min	Max	N
Any	7015.37	7054.71	281	39625	10500
Homicide	54.38	72.28	0	801	10500
Firearm Homicide	37.5	51.99	0	613	10500
Suicide	102.04	103.17	2	770	10500
Firearm Suicide	54.64	52.77	0	337	10500
Panel B:Rate	Mean	Std Dev	Min	Max	N
Any	12.275	1.868	4.928	18.966	10500
Homicide	0.082	0.0512	0	0.395	10500
Firearm Homicide	0.054	0.04	0	0.315	10500
Suicide	0.197	0.066	0.019	0.649	10500
Firearm Suicide	0.111	0.051	0	0.425	10500

NOTE: The analysis sample of 10,500 is comprised of 30 years * 50 states * 7 day-of-week observations.

The unadjusted Poisson regression indicates that repealing Sunday bans is associated with higher rates of homicide (Incidence rate ratio [IRR]=1.16; 95% confidence interval [CI]: 1.04–1.34). The estimated IRR remains similar with the addition of time-varying covariates (IRR=1.13; 95% CI: 1.02–1.24) (Table 2). The results for this common effect across days-of-the week are robust to our sensitivity analyses. Focusing on the more recent period with APIS policy data from 1998–2019 (IRR=1.08; 95% CI: 1.01–1.16) or adding state-specific trends (IRR=1.08; 95% CI: 1.00–1.16) produces similar IRRs. Estimating a Negative Binomial regression model produces more conservative and imprecise estimates than the Poisson (IRR=1.07; 95% CI: 0.98–1.17). Focusing on firearm homicides as the outcome of interest produces IRRs that are similar in magnitude to those for total homicides (adjusted IRR=1.17; 95% CI: 1.03–1.33), with similar results across sensitivity analyses (Table 2, Panel B).

Table 2:

Association between Repeal of Sunday Bans and Homicides: Common Effect Across Days-of-Week

	Main		Sensitivity
	Poisson	Poisson	Poisson	Poisson	Negative Binomial
Panel A: Homicides
Repeal of Sunday Ban	1.161* (1.004 – 1.342)	1.125* (1.021 – 1.240)	1.080* (1.010 – 1.156)	1.076* (1.001 – 1.158)	1.073 (0.981 – 1.173)
Covariates	No	Yes	Yes	Yes	Yes
State-Specific Trends	No	No	No	Yes	No
Analysis Period	1990–2019	1990–2019	1998–2019	1990–2019	1990–2019
Sample	10500	10500	7700	10500	10500
Panel B: Firearm Homicides
Repeal of Sunday Ban	1.211* (1.006 – 1.458)	1.169* (1.026 – 1.331)	1.097+ (0.997 – 1.206)	1.093+ (0.996 – 1.199)	1.103 (0.969 – 1.255)
Covariates	No	Yes	Yes	Yes	Yes
State-Specific Trends	No	No	No	Yes	No
Analysis Period	1990–2019	1990–2019	1998–2019	1990–2019	1990–2019
Sample	10500	10500	7700	10500	10500

NOTES: Time-varying covariates include: household firearm rate (3-year lag), income in 2000 dollars, GOP control (3-year lag), unemployment rate, percent of population aged 15 to 29, percent of population Black, percent of population Hispanic, percent of population with BA degree or higher, percent of population Catholic, and percent of population Evangelical Protestant.

All models clustered at the state-level. Clustering at the state-level produced larger confidence intervals than clustering at the state-year level and hence provide a more conservative finding.

^**p<.01;

^*p<.05,

⁺p<.10

In contrast, the models for suicides and firearm suicides consistently show smaller and insignificant associations with the repeal of Sunday bans (Table 3; eTable 2).

Table 3:

Association between Repeal of Sunday Bans and Suicides: Common Effect Across Days-of-Week

	Main		Sensitivity
	Poisson	Poisson	Poisson	Poisson	Negative Binomial
Panel A: Suicides
Repeal of Sunday Ban	1.037+ (0.993 – 1.083)	1.017 (0.994 – 1.041)	1.01 (0.988 – 1.032)	0.994 (0.969 – 1.018)	1.014 (0.992 – 1.037)
Covariates	No	Yes	Yes	Yes	Yes
State-Specific Trends	No	No	No	Yes	No
Analysis Period	1990–2019	1990–2019	1998–2019	1990–2019	1990–2019
Sample	10500	10500	7700	10500	10500
Panel B: Firearm Suicides
Repeal of Sunday Ban	1.042 (0.967 – 1.121)	1.029 (0.987 – 1.073)	1.018 (0.984 – 1.052)	0.995 (0.956 – 1.035)	1.022 (0.981 – 1.065)
Covariates	No	Yes	Yes	Yes	Yes
State-Specific Trends	No	No	No	Yes	No
Analysis Period	1990–2019	1990–2019	1998–2019	1990–2019	1990–2019
Sample	10500	10500	7700	10500	10500

NOTES: Time-varying covariates include: household firearm rate (3-year lag), income in 2000 dollars, GOP control (3-year lag), unemployment rate, percent of population aged 15 to 29, percent of population Black, percent of population Hispanic, percent of population with BA degree or higher, percent of population Catholic, and percent of population Evangelical Protestant.

All models clustered at the state-level. Clustering at the state-level produced larger confidence intervals than clustering at the state-year level and hence provide a more conservative finding.

^**p<.01;

^*p<.05,

⁺p<.10

In the eAppendix, eTable 2 shows the results for models estimated on death counts aggregated across weekdays to state-year totals. These models show the same magnitude and precision for the associations between the repeal of Sunday bans and violent deaths despite the smaller sample sizes.

Day-Specific Effects

The models with interactions between day-of-week and the repeal of Sunday bans also demonstrate an association with homicides. The adjusted Poisson model yields a significant association for Sundays (IRR=1.17; 95% CI: 1.06–1.29) (Table 4). The estimates for the interactions between repeal of Sunday bans and other weekdays are close to one and mostly insignificant, suggesting null or very small associations for other days. The corresponding association for firearm homicides on Sundays was also significant (IRR= 1.23; 95% CI: 1.08–1.40). However, for firearm homicides, the interactions were significant with IRRs less than 1 for several days-of-the-week; hence the association for those days suggests some offset from Sunday. There was no association between the repeal of Sunday bans on suicides with the exception of a small effect on Mondays. The Appendix shows similar results for Poisson models estimated over the shorter period of 1998–2019 as well as the Negative Binomial model estimated over both periods (see eTable 3). The results are also qualitatively similar for Sundays when we focus on the more common models in this literature, which posit an association for Sundays only; however, this model is not preferred for our outcome of interest due to potential delays in mortality (eTable4). Estimation on the sample stratified by day-of-the-week produces a relatively larger IRR for homicides that is statistically significant for Sundays despite the smaller sample (eTable 5). Positive associations with homicides were also found for other days-of-the-week, although they were imprecise for Wednesdays and Thursdays. The pattern of findings is similar for firearm homicides to that of total homicides.

Table 4:

Association between Repeal of Sunday Bans and Violent Deaths: Day-Specific Effects

	Homicide 1990–2019	Firearm Homicide 1990–2019	Suicide 1990–2019	Firearm Suicide 1990–2019
Repeal of Sunday Ban (Reference: Sunday)	1.172** (1.061 – 1.294)	1.229** (1.076 – 1.403)	1.002 (0.978 – 1.028)	1.011 (0.970 – 1.053)
Monday*Repeal of Sunday Ban	0.998 (0.993 – 1.002)	0.978 (0.929 – 1.029)	1.016 (0.991 – 1.041)	1.029* (1.001 – 1.057)
Tuesday*Repeal of Sunday Ban	1 (0.995 – 1.004)	0.939* (0.884 – 0.999)	1.016 (0.993 – 1.040)	1.018 (0.990 – 1.047)
Wednesday*Repeal of Sunday Ban	1 (0.996 – 1.004)	0.934* (0.874 – 0.999)	1.026+ (0.999 – 1.054)	1.031+ (1.000 – 1.063)
Thursday*Repeal of Sunday Ban	1.001 (0.998 – 1.004)	0.940+ (0.882 – 1.001)	1.016 (0.993 – 1.038)	1.006 (0.980 – 1.032)
Friday*Repeal of Sunday Ban	1 (0.998 – 1.003)	0.924** (0.872 – 0.979)	1.023+ (0.999 – 1.047)	1.023+ (0.999 – 1.048)
Saturday*Repealof Sunday Ban	0.997** (0.994 – 0.999)	0.937** (0.912 – 0.962)	1.008 (0.990 – 1.027)	1.016 (0.994 – 1.039)
Sample	10500	10500	10500	10500

NOTES: Time-varying covariates include: household firearm rate (3-year lag), income in 2000 dollars, GOP control (3-year lag), unemployment rate, percent of population aged 15 to 29, percent of population Black, percent of population Hispanic, percent of population with BA degree or higher, percent of population Catholic, and percent of population Evangelical Protestant.

Standard errors are clustered at the state-level.

^**p<.01;

^*p<.05,

⁺p<.10

Robustness Checks and Falsification Tests

Sensitivity analyses show estimates for the main models with three sets of comparisons (eTable 6): 1) states treated during analysis period versus always treated; 2) states treated during analysis period versus never treated; and 3) states treated during analysis period only (comparing to not-yet-treated). These three comparisons produce estimates highly similar to our main findings for homicides: (IRR=1.09; 95% CI: 1.05–1.14), (IRR=1.08; 95% CI: 1.03–1.14), and (IRR=1.09; 95% CI: 1.04–1.15), respectively. The corresponding estimates for firearm homicides also show similar effect sizes across the three comparisons: (IRR=1.16; 95% CI: 1.06–1.17), (IRR=1.12; 95% CI: 1.05–1.18), and (IRR=1.13; 95% CI: 1.07–1.19). As with our main analyses, estimated IRRs were small or insignificant for suicides.

Finally, as expected, our results for all-cause mortality show no relationship with repeal of Sunday bans. The IRRs are very close to one and precisely-estimated (eTable 7).

Discussion

The substantial geographic variation in homicides and suicides has raised the question of whether variation in state policies regarding firearms or alcohol could play a role²³. Our study provides the first quasi-experimental evidence on how day-based alcohol sales restrictions---specifically, the repeal of Sunday bans of off-premises alcohol sales---may contribute to violent deaths. This adds a new dimension to Sunday bans literature, which has focused largely on motor vehicles crashes and crime.^6–9

We find that repeals of Sunday bans are consistently associated with increased homicides, including firearm homicides. While the relationship of Sunday bans with homicides has not been examined in the US, evaluations of time-based alcohol restrictions from other countries are broadly consistent with our results. For example, an analysis of bar closing hours in Diadema Brazil found an approximately a 40–50% reduction in monthly homicides,²⁴ suggesting that the reduction in murders was not offset by increases during other times of the day. Similarly, there was an increased homicide risk in Cali Colombia (IRR=1.42) during periods of lax vs most restrictive alcohol availability based on time-of-day, as well as during periods of lax vs moderately restrictive alcohol availability (IRR=1.15).²⁵ More recently, an unexpected alcohol ban in South African during the COVID-19 pandemic was associated with at least a 14% reduction in violent deaths²⁶. We find smaller magnitude effects for Sunday sales bans, which may be due to the fact that on-premise consumption has different implications for social interactions or due to methodological differences such as focusing on a single smaller jurisdiction without considering spillover effects. More broadly, others have documented associations with consumption,

Importantly, our focus is on the total effect on homicides and firearm homicides. Our results are thus consistent with a net increase in homicides subsequent to repealing Sunday bans. We also find some evidence that other days are affected though these results are sensitive to the approach. At first glance, this may seem at odds with the literature on Sunday sales, which typically focuses on Sundays only. However, the literature’s focus on Sundays is likely an artifact of the nature of outcomes studied in the literature thus far, which are defined based on active drinking and/or intoxication levels and which are recorded at the time of event. This contrasts with deaths, which may be delayed relative to the alcohol purchase and even alcohol consumption.

On the other hand, our results do not support a substantial or significant association between the repeal of Sunday bans and suicide. That may follow from several factors. First, in contrast with homicides, suicides are more likely to occur during weekdays when alcohol was already available. Drinking motives also differ between weekday and weekend sessions, with coping motives more strongly associated with weekday drinking;²⁷ thus, allowing alcohol sales on Sundays may produce effects on types of behaviors, consumers, or settings less relevant for self-harm. Second, the role of alcohol in suicides may be too small to identify in our data or may be heterogenous, for example by gender,²⁸ which may produce muted effects on average.

Our study has a number of limitations. First, our data do not specifically identify the role of alcohol in these deaths, which precludes assessment of potential mechanistic pathways between Sunday sales bans, alcohol consumption, and homicides. Second, accuracy of our primary results relies on the assumption that the effects of state repeals of Sunday sales bans are immediate and similar across states. The structure of the identifying variation in our data are unsuitable to more recent methods that relax this assumption, though we can offer several analyses that support that our findings are unlikely to be undermined by heterogeneous or time-varying policy effects. Finally, we operationalize the policy as a dichotomous indicator for repealing Sunday bans using only repeals from the NIAAA APIS system (with the addition of New Mexico’s repeal in the longer analysis period only). Some researchers have defined the repeals differently, for example, to also include states that relax---but do not repeal---Sunday bans (e.g., alcohol type). And still others have considered the county’s ability to opt in/out of Sunday sales, albeit only in studies of a single state. But we are unable to take advantage of within-state variation in Sunday sales as our NVSS data do not include county identifiers nor is there systematic data on county bans available nationwide.

Public Health Implications.

Homicide and suicide remain the second- and third-leading cause of mortality among youth and young adults (aged 10–24), and increasing rates of both have contributed meaningfully to recent declines in life expectancy in the U.S.²⁹ Given the greater support for, and consequently feasibility of, enacting alcohol policies vis-à-vis gun policies, we assess whether existing alcohol policies can influence violent deaths overall and violent deaths involving firearms specifically. Our analysis provides quasi-experimental evidence that repealing bans on off-premises alcohol sales on Sundays was associated with an increase in homicides, including firearm homicides, but likely not suicides. These finding suggest that day- or time-based restrictions on the availability of alcohol may prove a useful policy tool in reducing homicides, but that these policies are unlikely to meaningfully affect population-level rates of suicide. While our findings suggest the viability of restricting alcohol sales as a means of violence prevention, they do not preclude the relevance of gun policies – on their own or in combination with alcohol policy – in addressing violent deaths in the U.S.

Highlights.

• Repeal of Sunday bans was associated with increases in homicides

• Similar associations were noted for firearm homicides

• Repeal of Sunday bans was not associated with significant changes in suicides

• Alcohol policy may represent an important tool to address the crisis of violence