Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jun 23.
Published in final edited form as: Health Place. 2022 May 20;75:102815. doi: 10.1016/j.healthplace.2022.102815

Draining the tobacco swamps: Shaping the built environment to reduce tobacco retailer proximity to residents in 30 big US cities

Todd B Combs a,*, Joseph T Ornstein b, Veronica L Chaitan a, Shelley D Golden c,d, Lisa Henriksen e, Douglas A Luke a
PMCID: PMC10288515  NIHMSID: NIHMS1905604  PMID: 35598345

Abstract

Combining geospatial data on residential and tobacco retailer density in 30 big US cities, we find that a large majority of urban residents live in tobacco swamps – neighborhoods where there is a glut of tobacco retailers. In this study, we simulate the effects of tobacco retail reduction policies and compare probable changes in resident-to-retailer proximity and retailer density for each city. While measures of proximity and density at baseline are highly correlated, the results differ both between effects on proximity and density and across the 30 cities. Context, particularly baseline proximity of residents to retailers, is important to consider when designing policies to reduce retailer concentration.

Keywords: Tobacco retailer density, Tobacco retailer proximity, Retail tobacco policy

1. Introduction

As in other countries, state and local governments in the United States (US) have the responsibility to reduce the public health harms of tobacco by enacting policies that affect the tobacco retail environment (also known as ‘point-of-sale’ policies). One goal of such regulations is to reduce the overall number and concentration of retailers that sell tobacco (Myers et al., 2015). We characterize areas with high tobacco retailer concentration as tobacco swamps, i.e., as opposed to food or pharmacy deserts. Tobacco retailer concentration can be described in terms of density (number of retailers per square meter, mile or population, e.g., retailers per 1000 people), or proximity (the distance between residents and retailers). Examples of retail reduction policies include a cap on the number of tobacco retailer licenses available, as in San Francisco (2014) and New York City (2018), and those that create buffer zones between schools and retailers or between tobacco retailers themselves, as in Chicago (2013) and Philadelphia (2016) (Bright Research Group, 2016; City of New York, 2021; Tobacco Control Legal Consortium, 2016; Lawman et al., 2020; Vyas et al., 2020). However, since most policies that affect the built environment for tobacco were only recently implemented and take time to achieve meaningful reductions, the evidence base is limited for which policies achieve their goals, how they actually work, and which policies work better than others.

Tobacco retail reduction policies may influence tobacco use through at least two mechanisms: increasing the distance that a consumer needs to travel to purchase tobacco, thus incurring additional costs in travel expenses and time; and reducing the frequency with which people observe environmental cues to use tobacco, like retail marketing (Marsh et al., 2021). However, policies that reduce the density of retailers may have little effect on the proximity of residents to retailers. That is, reduction of retailer density does not necessarily increase the distance that consumers must travel to purchase tobacco. A simple thought experiment illustrates this point. Suppose that there are four retailers at each of the four corners of a street crossing and two retailers on the next corner (Fig. 1, top). If a retail reduction policy removed tobacco sales from three of the four on the first corner, the retail density in the area would have decreased by 50%, but the distance that the consumer must travel (proximity) would not increase at all (Fig. 1, middle). Only removing sales from all four stores results in farther distances to travel in this scenario (Fig. 1, bottom).

Fig. 1.

Fig. 1.

Open in a new tab

Illustration of the impact of density reduction on proximity of residents to retailers. Policies that address density do not necessarily impact residents’ proximity to retailers.

Notwithstanding the claims of a historic cigarette advertising campaign (Fig. 2),(Stanford, 2021) very few consumers are willing to travel a significant distance – with any regularity – to purchase goods. Evidence from behavioral, transportation, and economic studies suggests that the majority of consumers make purchases close to home (Consumer, 2021; US Department of Transportation, 2021). For example, urban dwellers typically travel 8–10 min for frequent purchases and about 80% of trips to purchase goods made in metro areas are less than 15 min. This is relevant because behavioral health studies find that cessation attempts are more likely to be successful in neighborhoods where tobacco retailer concentration is relatively low (Pearce et al., 2016; Cantrell et al., 2015; Lee et al., 2021). Policies that aim to reduce concentration through density reductions should also consider the potential impact on residents’ proximity to retailers. Increased distance between residents and retailers may lead to behavioral changes in travel time and opportunity costs to obtain tobacco. To date, few empirical studies have focused on the distinction between tobacco retailer density and resident-to-retailer proximity.

Fig. 2.

Fig. 2.

Open in a new tab

I’d Walk a Mile for a Camel – 1970s advertising campaign (Source: Stanford Research into the Impact of Tobacco Advertising or SRITA).

To help fill this gap, this study focuses on measures of tobacco retailer proximity and how they compare to density metrics (Lawman et al., 2020; Shortt et al., 2016; Cantrell et al., 2016; Kirst et al., 2019; Glasser and Roberts, 2021). Recent evidence suggests that smoking prevalence in areas where residents live farther away from tobacco retailers is lower than in areas where residents are nearer to retailers (Lee et al., 2021). Proximity measures allow us to investigate potential changes to distances that current and potential smokers would need to travel to purchase tobacco products following a policy intervention. This focus also allows us to compare impacts on residents’ proximity to tobacco retailers across different types and strengths of retail reduction policies, as potentially implemented in different cities, and draw conclusions about which policies might have the strongest public health impact. Furthermore, our study is unique in tobacco control research in its use of synthetic populations for which realistic home and work locations are generated, paired with actual tobacco retailer location data, allowing us to calculate various measures of resident-to-retailer proximity before and after policy interventions.

Our primary goals are to 1) describe the existing tobacco retail environment in 30 big US cities, 2) explore the relationship between tobacco retailer density and proximity to residents, 3) simulate the effect of retail reduction policies on residents’ proximity to tobacco retailers and retailer density, and 4) illustrate the context dependency of potential policy impacts across 30 cities through comparisons of different urban environments and baseline measures of resident-to-retailer proximity. We achieve these goals by quantifying the median travel distance required to purchase tobacco and the percentage of residents living within 1 km (or a 10-min walk) of a tobacco retailer, as is common in the literature on built environments and health (Hou et al., 2021; Boisjoly et al., 2018; Wälty, 2018), as well as retailer density (retailers per km (Bright Research Group, 2016)). We then simulate a series of retail reduction policies (i.e., licensing caps, school-to-retailer buffers, and retailer-to-retailer buffers), estimating their likely effect on tobacco resident-to-retailer proximity as compared to retailer density. Results of this study are important for multiple audiences: 1) tobacco control policy researchers and evaluators; 2) future modeling investigators focusing on the role of the built environment in public health; 3) community policymakers making decisions about retail policy development and implementation; and 4) residents of the modeled cities.

2. Methods

Our study uses geospatial simulations to model the potential impact of various tobacco retail reduction policies in 30 US cities. This work is part of a multi-site center funded by the National Cancer Institute at the National Institutes of Health called ASPiRE (Advancing Science and Practice in the Retail Environment, #P01-CA225597) and is a collaboration between researchers, public health practitioners and organizations, and legal experts. The goal of ASPiRE is to build a strong scientific evidence base for effective policies in the retail environment to help reduce tobacco use, tobacco-related disparities, and the public health burden of tobacco, including cancer.

2.1. Communities

The study draws on data from 30 big US cities. At the beginning of the study, 27 cities were members of the Big Cities Health Coalition (BCHC) (Big Cities Health Coalition, 2017), two cities were then added for representation in the southeast (Memphis, Tennessee, and New Orleans, Louisiana), and Providence, Rhode Island, was added for early adoption of novel retail policies. Local tobacco control program managers and other public health professionals from these cities make up the Community Advisory Board (CAB) for the ASPiRE Center and represent diverse communities where a variety of retail reduction policies are being planned, implemented, and evaluated.1 The cities also vary in terms of 2018 adult smoking prevalence from 10% (Seattle) to 30% (Cleveland), with an average of 18% (SD: 4.5%), and are located in states with 2021 Tobacco Prevention and Cessation Funding grades ranging from B (California) to F (Florida) (American Lung Association, 2021).

2.2. Data

The geospatial simulations drew from multiple data sources: commercial retail databases (Dun & Bradstreet, 2021), state and local tobacco retailer licensing lists (obtained directly from localities and states), US Census data (US Census Bureau, 2021), synthetic populations (Gallagher et al., 2018), and public school location data (ArcGIS. Public Schools, 2022), all from 2019. For tobacco retailers in three cities without licensing (Charlotte, Detroit, Memphis), we obtained a national business database compiled by Data Axle (ReferenceUSA) and Dun & Bradstreet (Dun & Bradstreet, 2021; Axle, 2021). These data include US businesses along with their names, locations, and industry classification code (NAICS). Using these NAICS codes, we filtered the dataset for each of the 30 cities to include only those businesses that are likely to sell tobacco products: convenience stores, gas stations, grocery stores, discount warehouses, liquor stores, pharmacies, and tobacco specialty shops (D’ Angelo et al., 2014; Golden et al., 2021).

To compare the locations of retailers with the locations of households, we generated a synthetic population for each city. This technique, previously developed to inform transportation microsimulations (Barthelemy and Toint, 2013) and infectious disease epidemiology(Levy et al., 2016), produces a set of synthetic households whose locations and demographics reflect the observed distributions of those characteristics in the Census Public Use Microdata (PUMS). Crucially for our purposes, the assigned locations of these households yield an accurate and valid population density gradient for each city, and characteristics like race and ethnicity, income, and age are spatially realistic (Gallagher et al., 2018). We generated synthetic populations from the PUMS data representing 2% of the adult (21+) population. The supplemental appendix contains more information on the synthetic populations and compares them to real-world data. The process of synthesizing the populations assigns a set of latitude-longitude coordinates for each person or family as a residence. This is important because household-level spatial data on residents is unavailable due to privacy concerns.

Combining the two datasets, we computed the distance that each of the synthetic residents must travel to reach the nearest tobacco retailer. Distances were calculated using the GIS spatial analysis techniques contained in the sp and sf packages in R (Bivand et al., 2013; Pebesma, 2018; R Core Team, 2021). Transportation mode data for each of the 30 cities were obtained from the five-year estimates of the American Community Survey data (US Census Bureau, 2021). From this we derived two city-level measures of tobacco retailer proximity: (1) the median (Euclidean) distance that residents must travel from home to reach the nearest tobacco retailer and (2) the percentage of residents who live within 1 km of a tobacco retailer. The measure for tobacco retailer density that we used for comparison is number of retailers per square kilometer, as retailers per land area was more commonly used in a recent meta-analysis of the tobacco retailer density and proximity literature (Lee et al., 2021).

2.3. Policies

We selected policies to test, along with their features (e.g., size of retailer buffers), based on characteristics of implemented policies or those in the planning stages in US communities. Our policy implementation simulations fell into two categories: (1) license caps and (2) buffers between tobacco retailers and either schools or other tobacco retailers.

2.4. License caps

In a related 2019 study, license caps were the most commonly implemented retail reduction policy in the 30 ASPiRE cities (ASPiRE Center, 2019). These policies put a moratorium (i.e., cap) on tobacco retail licenses and limit the conditions to keep or transfer an existing license. These caps can then be lowered over time (i.e., winnowed), further reducing the number of retailers. In our simulations, these cap and winnow policies are modeled by randomly removing licenses from retailers until a percentage of the total baseline licenses is achieved. The percentages modeled range from a 10%–90% reduction of the status quo at baseline. Real-world examples of license caps have shown reductions in tobacco retailer density: San Francisco achieved an 8% reduction in initial density within the first ten months after limiting number of licenses available, and Philadelphia achieved a 20% reduction in three years after its 2012 license cap (Bright Research Group, 2016; Lawman et al., 2020; Kong and King, 2020).

2.5. Retailer and school buffers

Governments may prohibit tobacco sales within a certain distance (buffer) of schools or other tobacco retailers, and/or make these conditions to prohibit the transfer of an existing tobacco retail license. The buffer distances in our model are Euclidean distances of 300 m and 600 m, based on real-world examples like those passed in Newburgh, New York, (1000 feet or about 300 m from a school) and unincorporated Santa Clara County, California, (1000 feet or 300 m from a school and 500 feet or 150 m from another retailer), and stronger policies than currently implemented ones like a 600 m (about 2000 feet) buffer (Public Health Tobacco Policy Center, 2020; Santa Clara County Public Health, 2020).

2.6. Simulation modeling

2.6.1. Development of geospatial simulation

Once we constructed our distributions of households, retailers, and travel distances, we calculated baseline measures of retailer density and resident-to-retailer proximity. We then simulated a series of retail policy implementations for each city, and measured the resulting changes in measures of proximity i.e., median distance to nearest retailer and percentage of residents living within 1 km of a retailer (other distances are available in the supplemental appendix for sensitivity analyses). For comparison, we also calculated changes in tobacco retailer density. All simulations were performed in R: A language and environment for statistical computing (version 4.0) (R Core Team. R, 2019).

2.6.2. Development of tobacco swamps dashboard

To facilitate interpretation of the geospatial simulations, and to engage the members of the CAB in exploration of potential effects of retailer policy implementation in their own communities, we developed a Tobacco Swamps interactive dashboard. It uses the data described above and is programmed using the R Shiny environment (Chang et al., 2021). In addition to effects on proximity to retailers, the dashboard can be used to explore how different policies affect tobacco-related disparities in each community. The dashboard is publicly available: https://aspirecenter.org/tobacco-swamps/.

3. Results

3.1. Baseline geospatial model: Status quo in 30 major US cities

In land area, the cities are sized from Providence, Rhode Island, (48 km (Bright Research Group, 2016)) to Houston, Texas, (1553 km (Bright Research Group, 2016)) with a median of 363 km (Bright Research Group, 2016). At baseline, the percentage of residents within 1 km of a retailer ranged from 98.8% (Boston, Massachusetts) to 61.9% (Kansas City, Missouri). The median resident-to-retailer distance ranged from 0.14 km (New York City, New York) to 0.72 km (Kansas City, Missouri), and tobacco retailer density was highest in Miami, Florida, (13.0/km2) and lowest in Kansas City, Missouri (0.7/km2). Both measures of proximity are strongly correlated with the logarithm of retailer density (Fig. 3). As expected, the percentages of residents within 1 km of a retailer is positively associated with density, while the median resident-to-retailer distance is inversely related to density. Bivariate regression models returned R2 values of 0.75 and 0.86, respectively.

Fig. 3.

Fig. 3.

Open in a new tab

Relationship of tobacco retailer density to a) the percentage of residents within 1 km of a retailer, and b) median resident-to-retailer proximity to for 30 large US cities, at baseline.

3.2. Retail reduction simulations

3.2.1. License caps

To represent a cap-and-winnow policy, we randomly removed retailers until their total number falls below a predetermined cap. The cities in Fig. 4 are ordered from highest to lowest baseline percentage of residents within 1 km of a tobacco retailer (top to bottom, left to right). The curves show the changes to the percentage of residents who live within 500 m, 1 km, 1500 m, and 2 km of a tobacco retailer. The effect of random removal on retail proximity is nonlinear, especially for the shorter distances. However, even removing 50% of retailers in this fashion barely affects the proportion of residents who live in close proximity – within 500 m or 1 km – to a tobacco retailer. Only when 70–80% of retailers are removed do we begin to see a significant decrease in the percentage of residents living within 1 km of a retailer in most cities. For cities that start out with higher percentages of residents in close proximity to retailers, e.g., New York City, Miami, Philadelphia, and others toward the top of the figure, decreases in percentages of residents living within 1 km of a retailer are much smaller. Larger decreases for all distances are found in cities with lower baseline percentages of residents living within 500m or 1 km of a retailer, e.g., Memphis, Tennessee; San Diego, California; Kansas City, Missouri; and others near the bottom of the figure.

Fig. 4.

Fig. 4.

Open in a new tab

Predicted reductions of residents living within 500 m, 1 km, 1500 m, and 2 km of a tobacco retailer with license caps at 10–90% of baseline. Cities are ordered from highest baseline tobacco retailer concentration to lowest, top to bottom.

3.2.2. Buffers

We also considered policies that establish buffers between schools and tobacco retailers (school-to-retailer or S2R) or between tobacco retailers themselves (retailer-to-retailer or R2R) by stipulating a minimum distance between them. Fig. 5 (left) shows the changes in the 30 cities for the percentage of residents living within 1 km of a retailer for the two buffer policies and a license cap that removes tobacco sales from 50% of stores for comparison.

Fig. 5.

Fig. 5.

Open in a new tab

Reductions in resident-to-retailer proximity and tobacco retailer density under three retailer reduction policies. Cities are ordered from highest baseline retailer concentration to lowest, top to bottom.

In Fig. 5 cities are again ordered from highest to lowest (top to bottom, left to right) in baseline percentage of residents living within 1 km of a tobacco retailer. Longer bars represent higher relative percentage decreases in this metric after policy implementation. Randomly removing half of retailers through a license cap performs slightly better than the buffer policies at reducing residents in close proximity to retailers in some cities, especially in those cities toward the bottom of the plot, like Sacramento, Charlotte, and Fort Worth. By comparison, school-to-retailer buffers of 600 m appear to decrease the percentage of residents living within 1 km of a tobacco retailer more than the license cap, especially for cities near the top of the plot. For example, a 600 m S2R buffer is estimated to have large decreases of 50% in New York City and by over 25% in Washington DC, where the baseline percentages of residents living within 1 km of a retailer are relatively high at 99% and 97%, respectively. School-to-retailer buffers have the largest effects on residents’ proximity to retailers for cities with the highest concentration of retailers at baseline; for cities starting with lower retailer concentration, school-to-retailer buffers and license caps have similar effects. The appendix contains Figure A1, which illustrates sensitivity analyses at different proximities (500 m, 1 km, 1500 m, 2 km). The same trends are present at each level with regard to baseline retailer concentration. However, reductions to the percentage of residents who live within 500 m are markedly different – more than double – the changes to those living within 1 km.

3.2.3. Comparative policy impacts on proximity and density

In our simulations, retailer-to-retailer buffers have little to no impact on residents’ proximity to tobacco retailers. In Fig. 5 (right), in every community, a 600 m R2R buffer eliminates over 50% of tobacco retailers, but the percentage of residents living within 1 km of a retailer remains the same. For comparison, the right panel of Fig. 5 shows reductions in tobacco retailer density (retailers/km (Bright Research Group, 2016)) after the 50% cap, school-to-retailer buffers and retailer-to-retailer buffers. In all cities, density reductions after a 600 m R2R buffer are greater than both the 50% license cap and 600 m S2R buffers. However, as is clear from single-city comparisons, policy impact on residents’ proximity to retailers and retailer density differs between cities. For example, in Miami, none of the three policies results in substantial changes in the percentage of residents living within 1 km of a retailer, though retailers per square kilometer decrease by over 50% under the 2000 ft school-to-retailer buffer, and by over 75% under the retailer-to-retailer buffer.

To provide a more grounded example of these findings, Fig. 6 illustrates the differential impact of policies and the importance of local context – including city size, baseline tobacco proximity (both measures), and baseline tobacco retailer density. Baltimore, Maryland, (Fig. 6, top) and Portland, Oregon (Fig. 6, bottom), are mid-sized (210 km (Bright Research Group, 2016) and 376 km (Bright Research Group, 2016)). Baltimore has a baseline retailer density of 5.80/km2 and a baseline median resident-to-retailer proximity of 0.24 km, while Portland has a baseline retailer density of 1.66/km2 and an average median baseline resident-to-retailer proximity of 0.39 km. The percentages of residents within 1 km of a retailer at baseline are 96% (Baltimore) and 84% (Portland). In the maps, the heatmap is red where larger proportions of residents reside within 1 km of a retailer and yellow to green in areas where this proportion is smaller. From left to right for Baltimore, we see that the yellow areas increase the most under the school-to-retailer buffer policy; however, we see little change in the heatmaps for Portland under the same policy conditions. In both cities, the 50% licensing cap reduces density by about half; the median resident-to-retailer distance increases just 0.05 km in Baltimore, but more than double that (0.12 km) in Portland. Both cities see only small decreases under this policy in the percentage of residents living within 1 km of a retailer: from 96% to 94% (Baltimore) and 84%–80% (Portland).

Fig. 6.

Fig. 6.

Open in a new tab

Maps of two cities under three retailer reduction policies.

Notes: LC 50% = License Cap at 50% of baseline; R2R 600m = 600 m Retailer-to-Retailer Buffer; S2R 600m = 600 m School-to-Retailer Buffer. Red dots represent tobacco retailers and black dots represent a proportion of the city’s population (each dot in Baltimore, Maryland = 62 people; each in Portland, Oregon = 64 people). The heatmap overlay becomes darker red to indicate high proportions of residents in close proximity to retailers, i.e., Tobacco Swamps, and lighter yellow and green when proportions are lower.

In both cities, the 600 m R2R buffer reduces density the most of the three policies, though median resident-to-retailer distance in Portland under this policy increases less (0.39–0.55 km) than in Baltimore where it doubles (0.24–0.51 km). The impact on retailer density of the 600 m S2R is not as strong as that of the 600 m R2R in San Diego (change under S2R = 1.21/km2 v R2R = 0.86/km2) and density under the 600 m S2R buffer in Baltimore is double that of the 600 m R2R buffer (change under S2R = 1.1/km2 v R2R = 0.56). In Baltimore, the S2R buffer is most effective at clearing the swamp in some neighborhoods (turning areas from red to yellow in the map). The median resident-to-retailer distance triples (0.24–0.74 km) and the percentage of residents within 1 km of a retailer drops by about one-third (96%–68%). The impact in Portland is not as large as in Baltimore under the 600 m S2R buffer, however this policy produces the largest impact on proximity of the three policies in Portland. Median resident-to-retailer proximity increases from 0.39 km to 0.63 km, and the percentage of residents within 1 km of a retailer decreases from 84% to 76%. The impact of each of the policies differs in terms of changes in proximity and density within one city, and looking across cities, policy impacts depend on baseline tobacco retailer concentration.

4. Discussion

In this study, we described the existing tobacco retail environment in 30 major US cities and simulated the effect of retail reduction policies on residents’ proximity to tobacco retailers and retailer density. Although we found a strong relationship between tobacco retailer density and resident-to-retailer proximity (R2 =0.86), we also saw that the impact of the same policy on density and proximity are not equal within cities. Furthermore, identical policies implemented in different cities had diverse impacts on both retailer density and resident-to-retailer proximity.

4.1. The importance of context

In our simulations, some interventions had little effect on some measures of retailer concentration, though results differed across cities. This points to the importance of context. In this study, we investigated context as the built environment for tobacco at baseline in each of the 30 cities. Our results showed relatively different outcomes for one group of cities, those that currently have high tobacco retailer concentration, in terms of both density and proximity, and another group of cities with lower retailer concentration. For cities with high retailer concentration, we found that only the school-to-retailer buffer policy had large effects on proximity measures. Conversely, for the same policy in low-retailer-concentration cities, we saw greater effects on proximity as compared to the first group. Notably, all three policies reduced retailer density across all 30 cities. Our results also suggest that when policies in high-concentration cities reduce density by more than some threshold – around 50% – proximity, measured as the percentage of residents living within 500 m or 1 km of a tobacco retailer, also drops. However, this is not the case in low-concentration cities where the effects are much less apparent, especially in the case of retailer-to-retailer (R2R) buffers. When density is markedly reduced, proximity metrics barely change. Taken together, these findings indicate that other mechanisms are at work in the relationship between retailer density and resident-to-retailer proximity. Such mechanisms might include other contextual factors beyond the baseline built environment for tobacco, like tobacco industry targeting of racially marginalized or economically disadvantaged populations or historical reasons for areas with higher concentrations of tobacco retailers, e.g., redlining (Glasser and Roberts, 2021).

4.2. Implications for science and practice

These results have implications for tobacco control, both for future research and policy considerations for urban US environments. Systematic reviews investigating literature on associations between tobacco use behaviors and retailer concentration conclude that in areas with relatively lower density, or those with more distance between residents and retailers, tobacco use prevalence is lower than in areas of higher retailer concentration (Lee et al., 2021; Glasser and Roberts, 2021). Most studies are quasi-experimental and cross-sectional, and do not compare areas before and after a retail reduction policy is introduced. Simulation approaches like in this study and others are useful since most cities have not yet implemented these types of policies, and the few that have successfully implemented these policies have done so relatively recently (Glasser and Roberts, 2021; Caryl et al., 2020). Furthermore, in these cities, retailer density and proximity are strongly correlated at baseline. This might suggest that when the retailer market is relatively unregulated, as is the case in most of the US, density and proximity are correlated in most or all communities. Whether distance or clustering might be a more important tobacco control mechanism is difficult to disentangle when the measures are so highly correlated, as in the 30 cities studied here.

In addition, simulation studies find that certain policies in certain communities affect proximity or density in different ways (Glasser and Roberts, 2021; Caryl et al., 2020; Ribisl et al., 2017; Luke et al., 2017). However, less is known about the mechanisms at work over time and any causal relationships between or thresholds around the necessary amount of change needed in the built environment for tobacco and tobacco use behaviors. More attention to these mechanisms, and considerations of both density and proximity in each study will help to unpack how a changing built environment may change behavior.

Our study also highlights the importance of doing more research on actual purchasing behaviors and the built environment over time. While established surveillance systems, e.g., BRFSS (Behavioral Risk Factor Surveillance System, 2016) and PATH (PATH (Population Assessment of Tobacco and Health) Study - Home, 2021) include items about price paid and discounts, they do not ask about the distance or frequency of which consumers travel to purchase cigarettes. The dearth of information around travel and other purchasing behaviors prevented us from analyzing how changes in resident-to-retailer proximity might directly impact tobacco use behaviors immediately and further downstream.

Our work also has implications for policy development in urban environments. It draws attention to the value of examining the actual distribution of retailers in a community prior to policy development and implementation, as well as understanding the descriptive metrics for resident-to-retailer proximity and retailer density. Policymakers and other stakeholders should keep in mind that retail reduction policies will likely affect proximity and density differently when planning for and gauging successful policy outcomes. In addition, different measures of proximity, such as average resident-to-retailer distance, or the proximity of retailers to schools, may resonate differently across communities and groups.

In our simulations, changes to tobacco retailer density and resident-to-retailer proximity are instantaneous, but in reality, change is slower. For example, Philadelphia’s tobacco retailer density cap at one retailer per 1000 daytime population resulted in a 20% decrease in the total number of retailers in three years (2016–2019) (Lawman et al., 2020). It is difficult for policy makers to estimate how long it would take to achieve simulated outcomes, because rates of “natural attrition” for tobacco retailers are not well established, though one recent study found a rate of 7% natural attrition for tobacco retailers in the US (Golden et al., 2021). Retailer closures are likely affected by multiple factors, including the type of licensing scheme and fee structure (Burton et al., 2021; Bowden et al., 2014), strength of tobacco control (e.g., bans on flavors or other product characteristics, and the presence of litter mitigation fees for cigarette butts and e-cigarette waste). Economic stressors, including the Covid-19 pandemic, are also relevant because they may contribute to tobacco retailer closures (Barker et al., 2021). There are, however, strategies that governments could employ to accelerate tobacco retail reduction, such as license revocation for repeated violations of a minimum legal sales age or a tobacco-free generation. For example, Philadelphia’s policy eliminated exemptions for existing retailers that violated the minimum legal sales age law three times within 24 months, which rendered 149 retailers ineligible for license renewal and may have contributed to greater reduction in lower-income neighborhoods (Lawman et al., 2020; Feldman, 2020). Conversely, the US Food and Drug Administration has the authority to issue no-tobacco-sale orders to stores that repeatedly violate underage sales laws, but evidence suggests that this regulatory mechanism is underutilized (Hemmerich et al., 2015). Future simulations could incorporate retail category and neighborhood characteristics that are associated with higher odds of repeat violations.

4.3. Strengths and limitations

A major strength of this study is that it is grounded in real-world data from 30 major US cities. Although many studies compare places by urbanicity (urban v. suburban v. rural), the current study demonstrated within-category variation through the diverse results across 30 specific urban areas. Another strength is the generation of synthetic populations for each city that preserve demography at the US Census-designated Place and Census Tract levels (Gallagher et al., 2018). The process of generating the populations also assigns realistic geographic coordinates for residences and workplaces for each observation. We also compared the policies’ impact on resident-to-retailer proximity to that of retailer density. While some quasi-experimental studies consider both density and proximity (Cantrell et al., 2016; Shareck et al., 2016, 2020), most simulation studies of the tobacco retail environment focus on one or the other (Lee et al., 2021).

Obvious limitations of this simulation are relying on random selection to eliminate tobacco retailers from retailer-to-retailer (R2R) buffers, and assuming that policies result in immediate reduction of tobacco retailers, which simplifies the complicated realities of policy implementation in the real world. In San Francisco, an 8% reduction achieved within 10 months suggests that it could take years to achieve the 70–80% reduction that these simulations suggest are necessary to dramatically reduce the proportion of urban dwellers who live in close proximity to a tobacco retailer. Strong and explicit penalty and license revocation structures, like those in Philadelphia’s retail reduction policies, can accelerate the goal of maximizing the distance between residents and tobacco retailers. In addition, observed differences between the 30 cities in this study suggest that results from one city may not generalize to others. However, the patterns of similar policy impact in big cities with high retailer concentration and the same for those with low retailer concentration suggest that similar cities outside the 30 could compare their retail environments to those studied here and gain understanding of potential policy impact in their own community. For instance, policymakers and intermediaries could find the city of the 30 here that is closest to theirs in size, population, and tobacco retailer concentration to identify the potential and comparative impact a policy might have in their city. We do, however, acknowledge that such policies likely have different impacts in suburban and rural communities.

A limitation of the current study and direction for future research is to investigate the equity impact of tobacco retail reduction strategies (Cantrell et al., 2015; Kong et al., 2021). Indeed, we did not investigate how contextual factors (e.g., socioeconomics and demographics) might influence policy outcomes with regard to proximity and density. Research is needed to identify the strategies that eliminate (rather than exacerbate) place-based disparities in tobacco retail density and proximity.

Finally, we note that the use of Euclidean rather than Manhattan distances is a limitation. Although tobacco users likely experience distance based on city blocks (or roadways) or travel patterns, rather than Euclidean distances, many cities like Mounds View, Minnesota, have implemented retailer reduction policies based on Euclidean distances (§ 118, 2022). Fewer cities have implemented retailer proximity policies using Manhattan distance, such as Minneapolis, Minnesota, which measures distances between retailers door-to-door (City of Minneapolis Community Planning and Economic Development, 2019). With respect to residents’ proximity to retailers, we performed sensitivity analyses for proximity statistics at 0.5, 1, and 2 km, available in the appendix. However, we realize that straight-line (Euclidean) distances are likely shorter than road network distances and may, therefore, overestimate policy impact on city residents’ proximity to retailers.

4.4. Directions for future research

Findings from this study suggest many avenues for future research on the tobacco retail environment. Our findings around the importance of context regarding the built environment at baseline aligns well with the current body of literature around racial, ethnic, and economic disparities in tobacco retailer density and proximity to residents. This knowledge obliges us and other researchers to investigate how other contextual factors (such as racial, ethnic, and income composition of communities) are associated with policy impact, and to increase the study of actual and probable equity-oriented impact of policies and potential unwanted consequences (Caryl et al., 2020; Craigmile et al., 2020).

In addition, repeating these simulations and analyses with other retail reduction policies could add to the evidence base. Much of our data for retailers came from state or local licensing lists, which do not include information about store type. This prevented us from studying policies based on store type, e.g., pharmacy sales prohibitions or restricting all tobacco sales to adult-only tobacco specialty stores. Obtaining data on likely tobacco retailers, through a protocol cited above (Golden et al., 2021), would allow for exploring policy effects across different types of retailers.

Similarly, simulations like ours are beginning to illuminate possible mechanisms at work around resident-to-retailer proximity and retailer density, along with the differences between them. This suggests that some creativity for strengthening existing policies could be effective. For example, why limit buffer policies to schools and not all youth-friendly locales, e.g., parks, libraries, churches, public swimming pools, etc.? Communities that have focused on walkability and green spaces may be especially open to broadening these types of policies.

Finally, since most retail reduction policies are implemented with exemptions to existing retailers and depend on natural attrition, more evaluation of policies currently or imminently in place are needed to learn more about actual rates of decline. This would subsequently allow for simulation modeling in more realistic timeframes and could facilitate comparisons between policies with and without the revocation penalties discussed above. Many retail reduction policies are new, and a few studies have been completed (Bright Research Group, 2016; Lawman et al., 2020; Vyas et al., 2020), yet more are needed.

Given the vast saturation of tobacco retailers and products in many communities, we should continue studying tobacco retail reduction policies so that policymakers can have the evidence they need to support ‘clearing the tobacco swamp’ and improving the public health environments of their communities.

Supplementary Material

Appendix

Acknowledgements

The National Cancer Institute of the National Institutes of Health supported this research under Award Number P01CA225597 for ASPiRE (Advancing Science & Practice in the Retail Environment). We are grateful to Nina C. Schleicher, Monika Vishwakarma (Stanford Prevention Research Center) and Santosh E. Gummidipundi (Quantitative Sciences Unit) at Stanford University School of Medicine for curating data and identifying tobacco retailers in the 30 ASPiRE cities.

Footnotes

Declaration of competing interest

None.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.healthplace.2022.102815.

1

New York City (NYC), Los Angeles (LAX), Chicago (CHI), Houston (HOU), Phoenix (PHX), Philadelphia (PHL), San Antonio (SAT), San Diego (SD), Dallas (DAL), San Francisco (SFO), Fort Worth (FOW), Charlotte (CLT), Seattle (SEA), Detroit (DET), Denver (DEN), Washington (DC), Boston (BOS), Memphis (MEM), Portland (PDX), Las Vegas (LAS), Baltimore (BAL), Sacramento (SAC), Kansas City (KC), Atlanta (ATL), Miami (MIA), Oakland (OAK), Minneapolis (MIN), Cleveland (CLE), New Orleans (NO), and Providence (PRV).

References

  1. American Lung Association, 2021. Tobacco prevention and cessation funding | state of tobacco control. https://www.lung.org/research/sotc/state-grades/state-rankings/tobacco-prevention-funding. (Accessed 26 August 2021).
  2. ArcGIS. Public Schools. ArcGIS Hub, 2022. https://hub.arcgis.com/datasets/geoplatform::public-schools/about. (Accessed 6 April 2022). [Google Scholar]
  3. ASPiRE Center. Tobacco retail policy trends in 2019. ASPiRE Center. http://aspirecenter.org/resource/tobacco-retail-policy-trends-2019/. (Accessed 17 April 2021). [Google Scholar]
  4. Axle, Data. Data Axle reference solutions. Data Axle. https://www.data-axle.com/what-we-do/reference-solutions/. (Accessed 22 November 2021). [Google Scholar]
  5. Barker DC, Henriksen L, Voelker DH, et al. , 2021. Turning over a new leaf: vape shop closings, openings and transitions in six U.S. Metropolitan statistical areas. Prev. Med. Rep 23, 101428. 10.1016/j.pmedr.2021.101428. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Barthelemy J, Toint PL, 2013. Synthetic population generation without a sample. Transp Sci. 47 (2), 266–279. [Google Scholar]
  7. Behavioral Risk factor surveillance system: 2016. BRFSS Survey data and documentation. Centers for disease control and prevention (CDC). Published December 6, 2017. https://www.cdc.gov/brfss/annual_data/annual_2016.html. (Accessed 19 January 2018). [Google Scholar]
  8. Big Cities Health Coalition. Big Cities Health Coalition Members, 2017. https://www.bigcitieshealth.org/members/. (Accessed 23 May 2017).
  9. Bivand RS, Pebesma E, Gomez-Rubio V, 2013. Applied Spatial Data Analysis with R, ´ second ed. Springer-Verlag. 10.1007/978-1-4614-7618-4. [DOI] [Google Scholar]
  10. Boisjoly G, Wasfi R, El-Geneidy A, 2018. How much is enough? Assessing the influence of neighborhood walkability on undertaking 10-minute walks. J. Transp. Land Use 11 (1), 143–151. [Google Scholar]
  11. Bowden JA, Dono J, John DL, Miller CL, 2014. What happens when the price of a tobacco retailer licence increases? Tobac. Control 23 (2), 178–180. 10.1136/tobaccocontrol-2012-050615. [DOI] [PubMed] [Google Scholar]
  12. Bright Research Group, 2016. Reducing Tobacco Retail Density in San Francisco: A Case Study, 16. https://sanfranciscotobaccofreeproject.org/wp-content/uploads/Retail-Density-Case-Study-1.27.16-FINAL-to-TFP.pdf. [Google Scholar]
  13. Burton S, Phillips F, Watts C, et al. , 2021. Who sells tobacco, who stops? A comparison across different tobacco retailing schemes. Tobac. Control 30 (4), 392–398. 10.1136/tobaccocontrol-2019-055561. [DOI] [PubMed] [Google Scholar]
  14. Cantrell J, Anesetti-Rothermel A, Pearson JL, Xiao H, Vallone D, Kirchner TR, 2015. The impact of the tobacco retail outlet environment on adult cessation and differences by neighborhood poverty. Addiction 110 (1), 152–161. 10.1111/add.12718. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Cantrell J, Pearson JL, Anesetti-Rothermel A, Xiao H, Kirchner TR, Vallone D, 2016. Tobacco retail outlet density and young adult tobacco initiation. Nicotine Tob. Res 18 (2), 130–137. 10.1093/ntr/ntv036. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Caryl FM, Pearce J, Reid G, Mitchell R, Shortt NK. Simulating the density reduction and equity impact of potential tobacco retail control policies. Tobac. Control Published online November 2, 2020. doi: 10.1136/tobaccocontrol-2020-056002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Chang W, Cheng J, Allaire JJ, et al. , 2021. Shiny: web application framework for R. https://CRAN.R-project.org/package=shiny. (Accessed 17 April 2021).
  18. City of Minneapolis Community Planning and Economic Development. Tobacco density study June 2019. https://lims.minneapolismn.gov/. (Accessed 13 March 2022).
  19. City of New York. Tobacco retail dealer license - NYC business. NYC business. https://www1.nyc.gov/nycbusiness/description/cigarette-retail-dealer-license. (Accessed 11 May 2021). [Google Scholar]
  20. Consumer, Access. Access consumer spend study. https://cdn2.hubspot.net/hubfs/263750/Access_Consumer_Spend_Study_2016.pdf. (Accessed 17 April 2021).
  21. Craigmile PF, Onnen N, Schwartz E, Glasser A, Roberts ME. Evaluating how licensing-law strategies will impact disparities in tobacco retailer density: a simulation in Ohio. Tobac. Control Published online August 21, 2020. doi: 10.1136/tobaccocontrol-2020-055622. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. D’Angelo H, Fleischhacker S, Rose SW, Ribisl KM, 2014. Field validation of secondary data sources for enumerating retail tobacco outlets in a state without tobacco outlet licensing. Health Place 28, 38–44. 10.1016/j.healthplace.2014.03.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Dun & Bradstreet - accelerate growth and improve business performance. https://www.dnb.com/. (Accessed 22 November 2021).
  24. Feldman N. Philly health department busts 149 tobacco retailers for selling to kids. WHYY. Published January 8, 2020. https://whyy.org/articles/philly-health-department-busts-149-tobacco-retailers-for-selling-to-kids/. (Accessed 8 October 2021). [Google Scholar]
  25. Gallagher S, Richardson LF, Ventura SL, Eddy WF, 2018. SPEW: synthetic populations and ecosystems of the world. J. Comput. Graph Stat 27 (4), 773–784. 10.1080/10618600.2018.1442342. [DOI] [Google Scholar]
  26. Glasser AM, Roberts ME, 2021. Retailer density reduction approaches to tobacco control: a review. Health Place 67, 102342. 10.1016/j.healthplace.2020.102342. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Golden SD, Baggett CD, Kuo TM, et al. , 2021. Trends in the Number and Type of Tobacco Product Retailers. 10.1093/ntr/ntab150. United States, 2000–2017. Nicotine Tob Res, (ntab150). [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Hemmerich N, Jenson D, Bowrey BL, Lee JGL. Underutilisation of no-tobacco-sale orders against retailers that repeatedly sell to minors, 2015–2019, USA. Tobac. Control Published online June 8, 2021:tobaccocontrol-2020–056379. doi: 10.1136/tobaccocontrol-2020-056379. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Hou Y, Moogoor A, Dieterich A, Song S, Yuen B, 2021. Exploring built environment correlates of older adults’ walking travel from lifelogging images. Transp. Res. Part Transp Environ 96, 102850. 10.1016/j.trd.2021.102850. [DOI] [Google Scholar]
  30. Kirst M, Chaiton M, O’Campo P, 2019. Tobacco outlet density, neighbourhood stressors and smoking prevalence in Toronto, Canada. Health Place 58, 102171. 10.1016/j.healthplace.2019.102171. [DOI] [PubMed] [Google Scholar]
  31. Kong AY, King BA. Boosting the Tobacco Control Vaccine: recognizing the role of the retail environment in addressing tobacco use and disparities. Tobac. Control Published online September 23, 2020:tobaccocontrol-2020–055722. doi: 10.1136/tobaccocontrol-2020-055722. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Kong AY, Delamater PL, Gottfredson NC, Ribisl KM, Baggett CD, Golden SD, 2021. Sociodemographic inequities in tobacco retailer density: do neighboring places matter? Health Place 71, 102653. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Lawman HG, Henry KA, Scheeres A, Hillengas A, Coffman R, Strasser AA, 2020. Tobacco retail licensing and density 3 Years after license regulations in Philadelphia, Pennsylvania (2012–2019). Am. J. Publ. Health 110 (4), 547–553. 10.2105/AJPH.2019.305512. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Lee JG, Kong AY, Sewell KB, et al. Associations of tobacco retailer density and proximity with adult tobacco use behaviours and health outcomes: a meta-analysis. Tobac. Control Published online September 3, 2021. doi: 10.1136/tobaccocontrol-2021-056717. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Levy DT, Huang AT, Havumaki JS, Meza R, 2016. The role of public policies in reducing smoking prevalence: results from the Michigan SimSmoke tobacco policy simulation model. Cancer Causes Control 27 (5), 615–625. 10.1007/s10552-016-0735-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Luke DA, Hammond RA, Combs T, et al. , 2017. Tobacco town: computational modeling of policy options to reduce tobacco retailer density. Am. J. Publ. Health 107 (5), 740–746. 10.2105/AJPH.2017.303685. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Marsh L, Vaneckova P, Robertson L, et al. , 2021. Association between density and proximity of tobacco retail outlets with smoking: a systematic review of youth studies. Health Place 67, 102275. 10.1016/j.healthplace.2019.102275. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Myers AE, Hall MG, Isgett LF, Ribisl KM, 2015. A comparison of three policy approaches for tobacco retailer reduction. Prev. Med 74, 67–73. 10.1016/j.ypmed.2015.01.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. PATH (Population Assessment of Tobacco and Health) Study - Home. https://pathstudyinfo.nih.gov/. (Accessed 8 October 2021).
  40. Pearce J, Rind E, Shortt N, Tisch C, Mitchell R, 2016. Tobacco retail environments and social inequalities in individual-level smoking and cessation among scottish adults. Nicotine Tob. Res 18 (2), 138–146. 10.1093/ntr/ntv089. [DOI] [PubMed] [Google Scholar]
  41. Pebesma E, 2018. Simple features for R: standardized support for spatial vector data. R J 10 (1), 439–446. [Google Scholar]
  42. Public Health, Tobacco Policy Center, 2020. Tobacco retail licensing: promoting health through local sales regulations. http://www.tobaccopolicycenter.org/documents/TobaccoRetailLicensing.pdf. (Accessed 14 May 2021).
  43. R Core Team, 2021. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. [Google Scholar]
  44. R Core Team. R, 2019. A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/. (Accessed 15 February 2020). [Google Scholar]
  45. Ribisl KM, Luke DA, Bohannon DL, Sorg AA, Moreland-Russell S, 2017. Reducing disparities in tobacco retailer density by banning tobacco product sales near schools. Nicotine Tob Res Off J Soc Res Nicotine Tob 19 (2), 239–244. 10.1093/ntr/ntw185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Santa Clara County Public Health. Tobacco retailer laws: county of Santa Clara (unincorporated areas). Published online June 30, 2020. https://cpd.sccgov.org/sites/g/files/exjcpb706/files/Unincorporated%20Santa%20Clara%20County%20Tobacco%20Retail%20Program%20Fact%20Sheet.pdf. (Accessed 14 May 2021).
  47. Shareck M, Kestens Y, Vallee J, Datta G, Frohlich KL, 2016. The added value of ´ accounting for activity space when examining the association between tobacco retailer availability and smoking among young adults. Tobac. Control 25 (4), 406–412. 10.1136/tobaccocontrol-2014-052194. [DOI] [PubMed] [Google Scholar]
  48. Shareck M, Datta GD, Vallée J, Kestens Y, Frohlich KL, 2020. Is smoking cessation in young adults associated with tobacco retailer availability in their activity space? Nicotine Tob. Res 22 (4), 512–521. 10.1093/ntr/nty242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Shortt NK, Tisch C, Pearce J, Richardson EA, Mitchell R, 2016. The density of tobacco retailers in home and school environments and relationship with adolescent smoking behaviours in Scotland. Tobac. Control 25 (1), 75–82. 10.1136/tobaccocontrol-2013-051473. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Stanford, University. SRITA – Stanford research into the impact of tobacco advertising. https://tobacco.stanford.edu/. (Accessed 16 November 2021).
  51. Tobacco Control Legal Consortium, 2016. Chicago’s Regulation of Menthol Flavored Tobacco Products: A Case Study. Public Health Law Center at Mitchell Hamline School of Law. https://publichealthlawcenter.org/sites/default/files/resources/tclc-casestudy-chicago-menthol-2015_0.pdf. (Accessed 12 May 2021). [Google Scholar]
  52. US Census Bureau. American community Survey (ACS). The United States Census Bureau. https://www.census.gov/programs-surveys/acs. (Accessed 17 April 2021). [Google Scholar]
  53. US Department of Transportation. National household travel Survey. National household travel Survey. https://nhts.ornl.gov/. (Accessed 17 April 2021). [Google Scholar]
  54. Vyas P, Sturrock H, Ling PM, 2020. Examining the role of a retail density ordinance in reducing concentration of tobacco retailers. Spat Spatio-Temporal Epidemiol. 32, 100307. 10.1016/j.sste.2019.100307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Wälty S, 2018. Based on the Analysis of Ten Essential Elements: Does Greater Zurich Provide Healthy, 10-Minute Neighborhoods? ETH Zurich. 10.3929/ethz-b-000304251. [DOI] [Google Scholar]
  56. § 118.031 Proximity to other tobacco retailers. American Legal Publishing Corporation. Accessed March 31, 2022. https://codelibrary.amlegal.com/codes/moundsviewmn/latest/moundsview_mn/0-0-0-4816.

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Appendix
NIHMS1905604-supplement-Appendix.docx (197KB, docx)

RESOURCES