Background
Researchers have established that disadvantaged neighborhoods with a high concentration of residents with low socio-economic status (SES) often have relatively low average life expectancy [
1,
2], poor mental and physical health [
3‐
8], and low school completion rates [
9]. The reasons for this are not entirely clear. Using a life course approach, Seabrook and Avison [
3] highlight that individuals with differing SES have differing stressors, and differing resources to address those stressors, and that individuals often have “linked lives” with individuals of similar SES. Therefore, if an individual with, say, low SES, has friends, neighbors, and family members who also have low SES, that individual will likely be exposed to the stresses and challenges that her or his family, neighbors, and family members experience (e.g., job loss, food insecurity, lack of health care services, etc.). In addition to stress, the built environment in disadvantaged neighborhoods is often quite different from that of more advantaged neighborhoods. Specifically, disadvantaged neighborhoods are less likely to have infrastructure in place that promotes health, such as food stores with healthy options [
10], quality health care services [
11], and safe, walkable spaces [
12,
13].
One measure of neighborhood disadvantage, the area deprivation index (ADI), was originally developed by Gopal Singh [
14], and revised by Amy Kind and colleagues at the University of Wisconsin-Madison [
15]. The revised ADI is a validated instrument that has been used by researchers and government agencies such as the Centers for Medicare and Medicaid Services [
16]. Recently, a number of studies have been published using the ADI [
17‐
22].
Three unhealthy consumption behaviors are tobacco use [
23], alcohol use [
24], and eating unhealthy snack foods [
25]. All three behaviors are major risk factors for chronic disease and premature death [
26]. Given that people with low SES in the US are more prone to tobacco use [
27,
28], alcohol use [
27,
28], and unhealthy snack food intake [
25] than people with higher SES, and that all three risky health behaviors have been identified as mechanisms to cope with stress [
29], we would expect the sales of these items to be high in areas with higher neighborhood disadvantage.
Neighborhood deprivation has been associated with risky health behaviors including excessive alcohol consumption, physical inactivity, and high-fat diets [
29]; studies have examined the relationship between neighborhood disadvantage and the availability of unhealthy products. A study by Lee et al. [
30] found that as the proportion of African American residents increased and median household income decreased, the density of tobacco outlets per 1000 in census tracts increased. Similarly, Datta et al. found that higher levels of neighborhood poverty were associated with higher prevalence of smoking [
31]. Other researchers found that residents in areas with more tobacco outlets were more likely to start smoking than residents in other areas, and less likely to quit [
32]. There is also evidence that greater availability [
33,
34] and acceptability of alcohol (or “neighborhood norms”) are linked to higher alcohol use [
35,
36]. Research has also found that living in a disadvantaged neighborhood is associated with less access to high-quality food sources [
25], greater access to unhealthy snack foods [
37], and greater exposure to unhealthy snack food advertising [
38]. Separate from those neighborhood risks, there is evidence that low household SES is associated with greater access to unhealthy products [
25,
30,
32‐
39].
Despite the studies described above, there is still a gap in research that examines whether the greater access to unhealthy products actually results in greater purchasing of these products. Also, while living in a neighborhood with low SES is associated with a number of poor health outcomes, there is limited research identifying the specific chain of events that leads to poor health outcomes. In this study, we address a gap in this area of research by examining the purchase of unhealthy products and the role of neighborhood disadvantage. Specifically, the purpose of this study is to assess how neighborhood disadvantage is associated with the sales of unhealthy products (alcohol, tobacco, and unhealthy snacks) at a chain of discount variety stores. Discount variety stores (DVSs) sell a wide range of products. The items we included in our analysis are described in the “Outcome Measures” section. The specific stores in our study were selected by the DVS chain that made its data available for our study. The stores were located in different neighborhoods.
There have been few studies of the sales of unhealthy products in deprived neighborhoods using actual sales data. There is little research in this area likely because sales data is proprietary, and it is rarely in the best interest of the business to share these data with researchers. Because we had the opportunity to use proprietary sales data, we were able to address this gap in the research by examining actual sales data. From this study, the association between neighborhood deprivation and actual purchasing behavior in places where there is typically limited choice adds new sales-based evidence that sales of tobacco and unhealthy snacks are greater in deprived neighborhoods. We hypothesized that there would be a positive association of neighborhood disadvantage with the sales of unhealthy products.
Results
The unit sales of unhealthy products varied among stores, where the median weekly unit sales for tobacco were 412 [interquartile range (IQR) 205, 610], alcohol 339 [IQR 147, 428], and unhealthy snack foods 1378 [IQR 883, 2526] (Table
1). Results for the socio-demographic variables indicate that, in general, the DVSs were located in neighborhoods where the median ADI percentile was 87 [IQR 83, 89] (compared to median ADI percentile of 50 for all US communities). The racial composition of the communities was primarily Non-Hispanic African American, where the median percentage of African-Americans was 74% [IQR 61, 87]; that result differed greatly from the national average, 13% [
50]. The median percent of the population comprised of children (ages 0–17) was 25% [IQR 23, 29] (compared to 23% of the US population in 2017 [
51]). The median walk score was 48 [IQR 39, 59], indicating that the majority of communities were car dependent [
40].
Table 1
Neighborhood Socio-Demographic Characteristics of 16 Discount Variety Storesa
ADI national rank percentilec | 87 [83,89] |
Household income (in dollars) | 25,827 [20,387, 35,320] |
Population Size | 1068 [901,1889] |
Percent of population that is Non-Hispanic African-American | 74 [61,87] |
Percent population ages 0–17 | 25 [23,29] |
Walk Scored | 48 [39,59] |
Multivariate analyses indicated that as the ADI percentile increased, the sales of tobacco and unhealthy snacks in DVSs increased significantly. Specifically, for every 1% increase in ADI, the weekly unit sales of tobacco per store increased by 11.48 (95% Confidence Interval, 95% CI 5.02; 17.94) adjusted for population size and other covariates (Table
2). There was also a negative relationship between the racial composition of the neighborhood and tobacco sales: when the percent of Non-Hispanic African-Americans in a neighborhood increased by 1%, the weekly unit sales for tobacco decreased by 6.16 (95% CI -9.06; − 3.27). Finally, there was a negative relationship between neighborhood walk score and tobacco sales: when the walk score increased by 1%, the weekly unit sales for tobacco decreased by 6.18 (95% CI -10.36; − 2.01).
Table 2
Relationship between Area Deprivation Index and Neighborhooda Characteristics and the Number of Tobacco Units Sold on Average by Store per Week
ADI national percentileb | 11.48 | 0.02 | 4.37 | 18.52 |
Percent of population that is Non-Hispanic African-American | −6.16 | 0.01 | −9.35 | −2.98 |
Percent population ages 0–17 | 22.00 | 0.05 | 4.35 | 39.64 |
Walk scorec | −6.18 | 0.03 | −10.78 | −1.59 |
Conditional R-squared = 0.85 |
We did not find a relationship between alcohol sales and neighborhood characteristics (Table
3). We found that of the 16 DVSs, alcohol was sold at only 8 stores, which limited statistical power (degrees of freedom =2).
Table 3
Relationship between Area Deprivation Index and Neighborhood Characteristics and the Number of Alcohol Units Sold on Average by Store per Weeka
ADI national percentileb | −11.34 | 0.35 | −38.60 | 15.91 |
Percent of population that is Non-Hispanic African-American | −5.13 | 0.17 | −12.13 | 1.88 |
Percent population ages 0–17 | 21.00 | 0.26 | −18.87 | 60.79 |
Walk scorec | 0.49 | 0.91 | −11.01 | 11.98 |
Conditional R-squared = 0.87 |
For each 1% increase in ADI, the weekly unit sales of unhealthy snack food increased by 42.57 (95% CI 28.13; 57.01) adjusted for population size and other covariates (Table
4). In addition, each 1% increase in the percentage of children age 0–17 living in the census block was associated with increased weekly unit sales of unhealthy snacks of 62.73 (95% CI 26.88; 98.59).
Table 4
Relationship between Area Deprivation Index and Neighborhood Characteristics and the Number of Unhealthy Snack Food Units Sold on Average by Store per Weeka
ADI national percentileb | 42.57 | < 0.001 | 26.67 | 58.45 |
Percent of population that is African-American | −5.52 | 0.19 | −12.64 | 1.60 |
Percent population ages 0–17 | 62.73 | 0.02 | 23.28 | 102.18 |
Walk scorec | −7.35 | 0.22 | −17.61 | 2.92 |
Conditional R-squared = 0.64 |
Discussion
We used sales data from a small-format national DVS chain and neighborhood characteristic measures to examine the relationship between neighborhood disadvantage and sales of unhealthy products (tobacco, alcohol, and unhealthy snack foods). Tobacco sales were greater in neighborhoods with greater disadvantage. This may be due to more tobacco advertising in lower SES neighborhoods [
52,
53], higher levels of stress among residents [
54,
55], and/or greater access to tobacco products [
30,
32]. Also, as neighborhood deprivation increased, the sales of unhealthy snack foods increased [
44,
47]. Similar to tobacco sales this may be due to increased advertising for unhealthy snack foods found in lower SES neighborhoods [
38], increased levels of stress among residents [
39], and/or increased access to unhealthy snack products in these communities [
37]. There is also evidence that many people with lower income in the United States deliberately choose high fat foods to fulfill caloric needs, a factor that may have contributed to the results [
56].
We did not find a relationship between neighborhood characteristics and alcohol sales. However, this result may be due in part to limitations of our data: only 8 of the 16 stores sold alcohol, which limited our statistical power. There were generally no clear differences between the neighborhoods in which DVS locations chose to sell alcohol and those in DVS locations did not. The decision to sell alcohol in these DVSs may be dependent on factors such as store or local alcohol sale policies or local opposition to alcohol sales, although the DVSs in this study were located in states and municipalities that allowed alcohol sales in DVSs. Our study focused on stores located in low- to very low-SES neighborhoods. The lack of statistical significance for alcohol measures may also be due to the fact that we did not compare stores in high SES communities to stores in low SES communities; that feature of our analysis limited variation, and therefore also statistical power, although it offered the advantage of controlling for community SES. However, there may be granular levels of correlation among certain neighborhood characteristics (such as income level, demographic data, and level of neighborhood support or criticism of alcohol sales) and alcohol sales. In our future research, we would like to investigate whether or not that is the case. Future research to examine why stores like DVS choose not to sell alcohol when it is legal to do so would also be useful, particularly as that choice may be associated with local patterns of food consumption or tobacco use.
Neighborhoods with higher proportions of children had higher sales of unhealthy snacks. The majority of children (87%) and adults (87%) report snacking each day; other research has found that American adults eat over 500 cal per day while snacking [
44,
47]. It may be that the unhealthy snack foods assessed in this study appealed to children more than adults. For instance, from previous research we are aware that the majority of snack calories that children consume are from desserts, sweets, and salty foods [
47]. Research looking at the snacking calories consumed by adults includes alcohol [
44], which in the current study was captured in a separate outcome variable.
Also, there was a negative relationship between the percentage of the population that was non-Hispanic African American and tobacco sales [
57]. Previous studies found that tobacco outlet density increased as the proportion of African Americans increased [
30], and that residents in areas with more tobacco outlets were more likely to smoke [
32]. Other researchers found that in some racially integrated neighborhoods White residents were more likely to smoke than African American residents, and that these behaviors can be explained by the social environment [
45]. For example, LaVeist et al. [
45] found that when whites and African Americans lived in similar conditions, health disparities either decreased, or completely disappeared. The authors concluded that there may be few racial disparities when social factors are equalized [
45].
Limitations
Our results were based on sales at only 16 stores. However, researchers have rarely had access to detailed proprietary data such as the data we used, and our time series data were quite rich with weekly sales over 18 months. This study represented deprived areas, primarily in the Southeastern United States; the findings may not apply to less deprived areas or other regions, or to areas with more racial and socioeconomic diversity. However, the demographic characteristics of the neighborhoods we studied were similar to other low income neighborhoods throughout the Southeast. The data indicated whether products were purchased but did not confirm that the products were consumed. Also, individuals who make purchases at DVS may not live in the same census block group. It is quite possible that individuals who live and work in neighborhoods with a different ADI purchased the unhealthy products studied in this analysis.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.