The oversaturation of alcohol outlets continues to be a major public health problem that impacts the health and safety of communities (Campbell et al., 2009; Hippensteel et al., 2018). Increased alcohol outlet density has been associated with higher overall consumption and associated problems including drinking and driving, traffic crashes, pedestrian injuries, violence, suicide, and reduced life expectancy (Furr-Holden et al., 2016, 2019; Jennings et al., 2014; Popova et al., 2009). Alcohol outlet density and availability are also associated with adolescent purchasing of alcohol (Milam et al., 2014; Rowland et al., 2015) and youth violence (homicide, gunshot injuries, physical fighting, etc.) (Dahlberg, 1998). Indeed, high alcohol outlet density may pressure businesses to lower prices of alcohol due to the competitive market, and thus lead to more sales (Shortt et al., 2015). Interestingly, previous studies have also found that alcohol outlets are associated with neighborhood disorder and disadvantage, where the highest density of alcohol outlets is commonly found in the most deprived neighborhoods (LaVeist & Wallace, 2000; Morland et al., 2002; Pollack et al., 2005).

Generally, alcohol outlets are categorized into two types: on-premise and off-premise. On-premise alcohol outlets refer to locations that sell alcohol for on-site consumption (e.g., restaurants, breweries, wineries, etc.), whereas off-premise refers to locations where alcohol is bought but consumed elsewhere (e.g., from grocery stores, convenience stores, gas stations, etc.) (Guide for Measuring Alcohol Outlet Density, n.d.). Moreover, neighborhood characteristics such as income, racial/ethnic population and urbanicity tend to be indicators of the density and types of alcohol outlets in the area. Trangenstein and colleagues found that areas comprised of low-income populations were more likely to have off-premise alcohol outlets, while on-premise alcohol outlets were more common in high-income areas (Trangenstein et al., 2019). LaVeist and Wallace also found liquor stores were often located in low-income and predominantly African American communities (LaVeist & Wallace, 2000). Off-premise alcohol outlets such as liquor stores tend to have larger quantities of alcohol available than on-premise establishments and offer chilled alcoholic beverages (ready to drink) (LaVeist & Wallace, 2000). In particular, off-premise alcohol outlets have been associated with increased risk of unhealthy alcohol consumption (Halonen et al., 2013) and high rates of violence (Furr-Holden et al., 2019; Jennings et al., 2014). Furthermore, alcohol outlet density is higher in urban areas (Hay et al., 2009). Specifically, residents in more-deprived urban areas have easy access to alcohol, as alcohol outlets are often located within walking distance from residential areas (Hay et al., 2009). Unfortunately, access to alcohol outlets in disadvantaged communities makes vulnerable populations even more susceptible to alcohol-related problems. Further, alcohol outlet density in disadvantage areas may also be related to the lack of human and financial resources available in the community to accurately maintain data on the number of outlets, regulations, and most importantly, how these alcohol outlets are impacting the health of residents. Although alcohol outlets can be recognized as problematic, the negative effects of alcohol often remain salient in communities (Furr-Holden et al., 2019).

Researchers have found a decrease in alcohol-related problems when alcohol density is reduced. A community-based intervention in California that resulted in a decrease in alcohol outlets and improved enforcement of drinking and driving laws, found a subsequent decrease in alcohol-related injuries, specifically, assault-related emergency room visits and night-time motor vehicle crashes (Holder et al., 2000). Campbell and colleagues also suggested that a decrease in alcohol outlets could decrease excessive alcohol consumption as the distance between alcohol outlets would increase and thus make alcohol less accessible (Campbell et al., 2009). Even a modest reduction of alcohol outlets has shown to reduce violent crimes in neighborhoods (Zhang et al., 2015). Researchers have similarly examined alternative methods of reducing alcohol consumption, such as reducing the hours stores are allowed to sell alcohol (Popova et al., 2009; Wilkinson et al., 2016), regulating alcohol prices (Ramstedt, 2002), and banning alcohol sales (Campbell et al., 2009). These interventions have been associated with reduced rates of violence (Wilkinson et al., 2016), reduced motor vehicle collisions and alcohol intoxication among young people (age 10–19) (Ramstedt, 2002), and reduced alcohol-related harms and excessive consumption (Campbell et al., 2009).

Because so much of this research hinges on what we understand of and how we define alcohol outlet density, the availability of local, state, and national data for defining alcohol outlet density—and the corresponding way each defines said data—has important implications for policies and interventions. To better understand the impact of alcohol outlet density and reduce alcohol-related problems, researchers have started to examine the quality of the data available on alcohol outlets. Ponicki et al. examined California-based state records on alcohol outlets and were unable to locate 9% of the establishments listed, indicating many outlets were closed (Ponicki et al., 2014). Other researchers have explored alternative measures of alcohol outlet density outside of the state and local databases. Matthews et al. compared estimates of counts of alcohol outlets from local and state listings in Michigan (focusing on communities surrounding college campuses) to those on the U.S. Census Bureau’s ZIP Code Business Patterns (ZIP-BP) database and concluded that the ZIP-BP dataset adequately estimates alcohol outlet densities (Matthews et al., 2011). Additionally, Carlos and colleagues looked at alcohol outlets in California using commercial business lists and concluded that business lists were a reliable proxy for licensed alcohol outlets (Carlos et al., 2017).

In an effort to find reliable data on alcohol outlet density, researchers have noted a few limitations, such as accurate reporting of license type and operating characteristics (as some outlets act more as bars but are reported as restaurants because they serve food) (Ponicki et al., 2014) and lack of universal protocol to classify or maintain alcohol outlet data (Matthews et al., 2011). This study’s aims are to 1) examine discrepancies between alcohol outlet listings (local, state, and national databases) in three states; 2) validate state databases of alcohol licenses available online to better understand how the level at which alcohol outlet data is collected affects its accuracy; and 3) examine how neighborhood characteristics (i.e., income, majority racial population, urbanicity) are associated with discrepancies between alcohol outlet databases. Identifying discrepancies between alcohol outlet databases is essential to accurately inform community interventions and public health initiatives.

To measure the validity of state databases of alcohol licenses, data was collected from three states with varying levels of alcohol control: Maryland, Oregon, and Wisconsin. This included data from select counties in each state (for a local comparison) and the 2016 ZIP-BP data collected by the U.S. Census (for a national comparison). We also obtained race and ethnicity data from the 2010 U.S. Census; other demographic and population information by ZIP code was obtained from the 2012–2016 American Community Survey. No consent process was used as this study did not use human participants.

The number of alcohol outlets per 1000 residents by state and the level from which the data came from (local, state, or national database) and other demographics by ZIP code are given in Table 1. The average density of alcohol outlets (per 1000 residents) ranged from 0.71 to 2.17 in Maryland, 1.65 to 5.17 in Wisconsin, and 1.09 to 1.22 in Oregon based on different reporting resources.

Table 2 shows the Interclass Correlation Coefficient (ICC) comparing national, state, and local databases for alcohol outlets. The ICC estimates were calculated using a fixed-effect method with a single measure. The ICC between state and national databases ranged from 0.83 to 0.93 for three states. The highest level of agreement was found in Maryland (ICC = 0.93; ICC = 0.83 and ICC = 0.84 for Oregon and Wisconsin, respectively), Figure 1 shows the scatterplot of alcohol outlets in three states between the national and state databases. For local data reports, we were only able to collect enough detailed alcohol outlet counts from local Maryland sources. The ICC for Maryland comparing local databases to the national and state databases was 0.91 and 0.97 respectively indicating the local reporting in Maryland is similar to state reporting.

Table 3 shows the GEE analysis results identifying the factors affecting state/national alcohol outlet reports for all three states. The results show that wealthier areas (>200% Poverty line) have significantly lower alcohol outlets (Incidence Rate = 0.74, p < 0.01) while poorer areas (below the poverty line) had significantly more alcohol outlets (IR = 3.54, p <0.01). The percent of white population and urban/rural area status did not have a significant effect on the count of alcohol outlets. Alcohol outlet density (based on population) at the level of ZIP code for Oregon and Wisconsin was significantly higher compared with Maryland (IR = 2.60 and 2.08 respectively, p < 0.01). In addition, in all three states, the number of alcohol outlets reported in national databases was significantly lower compared to state databases (IR = 0.53, 0.48 and 0.70 for MD, OR, and WI respectively, p < 0.01). Table 5 shows the analysis results by each state separately. Poverty status still had a significant effect on alcohol outlet counts (such that poorer areas had more alcohol outlets) in Wisconsin and Oregon, but not in Maryland. The lack of an association in Maryland might be due to the control policy in the state. In Maryland, areas with ≤ 50% white population had significantly fewer alcohol outlets (IR = 0.66, p < 0.01). In Wisconsin, meanwhile, areas with ≤ 50% white population or <50% urban area had significantly fewer alcohol outlets (IR=0.54 and 0.68 respectively, p ≤ 0.01).

Table 4 shows the GEE regression results for the discrepancy of alcohol outlets using national and state databases after aggregating data from all three states. Areas with ≤ 50% or 51–80% of white population had significantly fewer discrepancies in alcohol outlets (state vs. national) comparing with areas with >80% white population (IR = 0.77 and 0.59 respectively, p < 0.01). Urban areas had significantly fewer discrepancies compared to rural areas (IR = 0.38, p < 0.01). Areas with higher income levels (where the median income was >200% of the poverty level) had significantly fewer discrepancies (IR= 0.78, p < 0.01) compared with medium income areas (100%−200% above poverty level), while poor areas (where the median income was below the poverty level) had much higher discrepancies (IR=4.99, p < 0.01). In addition, we found that Oregon and Wisconsin had more reporting discrepancies of alcohol outlets between state and national databases compared to Maryland (IR = 1.59 and 1.89 respectively, p < 0.01). Table 6 shows the analysis results by each state separately. Two noteworthy similarities existed. In Maryland, low-income areas were not significantly different than medium-income areas. In Oregon, meanwhile, high-income areas were not significantly different than medium-income areas. A map to visually compare the relative number of differences per person living in an area is presented in Figure 2.

This study compared the number and density of alcohol outlets using national, state, and local alcohol outlet databases. Overall, the national databases appeared to under-estimate the number of alcohol outlets. National databases only over-estimated alcohol outlets in higher populated areas. The state databases had some inconsistencies, where under-estimation of alcohol outlets was seen in areas with higher income and higher population and over-estimation was seen in areas with a higher white population. When compared to local databases, state databases also seemed to over-estimate the number of alcohol outlets in the area.

Many studies have obtained alcohol outlet data through a variety of administrative sources (e.g., liquor license boards) (Goldstick et al., 2016; Haley et al., 2017; Milam et al., 2016), Google (Chen et al., 2019), and national databases (Freisthler & Wolf, 2016). Only a few studies, however, have examined the accuracy of alcohol outlet data and the process to obtain such data (Carlos et al., 2017; Matthews et al., 2011; Ponicki et al., 2014). These studies are necessary, for example, Ponicki et al. studied the California-based state records on alcohol outlets and concluded that state-level data on alcohol outlets often include outlets that are no longer open (Ponicki et al., 2014). Alcohol outlet databases need to represent accurate data and be easily accessible, as this data is crucial to inform policies and create awareness on a variety of alcohol-related problems. Many studies have used alcohol outlet data to examine associations between alcohol outlet density and a variety of physical and behavioral health outcomes, crime, and violence (Furr-Holden et al., 2019; Goldstick et al., 2016; Jennings et al., 2014; Milam et al., 2014; Nesoff et al., 2018; Popova et al., 2009). In addition, alcohol outlet oversaturation is more commonly seen in minority-majority communities, adding to the urgency of organized data to help inform efforts to reduce health inequities (Hippensteel et al., 2018; Trangenstein et al., 2019).

The discrepancies between local, state, and national alcohol outlet listings noted in this study may have impacted the results of previous studies using alcohol outlet data by under- or over-estimating the problem. For example, studies have examined how alcohol outlet density is associated with consumption (Sherk et al., 2018), life expectancy (Furr-Holden et al., 2019; Gonzales et al., 2014), and violence (Furr-Holden et al., 2016; Livingston, 2011). These studies rely on accurate data to understand public health issues and make appropriate recommendations for future research and policy changes.

This study has a few limitations that warrant discussion. First, limited data were available on Maryland’s rural counties, limiting what we can infer from these locations (both concerning internal and external validity). When aiming to retrieve alcohol outlet data, a few contacts referred us back to the state listings, and some did not respond even after repeated attempts. Non-response may be associated with discordance between their records and the state listings or a lack of updated data. Also, data obtained was only from three states (an alcohol control state, a semi/mixed-control state, and a non-control state). In addition, the alcohol outlets were not visited in-person to confirm they were still in business. For instance, Furr-Holden and colleagues obtained data for off-premise alcohol outlets in Baltimore and found 22% of alcohol outlets were chronically closed. Without further investigation, those outlets would have been included in the study as active outlets (Furr-Holden et al., 2019); this further explains the importance of keeping alcohol outlet data updated. Similar challenges are seen with tobacco outlet data, and researchers often use secondary data sources (D’Angelo et al., 2014). Finally, the NAICS codes used in this analysis may not be an exhaustive list of codes for alcohol outlets which could impact the accuracy using this national database to estimate alcohol availability.

Future research should continue to examine discrepancies between local, state, and national alcohol outlet databases. In addition, the processes of obtaining alcohol outlet data should be documented to improve access to data.