Changes in Neighborhood Foodscapes From 2003 to 2023 by Area Rurality in the United States

Abstract

Objectives. To evaluate the change in the US foodscape from 2003 to 2023 overall and by area rurality, considering various modes of transportation.

Methods. We conducted this study using longitudinal data on all licensed food retailers, geocoded to census-tract level, and examined changes in the foodscape by estimating the percentages of tracts defined as food deserts (no grocery stores) and food swamps (mostly restaurants or fast-food locations).

Results. We found a considerable increase in the percentage of food swamps and no substantial progress in reducing food deserts. For example, using distance-based foodscape measure (one half of a mile from tract borders), between 2003 and 2023, the percentage of tracts categorized as food swamps increased from 80.2% to 88.5%, while the percentage of food deserts decreased minimally from 6.1% to 5.5%. Across all years, the percentage of food deserts was substantially higher in rural areas, and when people used public transit as a mode of transportation.

Conclusions. The findings underscore the need for improving accessibility to healthy foods across the United States, particularly in rural areas.

Public Health Implications. Expanding public–private partnerships to establish healthy food retailers in food deserts may improve the foodscape quality. (Am J Public Health. 2026;116(7):1004–1014. https://doi.org/10.2105/AJPH.2026.308495)

Unhealthy diet has been associated with adverse health outcomes, including excess body weight, diabetes, heart disease, and cancer.^1–5 Although dietary choices may vary by individuals’ preferences and food culture, the affordability and availability of foods are important factors contributing to those choices.^6–8 Area-level food environment, also referred to as retail foodscape, often acts as a proxy measure for geographic accessibility to healthy grocery food retailers. Therefore, studies on the retail foodscape often differentiate between food deserts—areas with no access to grocery retailers within a defined distance or travel time—and food swamps—areas with a substantially high number of restaurants and fast-food locations and only a limited number of grocery retailers.

Previous research generally indicates that dietary behaviors are associated with a neighborhood’s foodscape. In the United States, for example, the odds of consuming fast food near home increased by 11% to 61%, depending on the model adjustments, for every standard deviation increase in the fast-food density in the neighborhood.⁹ By contrast, in another study, fresh produce consumption increased with each additional supermarket in the neighborhood.¹⁰ Several studies have found an association between residing in food deserts or food swamps and adverse health outcomes, including higher body mass index,¹¹ cardiovascular health risk,⁴ obesity,¹² and diabetes-related hospitalization rates in the population.¹³

Given the rising burden of diet-related chronic diseases in the United States, it is essential to track changes in the neighborhood retail foodscape over time. So far, only 1 study has examined these changes to identify geographic and racial disparities in exposure to unfavorable food environment (i.e., neighborhoods with low concentrations of grocery stores) in the United States.¹⁴ However, no studies have examined changes in food deserts and food swamps over time across rural and urban areas and by persistent poverty status. Additionally, it remains unknown how different modes of transportation may influence geographic accessibility to healthy food stores, or if the COVID-19 pandemic coincided with any substantial changes in the neighborhood retail foodscape.

In this study, we examined changes in the US retail foodscape from 2003 to 2023, which includes the COVID-19 era. We also assessed trends in food deserts and food swamps by area rurality, persistent poverty status, and state, and compared accessibility by using distance-, automobile-, and public transit–based measures.

METHODS

We obtained annual data on all licensed businesses in the United States between 2003 and 2023 from DataAxle.¹⁵ Following earlier recommendations for business data refinement,¹⁶ we first selected all relevant business types using the North American Industry Classification System (NAICS) and Standard Industrial Classification (SIC) codes (Table 1) and business names. Next, records for each business were linked across all years based on full street address and geographic coordinates and geocoded to census tract level using the 2010 geometry. For each year, we calculated the number of healthy and unhealthy food retailers in each tract (Table 1). We then reconstructed the tract-level modified Retail Food Environment Index (mRFEI),¹⁷ which estimates the proportion of healthy food retailers from total food retailers in a given geographic area,^18,19 by using the following equation:

$$\mathrm{mRFEI}=100\times \frac{\mathrm{Healthy} \mathrm{Food} \mathrm{Retailers}}{\{\mathrm{Healthy} \mathrm{Food} \mathrm{Retailers}+\mathrm{Unhealthy} \mathrm{Food} \mathrm{Retailers}\}}$$ (1)

TABLE 1—

Changes in the Number of Food Retailers in the United States Between 2003 and 2023

Retail Category	Standard Industrial Classification Code	2003, No.	2020, No.	2023, No.	Change 2003–2020, %	Change 2020–2023, %	Change 2003–2023, %
All retailers combined		756 668	1 208 797	1 067 984	62.6	−11.6	41.1
Unhealthy food retailers
Total		623 738	1 035 370	919 199	66.0	−11.2	47.4
Bakery/coffee/tea shop	546101, 546105, 546107, 546108, 546109, 549915, 549938	26 157	42 468	37 796	62.4	−11.0	44.5
Convenience store (including gas station shops, dollar stores)	541103, 541104, 549999	98 473	137 258	120 119	39.4	−12.5	22.0
Dairy produce retailer (including ice-cream shops)	545101, 545102, 545103	3 874	10 214	8 241	163.7	−19.3	112.7
Full-service restaurant	581208	404 417	708 003	624 408	75.1	−11.8	54.4
Limited-service restaurant	581201, 581202, 581203, 581204, 581205, 581206, 581209, 581211, 581213, 581214, 581215, 581217, 581218, 581219, 581221, 581222, 581224, 581227, 581228, 581229, 581236, 581237, 581246, 581248, 581240, 581241, 581244, 581249, 581250, 581251, 581252, 581254, 581255	90 817	137 427	128 635	51.3	−6.4	41.6
Healthy food retailers
Total		132 930	173 280	148 785	30.4	−14.1	11.9
Department store/club store (carrying fresh produce)	531102, 531110	2 060	7 250	7 119	251.9	−1.8	245.6
Ethnic grocery store	541106, 541107, 541108, 549912, 549916, 549917, 549918, 549919, 549923, 549926, 549927, 549930, 549937, 549945	2 119	2 204	1 932	4.0	−12.3	−8.8
Fruit and vegetable market	543101, 543102, 543103, 543104, 543105, 543106	10 021	14 061	11 369	40.3	−19.1	13.5
Grocery store/supermarket	541101, 541105, 541109, 549928	104 044	131 213	112 729	26.1	−14.1	8.3
Meat/fish market	542101, 542104, 542105, 542106, 542107, 542108, 542109, 542110, 542111, 542113, 542114, 542116, 542118, 542117	14 686	18 552	15 636	26.3	−15.7	6.5

Note. Although full-service restaurants are categorized as “unhealthy,” several may offer healthy meals. This categorization was adopted from the initial modified Retail Food Environment Index (mRFEI)¹⁷ definition. A more granular categorization of full-service restaurants was not possible because information on food types served was not available.

In this study, we define accessibility as geographic access—the ease of reaching food retailers on the basis of proximity or travel time. We used 3 different methods: first, the traditional half-mile buffer zone around the tract boundary (distance-based mRFEI); second, 20-minute drive time zones from the population-weighted tract centroid by personal automobile (20-minute drive time corresponds to approximately 6–10 miles travel distance in urban areas and up to 20 miles travel distance in rural areas); third, 20-minute ride zones by public transit (including bus, local and regional rail, and ferries) using the TravelTime tool.^20,21 A 20-minute threshold was selected on the basis of empirical evidence indicating that after accounting for regional variation and urban–rural context, over 80% of the US population can access a supermarket within approximately 19.6 minutes (considered optimal access).²²

We then evaluated the US retail foodscape by estimating the percentages of tracts defined as food deserts (no healthy food retailers; mRFEI = 0), food swamps (mostly restaurants or fast-food locations; mRFEI < 33.3), and tracts with a good food environment (mRFEI ≥ 33.3), and the number of people living in food deserts and food swamps overall, by area rurality (based on the Urban Rural Indicator Codes, where rural census tracts are defined as those with an urban population of < 50%),²³ persistent poverty status (defined as a census tract with a poverty rate ≥ 20.0% in 4 consecutive measurement periods between 1990 and 2019),²⁴ and by state. Population data were derived from the American Community Survey’s 5-year estimates for 2015 to 2019 (for tract level) and 2019 to 2023 (for state level). We selected these mRFEI thresholds in accordance with an earlier study.²⁵ We also examined the association between changes in the percentage of food swamps and food deserts by calculating the swamp-to-desert ratio (overall, by area rurality, and by state), and we mapped changes in the number of healthy and unhealthy food retailers and in the proportion of food deserts and food swamps by state (including the District of Columbia) for the contiguous United States. We conducted data preprocessing using Python (Python Software Foundation, Wilmington, DE); data analysis and visualization were executed using R statistical software (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

The total number of food retailers in the United States steadily increased by 62.6%—from 756 668 in 2003 to 1 208 797 in 2020—but then decreased from 2020, coinciding with the start of the COVID-19 pandemic in the United States, by 11.6% to 1 067 984 in 2023 (Table 1). Similarly, the number of unhealthy and healthy food retailers rose from 2003 to 2020 by 66.0% (from 623 738 to 1 035 370) and 30.4% (from 132 930 to 173 280), respectively. Between 2020 and 2023, however, the number of both unhealthy and healthy food retailers decreased by 11.2% (to 919 199) and 14.1% (to 148 785), respectively. As a result, the overall increase in numbers from 2003 to 2023 was substantially greater for unhealthy food retailers (47.4%) than for healthy food retailers (11.9%).

Changes in the Foodscape Nationally

Using distance-based mRFEI (Figure 1), the percentage of food deserts first decreased by 1.9 percentage points between 2003 and 2020 (from 6.1% to 4.2%), but then increased to 5.5% in 2023. The percentage of food swamps increased between 2003 and 2020 by 9.6 percentage points (from 80.2% to 89.8%) but decreased to 88.5% in 2023. By contrast, the percentage of tracts with a good food environment decreased by 7.7 percentage points (from 13.7% to 6.0%) between 2003 and 2020 and remained at this level through 2023.

FIGURE 1—

Change in the US Foodscape From Overall, (b) Rural Areas, and (c) Urban Areas Using Distance-Based mRFEI Measure: 2003–2023

Note. mRFEI = modified Retail Food Environment Index. Dashed line indicates the beginning of the COVID-19 pandemic (2020).

Using the drive time–based mRFEI (Appendix Figure A, available as a supplement to the online version of this article at http://www.ajph.org), the percentage of food deserts was stable at 2.0% between 2003 and 2020 and increased minimally by 0.1 percentage point by 2023. The percentage of food swamps rose by 2 percentage points between 2003 and 2020 (from 94.8% to 96.8%) but then decreased by 0.1 percentage point by 2023. The percentage of tracts with a good food environment decreased by 2.0 percentage points (from 3.2% to 1.2%) between 2003 and 2020 and remained unchanged thereafter.

When using the public transit-based mRFEI (Appendix Figure A), the percentage of food deserts decreased by 2.0 percentage points, from 53.1% to 51.1%, between 2003 and 2020, but increased to 52.7% in 2023. The percentage of food swamps increased by 6.9 percentage points between 2003 and 2020, from 33.3% to 40.2%, but decreased to 38.7% in 2023. The percentage of tracts with a good food environment decreased by 5.0 percentage points, from 13.7% to 8.7%, during 2003 to 2020, and decreased further to 8.6% in 2023.

Stratified by rurality status (Figure 1, Appendix Figure A), the overall patterns were generally similar and reflected the nationwide trajectories. Regardless of the mode of transportation, in both rural and urban areas there was a minimal decrease or no change in the percentage of food deserts between 2003 and 2020, but an increase from 2020 to 2023. By contrast, the percentage of food swamps increased substantially between 2003 and 2020, then decreased slightly by 2023. The percentage of tracts with a good food environment either consistently decreased over time or remained consistently low, depending on the mode of transportation. However, in rural areas, the percentage of food deserts was substantially higher than in urban areas, whereas the percentage of food swamps was generally lower. In 2023, for example, the percentage of food swamps in rural and urban areas was 66.8% and 93.8% and the percentage of food deserts was 15.8% and 2.9%, respectively. The percentage of food deserts was lower in both rural (10.2%) and urban (0.2%) areas when we used automobile-based mRFEI, but it was substantially higher when we used public transit–based mRFEI (93.8% and 42.8%, respectively). Regardless of the mRFEI measure used, we found that a decrease in the percentage of food deserts was generally associated with an increase in the percentage of food swamps, as indicated by the swamp-to-desert ratio (Appendix Figure B). This relationship was more pronounced in urban than rural areas.

In 2023, the number of people residing in food deserts ranged from 4.9 million (1.5% of the total US population) using drive time–based mRFEI to 14 million (4.4%) using distance-based mRFEI to 178 million (55.3%) using public transit–based mRFEI (Appendix Table A). In rural areas, the total population residing in food deserts ranged from 4.5 million (8.1% of the total rural population) using drive time–based mRFEI to 7.1 million (2.2%) using distance-based mRFEI to 51.6 million (93.9%) using public transit–based mRFEI. Food swamp residents included over 290 million people, or more than 90% of the total US population, using distance-based or drive time–based mRFEI, and approximately 116 million (36%) using public transit–based mRFEI. (Appendix Table A).

Changes in the Foodscape by Poverty Status

Among 68 106 (94.4%) census tracts with available poverty data, 8244 tracts (11.4%) were categorized as persistent poverty areas (> 30 million residents). Regardless of the mRFEI measure used, the overall patterns in both persistent poverty tracts and nonpersistent poverty tracts reflected the nationwide trajectories (Appendix Figure C). However, selection of the mode of transportation substantially influenced the number of food deserts, especially in persistent poverty areas. For example, using distance-based or drive time–based mRFEI, 1.9% of tracts with persistent poverty were categorized as food deserts in 2003 and 2023. When we used public transit–based mRFEI, this percentage was 22.0% in 2003 and increased to 23.3% by 2023. The total population residing in food deserts and food swamps in persistent poverty areas in 2023 varied by the mRFEI measure used (Appendix Table A). The estimated numbers of people living in food deserts were approximately 476 000 (distance-based mRFEI), 0.5 million (drive time–based mRFEI), and 7.4 million (public transit–based mRFEI). The numbers of people living in food swamps were 26 million (distance-based mRFEI), 29 million (drive time–based mRFEI), and 18 million (public transit–based mRFEI).

Changes in the Foodscape by State

Despite some variation in foodscape changes between states (Figure 2, Appendix Figure D), most state-specific trajectories reflected nationwide patterns (i.e., a minimal decrease to a slight increase in food deserts and a substantial increase in food swamps). Between 2003 and 2023, the percentage of food deserts decreased in 26, 14, and 25 states using distance-based (Figure 2a), drive time–based, or public transit–based mRFEI, respectively (Appendix Figure D). The percentage of food swamps increased in 48, 38, and 47 states using distance-based (Figure 2b), drive time–based, or public transit–based mRFEI, respectively (Appendix Figure D). In the majority of states, the foodscape patterns during the COVID-19 period were similar to the national patterns. From 2020 to 2023, regardless of the mRFEI measure used, the percentage of food deserts generally increased, whereas the percentage of food swamps decreased (Appendix Tables B–D).

FIGURE 2—

State-Specific Change in the Proportion of (a) Food Deserts and (b) Food Swamps Using Distance-Based mRFEI: United States, 2003–2023

Note. mRFEI = modified Retail Food Environment Index.

Geographically, by 2023, regardless of the mRFEI measure used, states with the highest percentage of food deserts were predominantly located in the North and Great Plains regions, whereas those with the lowest percentage were concentrated in the mid-Atlantic region and along the West Coast (Figure 3, Appendix Figure E). States with a lower percentage of food deserts often had the highest percentage of food swamps. In 2023, approximately 14.1 million people (4.4% of the total population) in the United States resided in food deserts and 292.1 million (90.7%) resided in food swamps (Appendix Table A). Across states (Appendix Tables B–D), Texas had the largest population living in food deserts (> 1 million; 3.6% of the state population). The largest populations living in food swamps were found in California (36 million; 91.2% of the state population) and in Texas (26.7 million; 94.6%).

FIGURE 3—

Geographic Distribution of (a) Food Deserts and (b) Food Swamps Using Distance-Based mRFEI: United States, 2023

Note. mRFEI = modified Retail Food Environment Index. NA indicates that the percentage of census tracts is 0 (only in the District of Columbia).

DISCUSSION

Our findings indicate little progress in reducing the number of food deserts and a substantial increase in the number of food swamps in the United States in the past 2 decades, resulting in at least 4.9 million people living in food deserts and up to 314 million people living in food swamps. This trend is largely driven by a substantially greater increase in the number of restaurants and fast-food locations compared with grocery retailers nationwide, which is consistent with earlier findings.¹⁴ The slower growth in the number of grocery retailers may be attributed to the complexities of their business structure, including logistics, supply chains, and financial sustainability,²⁶ or to challenges finding appropriate property size, especially in urban areas.¹⁹ Similar to previous research, we found that among healthy food retailers, the greatest increase was observed in the number of club stores and department stores,²⁷ many of which operate as franchises or chains. Although the establishment of such large club “superstores” may allow them to serve a broader customer base and act as anchor locations for several areas, their presence does not necessarily improve the geographic accessibility to healthy foods.²⁷ These stores are frequently located in commercial districts rather than residential neighborhoods, which may limit their role in alleviating food deserts,²⁸ especially for individuals with limited access to transportation. Additionally, club stores generally require a membership fee, which may not be affordable to some individuals; therefore, they are not considered a potential source for grocery shopping.^27,29 Public–private healthy food financing programs to improve access to grocery retailers have been implemented at the federal and state levels. For example, the Healthy Food Financing Initiative and related programs (including Pennsylvania’s Fresh Food Financing Initiative) provide grants and low-interest financing to support the development of healthy grocery retailers in rural and urban underserved areas.^30–32 Such investments could improve healthy food access in food desert settings and may yield measurable benefits for residents.³³ Additionally, local policies to control the growth of unhealthy food retailers have been proposed and implemented across many US jurisdictions, including zoning, land-use restrictions, or partial and temporal bans on the opening of new fast-food retailers.³⁴ However, their success must be further evaluated because many policies, especially bans, have generally been subject to strong opposition from businesses and other stakeholders.³⁴ Moreover, although higher taxes on unhealthy foods may not directly change the neighborhood foodscape, it has the potential to reduce unhealthy food consumption.³⁵

Although the overall pattern of foodscape changes was similar across urban and rural areas, food deserts were disproportionately concentrated in rural areas regardless of the mRFEI measure used, but especially when we used public transit-based mRFEI because of generally underdeveloped public transit networks in rural areas. This disparity is likely attributable to the larger sizes of census tracts in rural areas, which results in longer travel distances to healthy food retailers.³⁶ Additionally, rural tracts with small populations may lack the consumer base necessary to sustain large grocery stores, further exacerbating limited food access.³⁷

Our findings also highlight that accessibility to healthy food retailers varies by the mode of transportation. When using a private automobile, only a small fraction of the US population lacks access to healthy food retailers. Automobile-based measures, however, assume universal car ownership and may underestimate inequities in food access because many individuals do not have access to a private automobile³⁶; approximately 8% of the US population and 19% of individuals with incomes below the federal poverty level do not own a car.³⁸ Our findings also show that populations living in persistent poverty areas have more limited accessibility to grocery stores when relying on public transit only. Additionally, the majority of census tracts with available public transit connections were categorized as food swamps, confirming previous findings that many public transit users, especially bus riders, have easier access to unhealthy than to healthy food retailers.³⁹ Prior work also suggests that food shopping may be influenced by the design of the public transit routes and their interaction with access to fast-food retailers and convenience stores.³⁹ Evidence suggests that transit-focused interventions—such as modifying fixed-route bus service to connect socioeconomically disadvantaged areas to supermarkets⁴⁰ or establishing grocery shuttles and “shopper links”⁴¹—could enhance access to healthy food options for public transit users.

The nationwide foodscape trends were also largely reflected at the state level. In general, by 2023, states with the highest percentages of food deserts were predominantly located in the North and Great Plains regions (including states with the highest number of food deserts nationwide³⁷), whereas those with the lowest percentages were concentrated in the mid-Atlantic region and along the West Coast. We also observed that states with a lower percentage of food deserts often had a higher percentage of food swamps. This pattern may be related to population distribution. States with higher population densities and large urban centers tend to have a greater number of food retailers, particularly fast-food restaurants and convenience stores, but also healthy food retailers.⁴² Thus, having fewer deserts in a state may not necessarily mean having a healthier food environment. Additionally, a dense network of major arterial roads may contribute to a higher proportion of food swamps, as these areas often attract fast-food establishments serving the bypassing traffic.⁴³

The COVID-19 pandemic coincided with substantial changes in the foodscape, as the nationwide lockdown, reduction of dining-in opportunities, and decrease in the number of customers resulted in the closure and bankruptcy of many food retailers. The reduction of healthy food retailers during the pandemic is likely among the reasons for an overall increase in the number of food deserts during the pandemic and postpandemic periods, especially in rural areas, leaving many residents with very limited accessibility to healthy food. By contrast, a substantially larger decrease in the number of unhealthy food retailers resulted in an overall decrease in the number of food swamps. The number of food retailers should be monitored for a longer period, as some businesses continue to recover from the economic impact from the COVID-19 pandemic. Future monitoring should also account for structural changes such as the growth of e-grocery shopping and food delivery platforms,⁴⁴ which may reshape access independent of physical retail density. Increasing the advertising of food deliveries (beyond fast-food and restaurant delivery) and the number of grocery retailers participating in food delivery programs may be beneficial for populations with limited geographic accessibility to healthy food and for individuals with limited mobility.

Limitations

This study is the first to examine temporal changes in the US neighborhood-level foodscape, nationally and by area rurality, persistent poverty status, and state, over the past 2 decades, including the COVID-19 pandemic era. However, several limitations related to data and methodology should be acknowledged. First, some businesses might have changed their NAICS or SIC codes over time, potentially leading to misclassification within specific retail subcategories. For example, although the number of ethnic grocery stores appeared to decrease, many were reassigned to other categories, such as fruit and vegetable stores or meat markets. However, these reclassifications did not affect the broader “healthy” or “unhealthy” classification of food retailers. Second, defining the foodscape at the tract level may not fully capture actual food accessibility and availability. Additionally, using the number of food retailers alone does not allow us to evaluate the quality and volume of the supplied foods and may not guarantee a healthy dietary pattern. According to a study on the consumption behavior of approximately 60 000 US households in 2020, 48% of purchased foods in supermarkets are considered unhealthy ultraprocessed foods.⁴⁵ Higher taxes or implementing front-of-package warning labels, similar to practices in some other countries, may limit the consumption of these foods.^46,47

We also did not account for potential changes in access to healthy food retailers among the commuting population. However, a study from Florida reported only minimal changes in accessibility to healthy food retailers when considering commuting patterns as compared with primary residency only.⁴⁸ We were also not able to include digital food access (e.g., delivery zones) into our study because of the limited data. Although a recent report by the Brookings Institution states that over 90% of the US population should have access to at least 1 grocery delivery option,⁴⁹ actual use of these services depends on various factors, including broadband access or number of consumers in the areas, and thus remains largely unknown.⁵⁰ Lastly, there is no universally accepted mRFEI threshold for defining food swamps. Despite this, the substantially greater increase in the number of restaurants and fast-food locations compared with healthy food retailers will result in an increase in the number of food swamps, regardless of the threshold used.

Conclusions

In conclusion, our findings highlight the need to improve access to healthy food retailers across the United States, particularly in rural and socioeconomically disadvantaged areas. Expanding public–private partnerships to establish new healthy food retailers in food deserts, including in sparsely populated areas or areas with lower-income populations, may help increase access to healthy foods. Additionally, further expansion and strategic planning of public transit networks could help reduce accessibility barriers for socioeconomically marginalized populations.