Abstract
Introduction:
Numerous studies have found associations between alcohol outlet density and violence, but it is unknown whether alcohol advertisements (ads) visible outside outlets are also associated with violent crime. Baltimore City, Maryland enacted restrictions on retail alcohol establishment advertising practices as of June 5, 2017. This study examines the association between alcohol ads visible outside off-premise alcohol outlets and violent crime prior to this restriction.
Methods:
Outlet observations (n=653) were conducted in summer 2015, and violent crime data (n=24,085) were from June 5, 2015 through June 4, 2017. The number of violent crimes per square mile within 1,000 feet of outlets was summed using kernel density estimation. In 20182019, authors used mixed models with Simes-Benjamini-Hochberg correction for multiple testing.
Results:
Roughly half (47%, n=267) of the outlets with complete data (n=572) had alcohol ads visible from the exterior. Outlets with alcohol ads had 15% more violent crimes per square mile within 1,000 feet (ê=1.15, 95% CI=1.07, 1.25, q<0.001) after adjuŝ’ng for neighborhood context. All associations between alcohol ads and specific type’ of violent crime were significant, with the association strongest for homicides (ê=1.28, 95% CI=1.13, 1.46, q<0.001). There was no association between cigarette ads and violent crime (eB=1.08, 95% CI=0.92, 1.26, 0=0.43).
Conclusions:
Alcohol ads visible outside off-premise outlets were associated with increased violent crime over and above the ass ‘ciation between the outlets themselves and violent crime. Reducing alcohol advertising visible from the street may decrease risk of violent crime that is associated with alcoho1 outlets.
INTRODUCTION
In the U.S., approximately 21 homicides a day are considered attributable to excessive alcohol use.1 Each additional liter of absolute alcohol consumed per capita is associated with an 8% increase in the homicide rate in the U.S.2 and a 9% increase in Australia.3 At the individual level, there is a dose-response relationship between alcohol consumption and aggression.4 The density of stores selling alcohol (i.e., alcohol outlets) is also closely linked with level of violence,5 but it is unlikely that all alcohol outlets contribute to this violence equally.6 Identifying modifiable characteristics of “problem” outlets could help tailor prevention strategies.
Alcohol advertising may be a feature of outlets that contributes to violent crime.6–8 Off-premise alcohol outlets (e.g., packaged goods stores where patrons buy alcohol to consume off site), where approximately 76% of alcohol in the U.S. is sold,9 are a common site for alcohol advertising. Though two systematic reviews have found relationships between alcohol advertising exposure and youth drinking,10,11 there are three reasons why alcohol ads may increase violence: (1) Advertisement (ad) exposure increases alcohol consumption,10,11 which could then increase violent crime12; Ads often display price discounts, and low-price alcohol is independently associated with higher levels of consumption13 and violence14; and (3) Ad content shapes social norms in ways that promote aggression and violence.8 Exposure to violence and alcohol advertising are not the same to • all communities, with risks significantly higher in black communities.5,15–20
To date, the authors are aware of only three manuscripts describing an association between advertising on alcohol outlets and violent crime, only two of which empirically test that association.8 A 2018 analysis of survey data from South Africa found that women reporting high exposure to alcohol ads in their communities were 4.2 times more likely also to report being victims of intimate partner violence.7 A study of three California communities found that ads with Latina models were associated with higher rates of sexual violence against Latina and non-Latina women.8 These studies demonstrate an association between alcohol ads and violence may exist. However, the South African study relied on self-reported ad and violence data, which are subject to under-reporting. Both studies were in unique contexts that may not generalize to the majority of the U.S. Finally, the South African study did not adjust for alcohol outlet density, so their measure of alcohol ads may be a proxy for the outlets themselves. The California study adjusted for outlets, but did so using unadjusted counts of outlets, which the authors’ prior work has demonstrated to be prone to error.21
Baltimore passed a comprehensive zoning code update called TransForm Baltimore on December 5, 2016, with an estimated 2-year implementation period.22 Under the new regulations, it is illegal to post signs, posters, graphics or other items that advertise alcoholic beverages in a publicly visible location, with one exception for licensed premises, which may display an external sign if: the purpose is to identify the business, using the business’ name, slogan, or logo; the size is no greater than 1% of the exterior wall or window space; and the outlet has no other window or exterior signs on the same street frontage.23
The primary research question is whether there is an empirical relationship between presence of alcohol ads visible outside off-premise alcohol outlets and density of violent crime around the outlet. The unit of an alysis is the outlet itself, which is nested within Census tracts (CTs). Mixed models are used to separate outlet-specific associations from contextual effects, and adjust for off-premise outlet clustering.
METHODS
Study Sample and Measures
Data were obtained from the Board of Liquor License Commissioners for Baltimore City for establishments permitted to sell alcohol for off-site consumption in 2015 (N=683). Teams of research assistants conducted alcohol outlet observations in 2015. The observation instrument collected data about ads visible from outside the outlet. Complete exterior advertising data were available for 572 (83.7%) alcohol outlets.
To understand whether these alcohol ads were still present after TransForm Baltimore went into effect (June 5, 2017), the authors compared 2015 and 2018 observation data (more than a fullyear after TransForm Baltimore went into effect) for a subset of outlets (n=45), and found little change. The agreement rate was 86% (Cohen’s k=0.73). Five outlets rated differently in 2015 and 2018: Four had ads in 2018 but not <015, and one had ads in 2015 but not 2018.
Victim-based violent crime incident data were obtained from the Baltimore City Police Department. Violent dime was defined using the Federal Bureau of Investigation Uniform Crime Reporting definition: homicide, forcible rape, aggravated assault (including non-fatal shootings), and robbery.24 Authors measured violent crime after the outlet observations (summer 2015) and before TransForm Baltimore went into effect. Thus, the violent crime data spanned June 1, 2015 through June 4, 2017. The level and distribution of violent crime were compared for 2015–2016 and 2016–2017 by type of crime. There were slightly more crimes in 2016–2017 but the distribution was the same.
Kernel density estimation was used to measure the expected count of violent crimes within 1,000 feet of each off-premise outlet, given that a quarter mile (1,320 feet) is considered walking distance in urban areas.25 The violent crime variable summed the kernel density estimation raster values that fell within 1,000 feet of each outlet. The authors then calculated crime density per square mile by dividing these counts by the area of the buffer (0.11 square miles). They transformed all violent crime variables using the natural logarithm to adjust for positive skew.
Authors obtained demographic covariates from the 2017 American Community Survey 5-year estimates at the CT level because the margins of errors are lower at the CT level than at the Census block group level. They included percentage black, median annual household income, and percentage of the population who were male are d aged 15–24 years because blacks, low-income individuals, and male youth tend to have the highest levels of violence exposure.26 The median annual household income variable was scaled so that a 1-unit increase corresponded to $10,000.
The authors used a social disadvantage index calculated by subtracting the sum of two percentages for positive outcomes (i.e., adults aged ≥25 years with a college degree and owner-occupied housing) from two negative outcomes (i.e., households with incomes below the federal poverty line and female-headed households).27 A 1-unit increase in this scale corresponded to a 10% increase in each of the four components. Figure 1 shows the distribution of alcohol outlets with alcohol ads visible outside and the social disadvantage index, where darker areas indicate higher disadvantage.
Figure 1.
Distribution of alcohol utlets permitted to sell alcohol for off-premise consumption by presence of alcohol ads on the building exterior and Social Disadvantage Index, Baltimore City 2015–2017.
Ads, advertisements.
Four covariates describing the built environment were included: (1) off-premise alcohol outlet clusters, (2) drug arrest density, (3) vacant housing density, and (4) land use zoning. The authors calculated off-premise alcohol outlet density using the Getis-Ord Gi,28 which is a local measure of spatial dependence that identifies “hot spots” where the density of points is greater than the study area. Variables for drug arrest and vacant property density at the CT level were calculated and transformed using the same methods as used to measure violent crime. Finally, land use zoning shapefiles were obtained from the Baltimore City Department of Planning and a spatial join was performed to determine the zoning type in which each outlet was located.
Statistical Analysis
The primary research question was: Do off-premise alcohol outlets with alcohol ads visible outside have higher violent crime density within 1,000 feet? Secondarily, the authors asked whether this association was from the themselves or the context in which the ads are located. Random effects modeling was used to account for the lack of independence between alcohol outlets nested within the same CT. Random effects models can separate outlet-level and cluster (CT)-level associations. these models, the CT proportion tested whether outlets with alcohol ads located in a C± vuh a higher proportion of outlets with alcohol ads had a stronger association with violent crime. This serves as a proxy for neighborhood effects such as social milieu.
Stata, version 14.2 was used to fit the mixed models in 2018–2019.29 The random effects models used random intercepts and fixed effects for the independent variables (ad exposure) and covariates. The models assumed that alcohol outlets within a CT had an identity structure.
The authors first ran a series of unadjusted random effects models with no covariates. They then ran four sets of models that each had five dependent variables: total crime, homicide, aggravated assault, forcible rape, and robbery. The first set of models examined alcohol advertising and the second examined cigarette advertising. The second set of models with cigarette ads tested whether it was the ad content or the characteristics of outlets that display ads that drove any associations with violent crime. All regressions report a g-value estimated using the Simes-Benjamini-Hochberg correction for multiple testing,30 with g-values <0.05 considered statistically significant. Authors repeated our statistical models using generalized estimating equations specifying a Gaussian family, identity link, and exchangeable correlation structure and obtained nearly identical results (data not shown).
The authors used R, version 3.5.2 for spatial analyses.31 Spatial dependence in the violent crime outcomes and regression residuals was assessed using Moran’s Index (Moran’s I). There was positive spatial dependence in the outcomes (Moran’s I=0.41, p<0.001). The regression covariates explained all of the spatial dependence for models of homicide and aggravated assault and most of the spatial dependence for total violent crime, rape, and robbery (Moran’s I<0.03, p<0.05). A Hubf r-White sandwich estimator of the variance was used to obtain robust SEs to account for the r sidual spatial dependence. Even if the robust SEs did not account for all of this residual spatial dependence, the inferences should be approximately accurate because the residual dependence was small and most of the associations were highly significant.
RESULTS
The final sample included 572 alcohol outlets selling alcohol for off-site consumption (Table 1). Nearly half (46.5%) had alcohol advertising visible from outside the building. The average CT in which these outlets with alcohol ads were located had a population density of 3,131 residents and (like Baltimore overall)32 was mostly composed of black residents (61.7%).
Table 1.
Summary of Alcohol Outlets and Census Tracts Included in Analysis, Baltimore City 2015–2017
| Characteristics | Percent or mean | SD | Minimum | Maximum |
|---|---|---|---|---|
| Outlets (n=572) | ||||
| Any alcohol ads visible from outlet exterior, % (n) | 46.6 (267) | |||
| Any tobacco ads visible from outlet exterior, % (n) | 7.2 (41) | |||
| Census tracts (n=153) | ||||
| Population, n | 3,131 | 1,405 | 697 | 7,144 |
| Population who are Black, % | 61.7 | 34.6 | 0.5 | 100.0 |
| Population who are Hispanic/Latino, % | 5.4 | 7.6 | 0.0 | 43.1 |
| Population who are white, % | 28.8 | 29.1 | 0.0 | 91.8 |
| Median annual household income | $49,537 | $27,134 | $14,452 | $145,966 |
| Drug arrest density per square mile, mean | 44 | 41 | 1 | 211 |
| Vacant housing density per square mile, mean | 147 | 130 | 8 | 684 |
The intraclass correlation coefficient was 0.79, suggesting that random effects models were appropriate for the nested data structure. Outlets with alcohol ads visible outside had higher levels of aggravated assault (eB =1.17, 95% CI=1.07, 1.28, p<0.001), homicide (eB = 1.31, 95% CI=1.26, 1.52, p<0.001), and robbery density (eB= 1.18, 95% CI=1.05, 1.31, p<0.01) within 1,000 feet compared with alcohol outlets without visible ads.
The presence of alcohol ads was associated with higher levels of violent crime density immediately around the outlet after adjusting for outlet characteristics, off-premise alcohol outlet clustering, and neighborhood characteristics (eB=1.15, 95% CI=1.07, 1.25, g<0.001, Table 2). The association between alcohol ads and violent crime was strongest for homicide; the presence of alcohol ads vi siblt outside an off-premise outlet was associated with 28.4% more homicides within 1,000 feet than outlets without these ads (eB=1.28, 95% CI=1.13, 1.46, g<0.001). Alcohol outlets with alcohol ads were also associated with 16.2% more aggravated assaults (eB =1.16, 95% CI=1.07, 1.26, g<0.001) and forcible rapes (eB=1.16, 95% CI=1.04, 1.28, g=0.01) and 15.0% more robberies (eB=1.15, 95% CI=1.06, 1.25, q<0.001).
Table 2.
Regression Results for Level of Crime Around the Outlet and Any Alcohol Ads Visible Outside in Baltimore City, 2015–2017
| Variable | Violent crime | Homicide | Aggravated assault | Forcible rape | Robbery | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| exp(B)a (95% CI) |
Q-valb | exp(B) (95% CI) |
Q-val | exp(B) (95% CI) |
Q-val | exp(B) (95% CI) |
Q-val | exp(B) (95% CI) |
Q-val | |
| Alcohol ads | ||||||||||
| No | ref | ref | ref | ref | ref | |||||
| Yes |
1.15 (1.07, 1.25) |
<0.001 |
1.28 (1.13,1.46) |
<0.001 |
1.16 (1.07, 1.26) |
<0.001 |
1.16 (1.04,1.28) |
0.012 |
1.15 (1.06,1.25) |
0.001 |
| Mean prevalence of ads |
0.75 (0.58, 0.97) |
0.043 |
0.50 (0.27, 0.90) |
0.045 | 0.79 (0.54, 1.17) |
0.318 | 0.77 (0.54, 1.09) |
0.218 |
0.70 (0.53, 0.93) |
0.021 |
| Off-premise outlet cluster | ||||||||||
| No | Ref | ref | ref | ref | ref | |||||
| Yes |
1.73 (1.42, 2.10) |
<0.001 | 1.67 (1.03,2.69) |
0.069 |
2.27 (1.67, 3.06) |
<0.001 |
1.65 (1.27, 2.16) |
0.001 |
1.75 (1.42, 2.16) |
<0.001 |
| Percent black | 1.16 (0.79, 1.72) |
0.524 |
9.03 (3.60, 22.42) |
<0.001 | 1.79 (0.98, 3.22) |
0.078 | 1.28 (0.76,2.16) |
0.390 | 0.84 (0.55,1.26) |
0.415 |
| Median annual incomec | 0.99 (0.92, 1.05) |
0.646 | 1.08 (0.92, 1.27) |
0.436 |
0.84 (0.76, 0.93) |
0.001 | 1.03 (0.94, 1.12) |
0.511 | 1.00 (0.93, 1.07) |
0.992 |
| Percent male aged 15–24 years | 1.22 (0.97, 1.52) |
0.112 | 1.27 (0.74,2.18) |
0.436 | 1.11 (0.78, 1.55) |
0.592 | 1.35 (1.00, 1.84) |
0.089 |
1.32 (1.03,1.68) |
0.033 |
| Social Disadvantage Indexd |
1.30 (1.09,1.52) |
0.003 |
1.88 (1.25,2.80) |
0.007 | 1.08 (0.84, 1.39) |
0.587 |
1.46 (1.16,1.82) |
0.003 |
1.23 (1.03,1.48) |
0.027 |
| Drug arrest density |
1.35 (1.21,1.49) |
<0.001 |
1.88 (1.46, 2.44) |
<0.001 |
1.35 (1.14, 1.60) |
0.001 |
1.49 (1.28, 1.72) |
<0.001 |
1.32 (1.17,1.48) |
<0.001 |
| Vacant housing density |
0.84 (0.75, 0.93) |
0.003 |
0.72 (0.55, 0.94) |
0.042 |
0.81 (0.68, 0.96) |
0.022 |
0.79 (0.68, 0.91) |
0.005 |
0.78 (0.69, 0.89) |
<0.001 |
| Zoning | ||||||||||
| Commercial | ref | ref | ref | ref | ref | |||||
| Industrial |
0.53 (0.43, 0.64) |
<0.001 | 0.83 (0.57, 1.19) |
0.436 |
0.53 (0.43, 0.66) |
<0.001 | 0.84 (0.63, 1.12) |
0.296 |
0.70 (0.57, 0.88) |
0.003 |
| Residential | 0.97 (0.90, 1.05) |
0.524 | 1.00 (0.86, 1.16) |
0.977 | 0.95 (0.87, 1.04) |
0.346 | 1.04 (0.92, 1.17) |
0.511 | 0.93 (0.85, 1.02) |
0.174 |
| Other |
0.66 (0.51, 0.86) |
0.003 | 0.80 (0.49, 1.30) |
0.436 |
0.68 (0.52, 0.91) |
0.015 | 1.25 (0.87,1.79) |
0.298 |
0.67 (0.51, 0.89) |
0.009 |
| Moran’s Index |
0.02 (−0.29) |
0.04 | −0.02 | 0.99 | <0.01 | 0.56 | 0.03 | 0.001 | 0.03 | <0.001 |
Notes: Boldface indicates hat the corrected q-value<0.05.
“Exponentiated regression coefficients presented because the regression outcome (violent crime density) was transformed using the natural logarithm.
Adjusted probability score using a Benjamini-Hochberg-Simes false discovery rate correction.
Median annual household income, censored at $250,000 and scaled so a unit increase equals $10,000.
Calculated as ([(% female-headed households/10) + (% families living in poverty/10)] − [(% owner-occupied housing/10) + (% adults aged 25+ with college degree/10)]) / 4.
Ads, advertisements.
The association between alcohol ads and violence held after adjusting for the CT-level proportion of outlets with alcohol ads. Increases in the CT-level proportion of outlets with alcohol ads visible outside were associated with lower levels of violent crime density (eB=0.75, 95% CI=0.58, 0.97, q=0.04). This suggests a saturation model where the first outlet(s) with alcohol ads in a CT has (have) a stronger association with violent crime than additional outlets with ads in CTs where more than half of the outlets already display ads.
The authors tested interaction terms between the alcohol outlet ad indicator and off-premise alcohol outlet clustering indicator for each type of violent crime.None were significant, suggesting that the association between alcohol ads visible outsidet crime does not differ within and outside of clusters of off-premise outlets. However, the association between off-premise alcohol outlet density clusters and violent crime was significant for all types of violent crime, and strongest for aggravated assaults: Outlets with alcohol ads visible outside and located inside a cluster of off-premise outlets had 127.0% more aggravated assaults per square mile (eB=2.27,95% CI=1.35, 3.82, q<0.001).
Authors explored associations between cigarette ads and violent crime to test whether the associations for alcohol ads could derive from characteristics of outlets that display ads in general (Table 3). Overall, there was no association (eB=1.08, 95% CI=0.92, 1.26, q=0.43). Models that stratified by type of violent crime also found no association.
Table 3.
Regression Results for Level of Violent Crime Around Off-Premi”e owlets and Cigarette Ads, Baltimore City 2015–2017
| Variable | Violent crime | Homicide | Aggravated assault | Forcible rape | Robbery | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| exp(B)a (95% CI) |
Q-valb | exp(B) (95% CI) |
Q-val | exp(B) (95% CI) |
Q-val | exp(B) (95% CI) |
Q-val | exp(B) (95% CI) |
Q-val | |
| Cigarette ads | ||||||||||
| No | ref | ref | ref | ref | ref | |||||
| Yes | 1.08 (0.92, 1.26) |
0.426 | 1.05 (0.81, 1.38) |
0.815 | 1.06 (0.90, 1.25) |
0.551 | 1.07 (0.87, 1.32) |
0.558 | 1.11 (0.94, 1.31) |
0.258 |
| Mean prevalence of ads | 0.84 (0.56, 1.25) |
0.462 | 0.92 (0.38, 2.23) |
0.908 | 0.95 (0.54, 1.67) |
0.856 | 0.84 (0.49, 1.43) |
0.558 | 0.74 (0.48, 1.14) |
0.212 |
| Off-premise outlet cluster | ||||||||||
| No | ref | ref | ref | ref | ref | |||||
| Yes |
1.72 (1.40, 2.10) |
<0.001 | 1.67 (1.03,2.69) |
0.098 |
2.27 (1.67,3.06) |
<0.001 |
1.65 (1.26, 2.16) |
0.001 |
1.73 (1.40, 2.16) |
<0.001 |
| Percent black | 1.17 (0.79, 1.73) |
0.462 |
9.21 (3.63, 23.10) |
<0.001 | 1.79 (0.99, 3.25) |
0.083 | 1.30 (0.77,2.18) |
0.442 | 0.84 (0.55, 1.28) |
0.456 |
| Median annual incomec | 0.99 (0.93, 1.05) |
0.789 | 1.11 (0.94, 1.30) |
0.419 |
0.85 (0.77, 0.93) |
0.001 | 1.03 (0.95, 1.13) |
0.547 | 1.01 (0.94, 1.07) |
0.819 |
| Percent male aged 15–24 years | 1.23 (0.98, 1.54) |
0.117 | 1.32 (0.77, 2.27) |
0.421 | 1.11 (0.79, 1.57) |
0.599 | 1.36 (1.00, 1.84) |
0.094 |
1.34 (1.04, 1.70) |
0.035 |
| Social Disadvantage Indexd |
1.30 (1.11,1.54) |
0.004 |
1.90 (1.26,2.83) |
0.008 | 1.08 (0.84, 1.39) |
0.599 |
1.46 (1.17,1.82) |
0.002 |
1.25 (1.04, 1.49) |
0.026 |
| Drug arrest density |
1.35 (1.21,1.49) |
<0.001 |
1.88 (1.45, 2.44) |
<0.001 |
1.35 (1.14,1.60) |
0.001 |
1.49 (1.28,1.72) |
<0.001 |
1.31 (1.17,1.48) |
<0.001 |
| Vacant housing density |
0.84 (0.74, 0.93) |
0.003 | 0.73 (0.55, 0.95) |
0.066 |
0.81 (0.68, 0.96) |
0.028 |
0.79 (0.68, 0.91) |
0.005 |
0.78 (0.69, 0.88) |
<0.001 |
| Zoning | ||||||||||
| Commercial | ref | ref | ref | ref | ref | |||||
| Industrial |
0.52 (0.43, 0.64) |
<0.001 | 0.81 (0.56, 1.17) |
0.421 |
0.53 (0.42, 0.65) |
<0.001 | 0.83 (0.63, 1.11) |
0.332 |
0.70 (0.56, 0.88) |
0.004 |
| Residential | 0.96 (0.89, 1.05) |
0.462 | 0.99 (0.85, 1.16) |
0.908 | 0.94 (0.86, 1.04) |
0.323 | 1.03 (0.92, 1.16) |
0.578 | 0.93 (0.85, 1.02) |
0.159 |
| Other |
0.66 (0.51, 0.86) |
0.004 | 0.81 (0.50, 1.32) |
0.502 |
0.68 (0.51, 0.90) |
0.016 | 1.23 (0.86, 1.79) |
0.381 |
0.67 (0.51, 0.89) |
0.010 |
| Moran’s Index | 0.02 | 0.03 | <0.01 | 0.93 | <0.01 | 0.50 | 0.03 | <0.001 | 0.03 | <0.001 |
Notes: Boldface indicates that the corrected q-value<0.05.
Exponentiated regression coefficients presented because the regression outcome (violent crime density) was transformed using the natural logarithm.
Adjusted probability score using a Benjamini-Hochberg-Simes false discovery rate correction.
Median annual household income, censored at $250,000 and scaled so a unit increase equals $10,000.
Calculated as ([(% female-headed households/10) + (% families living in poverty/10)] − [(% owner-occupied housing/10) + (% adults aged 25+ with college degree/10)]) / 4.
Ads, advertisement.
DISCUSSION
This study sought to explore associations between two contextual risk factors—outlet-level alcohol advertising and alcohol outlet clustering—and violence, and found both were independently and positively associated with violent crime. Publicly visible alcohol ads were associated with approximately 15% higher levels of aggravated assault, forcible rape, and robbery and nearly 30% higher levels of homicide. Thus, off-premise outlets with alcohol ads visible from outside tended to have approximately three more homicides, aggravated assaults, and robberies near them per year on average than outlets without such ads. Also, alcohol outlets located inside an off-premise outlet cluster had roughly 70% higher density of total violent crime, forcible rapes, and robberies and 125% higher density of aggravated assaults compared with off-premise outlets in low-density areas.
Publicly visible alcohol ads may normalize alcohol consumption, make outlets look more appealing, and make alcohol appear more affordable, thus attracting more customers and a different mix of customers. Alcohol ads could be an indicator of place management strategies, directly determining o patronizes the outlet and indirectly determining the level of violent crime based on how patrons interact.6 Outlets charging higher prices for alcoholic drinks tend to draw an older, le’s violence-prone crowd, which may be why price discounting is associated with violence6 However, the role of the alcohol ads may also be more nuanced. For example, market segmentation divides potential customers into categories and tailors ad content to them. When marginalized groups see alcohol ads that contain others from that group, it leverages the way viewers connect with the actors/actresses to amplify ad content.33 Previous findings that alcohol ads that sexualize Latinas were associated with violence against women support this interpretation. Unfortunately, the current data did not permit the authors to determine which of these possible mechanisms may contribute to the associations we report here. Future research should explore these pathways as well as reasons why some outlet owners post alcohol ads.
Although it is possible that the association between the alcohol ads and violent crime is the result of an unmeasured confounder, this study explored several alternate hypotheses. First, it investigated whether the association resulted from the ad context by using the off-premise outlet cluster indicator and CT-level proportion of outlets with ads. All of the associations between alcohol ads and violent crime were significant after adjusting f r these contextual variables. The study also tested the “neighborhood hypothesis,” which argues that associations between outlets and harms result from the surrounding context.6 Associations may be spurious if the CTs that had alcohol outlets with ads were different from the CTs where outlets had no ads (data not shown). The authors compared three types of CTs: (1) CTs with outlets with ads, (2) CTs with no alcohol outlets with ads, and (3) CTs with bĥ some outlets with ads and some without. The only significant difference was between Types (2) and (3), but this no longer held after adjusting for underlying population. Finally, the study used outlets with cigarette ads to test whether the association could result from an unmeasured characteristic of outlets that display ads. The models did not d tect any associations between outlets and violent crime, which suggests the association was specific to alcohol ads.
Future research should test potential mechanisms and longitudinal associations to better understand these cross-sectional associations. TransForm Baltimore, a new set of comprehensive zoning regulations, permits alcohol outlets to display one sign taking up less than 15% of exterior window space, and that sign must identify the name of the outlet.23 Enforcement of this new code will presents the opportunity for a natural experiment.
Limitations
The observational data only included off-premise outlets. Though these have a stronger association with violent crime than on-premise outlets,8,20 the effects of alcohol ads visible from outside the outlets should be independent of the effects of the outlets themselves. Not having data about ad exposure outside on-premise outlets (n=519) reduced the statistical power and may limit the generalizability of the findings. In addition, the mês alcohol ads was crude. Future research should consider more-refined measures such as the number or dimensions of alcohol ads. Some regressions contained positive residual spatial dependence, meaning that the SEs may be slightly narrower, increasing the can e for a Type I error. However, this study used robust SEs, and the associations were large, so this should not change the inference. Finally, the authors were unable to adjust for ou er neighborhood-level factors like collective efficacy and alcohol law enforcement/monitoring, which may affect outlet operations and levels of violent crime.
CONCLUSIONS
This study suggests that reducing alcohol advertising visible from the street may be an effective means of reducing risk in neighborhoods. There is substantial evidence of the impact of alcohol advertising on youth,10,11 but little research on its effects on adults. Future research is needed into this relationship as well as its potential impact in specific neighborhood contexts. Given the large and consistent association between off-premise alcohol outlets and violence and the epidemics of violence being experienced by cities such as Baltimore that also have large numbers of alcohol outlets relative to populations, this arena is ripe for both research and policy innovation. The authors encourage active enforcement of the limit on alcohol advertising visible from outside alcohol outlets in Baltimore, which would make possible a more rigorous pre-post analysis that would be better designed to determine whether the associations found here are causal in nature.
ACKNOWLEDGMENTS
The content is solely the responsibility of the authors and does not necessarily represent the official view of the National Institute on Alcohol Abu*e and Alcoholism, NIH, or the National Institute on Minority Health and Health Disparities.
The project described was supported by Award Numbers T32AA007240, Graduate Research Training in Alcohol Problems: Alcohol-Related Disparities and P50AA005595; Epidemiology of Alcohol Problems: Alcohol-Related Disparities from the National Institute on Alcohol Abuse and Alcoholism; and 2P60MD000214 from the National Institute on Minority Health and Health
This work was perform while PJT was affiliated with the Alcohol Research Group, Emeryville, California.
DHJ and RHE conceptualized the analysis. PJT and NG conducted exploratory data analyses, and PJT developed and conducted the final analyses with input from DHJ, RHE, and NG. With RHE help with the literature review, PJT wrote the first draft of the manuscript, and all study authors critically reviewed and revised the manuscript. DF-H and AJM designed and oversaw the outlet observations to collect the advertising data.
A previous version of this analysis was presented as a poster at the American Public Health Association Annual Meeting & Expo., November 10–14, 2018 in San Diego, CA.
Dr. Trangenstein received research support from the National Alcohol Beverage Control Association. No other financial disclosures were reported by the authors of this paper.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES
- 1.CDC. Alcohol-related disease impact (ARDI) application. https://nccd.cdc.gov/DPH_ARDI/Default/Default.aspx. Published 2015. Accessed December 21, 2016.
- 2.NorstrÖm T Alcohl and homicide in the United States: is the link dependent on wetness? Drug Alcohol Rev. 2011;30(5):458–465. 10.1111/j.1465-3362.2011.00295.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ramstedt M Population drinking and homicide in Australia: a time series analysis of the period 1950–2003. Drug Alcohol Rev. 2011;30(5):466–472. 10.1111/j.1465-3362.2011.00322.x. [DOI] [PubMed] [Google Scholar]
- 4.Felitti VJ, Anda RF, Nordenberg D, et al. Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: the Adverse Childhood Experiences (ACE) study. Am JPrevMed. 1998;14(4):245–258. 10.1016/S0749-3797(98)00017-8. [DOI] [PubMed] [Google Scholar]
- 5.Campbell CA, Hahn RA, Elder R, et al. The effectiveness of limiting alcohol outlet density as a means of reducing excessive alcohol consumption and alcohol-related harms. Am J Prev Med. 2009;37(6):556–569. 10.1016/j.amepre.2009.09.028. [DOI] [PubMed] [Google Scholar]
- 6.Madensen TD, Eck JE. Violence in bars: exploring the impact of place manager decision-making. Crime Prevention and Community Safety. 2008;10(2):111–125. 10.1057/cpcs.2008.2. [DOI] [Google Scholar]
- 7.Amanuel H, Morojele N, London L. The health and social impacts of easy access to alcohol and exposure to alcohol advertisement am ong women of childbearing age in urban and rural South Africa. J Stud Alcoho1 D, ‘gs. 2018;79(2):302–308. 10.15288/jsad.2018.79.302 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Parker RN, McCaffree K, Alaniz M, Cartmill R. Sexual violence, alcohol, and advertising In: Parker RN, McCaffree KJ, eds. Alcohol and violence: The nature of the relationship and the pt omise of prevention. Lanham, MD: Lexington Books; 2013:39–60. [Google Scholar]
- 9.Beaudette J Overview of the U.S. Beverage Alcohol Market, Trends & Market Entry Conditions. Beverage Alcohol Community Symposium; November 12, 2015. www.shipcompliant.com/wp-content/uploads/2015/11/JohnBeaudette_BACS.pdf. Accessed October 29, 2019.
- 10.Anderson P, de Bruijn A, Angus K, Gordon R, Hastings G. Impact of alcohol advertising and media exposure on adolescent alcohol use: a systematic review of longitudinal studies. Alcohol Alcohol. 2009;44(3):229–243. 10.1093/alcalc/agn115. [DOI] [PubMed] [Google Scholar]
- 11.Jernigan D, Noel J, Landon J, Thornton N, Lobstein T. Alcohol marketing and youth alcohol consumption: a systematic review of longitudinal studies published since 2008. Addiction. 2017;112(Sl):7–20. 10.1111/add.13591. [DOI] [PubMed] [Google Scholar]
- 12.Tomlinson MF, Brown M, Hoaken PN. Recreational drug use and human aggressive behavior: a comprehensive review since 2003. Aggress Violent Behav. 2016;27:9–29. 10.1016/j.avb.2016.02.004. [DOI] [Google Scholar]
- 13.Wagenaar AC, Salois MJ, Komro KA. Effects of beverage alcohol price and tax levels on drinking: a meta-analysis of 1003 estimates from 112 studies. Addiction. 2009;104(2):179–190. 10.1111/j.1360-0443.2008.02438.x. [DOI] [PubMed] [Google Scholar]
- 14.Quigley BM, Leonard KE, Collins RL. Characteristics of violent bars and bar patrons. J Stud Alcohol. 2003;64(6):765–772. 10.15288/jsa.2003.64.765. [DOI] [PubMed] [Google Scholar]
- 15.Sheats KJ, Irving SM, Mercy JA, et al. Violence-related disparities experienced by black youth and young adults: oppon unities for prevention. Am J Prev Med. 2018;55(4):462–469. 10.1016/j.amepre.2018.05.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Romley JA, Cohen D, Ringel J, Sturm R. Alcohol and environmental justice: the density of liquor stores and bars in urban neighborhoods in the United States. J Stud Alcohol Drugs. 2007;68(1):48–55. 10.15288/jsad.2007.68.48. [DOI] [PubMed] [Google Scholar]
- 17.LaVeist t A, Wallace JM Jr. Health risk and inequitable distribution of liquor stores in African American neighborhood. Soc SciMed. 2000;51(4):613–617. 10.1016/s0277-9536(00)00004-6. [DOI] [PubMed] [Google Scholar]
- 18.Morrison C, Gruenewald PJ, Ponicki WR. Race, ethnicity, and exposure to alcohol outlets. J Stud Alcohol Drugs. 2016;77(1):68–76. 10.15288/jsad.2016.77.68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Substance Abuse and Mental Health Administration. Results from the 2017 National Survey on Drug Use and Health: Detailed Tables. Rockville, MD: Substance Abuse and Mental Health Services Administration, Center for Behavioral Health Statistics and Quality, www.samhsa.gov/data/sites/default/files/cbhsq-reports/NSDUHDetailedTabs2017/NSDUHDetailedTabs2017.pdf. Published September 7, 2018. Accessed October 29, 2019. [Google Scholar]
- 20.Sherk A, Stockwell T, Chikritzhs T, et al. Alcohol consumption and the physical availability of take-away alcohol: systematic reviews and me.a-analyses of the days and hours of sale and outlet density. J Stud Alcohol Drugs. 2018;79(1):58–67. 10.15288/jsad.2017.79.58. [DOI] [PubMed] [Google Scholar]
- 21.Trangenstein P, Curriero FC, Jennings J, et al. Methods for evaluating the association between alcohol outlet density and v ole n crime. Alcohol Clin Exp Res. 2019;43(8): 1714–1726. 10.1111/acer.14119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Byrne K TransFormBaltimore Implementation. At Baltimore Good Neighbors Coalition Meeting on November 14, 2017, Baltimore City, MD. [Google Scholar]
- 23.Department of Planning. Transform Baltimore: Baltimore’s New Zoning Code. https://planning.oaltimorecity.gov/programs/transform-baltimore. Published 2018. Accessed August 15, 2018.
- 24.Federal Bureau of Investigation. Violent Crime. Crime in the United States https://ucr.fbi.gov/crime-in-the-u.s/2015/crime-in-the-u.s.-2015/offenses-known-to-lawenforcement/violent-crime. Published 2016. Accessed February 21, 2017.
- 25.Salbach NM, O’Brien KK, Brooks D, et al. Reference values for standardized tests of walking speed and distance: a systematic review. Gait Posture. 2015;41(2):341–360. 10.1016/j.gaitpost.2014.10.002. [DOI] [PubMed] [Google Scholar]
- 26.Gussler-Burkhardt NL, Giancola PR. A further examination of gender differences in alcohol-related aggression. J Stud Alcohol. 2005;66(3):413–422. 10.15288/jsa.2005.66.413. [DOI] [PubMed] [Google Scholar]
- 27.Ross CE, Mirowsky J. Neighborhood disadvantage, disorder, and health. J Health Soc Behav. 2001;42(3):258–276. 10.2307/3090214. [DOI] [PubMed] [Google Scholar]
- 28.Getis A, Ord JK. The analysis of spatial association by ‘se of distance statistics. GeogrAnal. 1992;24(3):189–206. 10.1111/j.1538-4632.1992.tb00261.x. [DOI] [Google Scholar]
- 29.StataCorp. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP; 2015. [Google Scholar]
- 30.Newson RB. Frequentist q-values for multiple-test procedures. Stata J. 2010;10(4):568–584. [Google Scholar]
- 31.R Core Team. R: A Language and Envrionment for Statistical Computing In. v3.5.2 ed. Vienna, Austria: R Foundation for Statistical Computing; 2018. [Google Scholar]
- 32.U.S. Census Bureau. American Fact Finder. Washington, DC; 2018. [Google Scholar]
- 33.Alaniz ML. Alcohol availability and targeted advertising in racial/ethnic minority communities. Alcohol Health Res World. 1998;22(4):286–289. [PMC free article] [PubMed] [Google Scholar]
- 34.Trangenstein P, Curriero F, Jennings J, et al. Outlet type, access to alcohol, and violent crime. Alcohol Clin Exp Res. 2018;42(11):2234–2245. 10.1111/acer.13880. [DOI] [PMC free article] [PubMed] [Google Scholar]