Obesity is shaped by many interacting factors: food prices and availability, sleep, physical activity, stress, medical conditions, genetics, medications and the broader environment. A county-level analysis cannot identify what caused an individual person’s weight. It can, however, reveal whether community conditions tend to appear together and which places contradict the national pattern.
Why compare food deserts with food swamps?
A food-desert measure asks whether people—especially people with low incomes—live far from a supermarket or large grocery store. That matters, but distance alone does not describe everything people encounter. A neighbourhood may have a grocery store and still be saturated with quick-service restaurants, convenience outlets and dollar stores.
The important question is not only “Is healthy food far away?” It is also “What choices dominate the local retail environment?”
For this analysis, BMITDEE builds two transparent descriptive scores. The food-desert score combines low-income/low-access population, households without a vehicle and low access, and the inverse of grocery-store density. The food-swamp score averages the national percentiles for fast-food restaurants, convenience stores and dollar stores per 1,000 residents.
A widely cited 2017 county-level study reported that its food-swamp measures predicted obesity rates more consistently than supermarket absence alone. This BMITDEE investigation is not a replication: it uses newer official releases and publishes its own clearly labelled percentile composites so readers can inspect the pattern and its exceptions.
Interactive test
Change the hypothesis and inspect every county
Choose a food-environment indicator. The chart recalculates the raw correlation and a partial correlation after accounting for income, poverty, metro status and population.
The average relationship is only half the story
A national correlation compresses thousands of places into one number. Counties with similar food environments can have very different obesity estimates, and counties with similar obesity estimates can have very different retail landscapes. Those exceptions are where local policy, culture, income, transport, healthcare and other factors may be especially important.
Two-dimensional view
Food desert and food swamp are not the same thing
The horizontal axis shows the food-desert score; the vertical axis shows the food-swamp score. Darker points represent higher adult obesity estimates.
Find counties that challenge the simple explanation
If food environments were the only explanation, counties would fall neatly along one line. They do not. Use the outlier finder to locate places that fit or contradict the expected pattern.
Outlier finder
Where the pattern holds—and where it breaks
Quartiles are recalculated from the joined national dataset. Select a group to inspect counties.
| County | Obesity | Food swamp | Food desert | Poverty | Metro |
|---|
What does the pattern look like in your county?
Search for a county to see its national percentiles and compare it with counties that have similar population, income, poverty and metro status. This matched-peer view is more informative than comparing a small rural county directly with a large urban county.
County profile
Search, compare and share a local view
Type a county name, state abbreviation or five-digit FIPS code.
Select a county to generate its profile and matched peers.
What the analysis can—and cannot—tell us
A positive relationship means counties with higher values on one measure tend, on average, to have higher values on another. It does not prove that adding or removing a particular store would change obesity. The datasets cover different observation years, individual dietary purchases are not observed, and county averages cannot capture neighbourhood variation.
The adjusted statistic is also descriptive. It removes the linear relationships of income, poverty, metropolitan status and population from both variables, then correlates the remaining variation. It does not control for every possible confounder and should not be interpreted as a causal effect.
Transparent methodology
How BMITDEE built this analysis
Health outcome
Adult obesity crude prevalence from CDC PLACES, 2025 release, primarily based on 2023 BRFSS data and model-based small-area estimation.
Food access
USDA Food Environment Atlas indicators for low food-store access and low-income/low-access populations, observation year 2019.
Retail exposure
Grocery, convenience, dollar-store and fast-food densities from the USDA Atlas, observation year 2020.
Adjustment variables
Median household income and poverty rate (2021), metropolitan classification (2023), and CDC total population.
Composite scores
Each input is converted to a national percentile. Food-swamp score averages fast-food, convenience and dollar-store percentiles. Food-desert score averages low-income/low-access, no-vehicle/low-access and inverse grocery-density percentiles.
Missing data
USDA N/A, −9999 and −8888 codes are treated as missing. A composite requires at least two valid components. Results use counties with valid values for the selected analysis.
This is an independent BMITDEE analysis. BMITDEE is not affiliated with or endorsed by CDC, USDA or the Census Bureau.
Download and reproduce the work
The food-environment file below is a compact, machine-readable extract of selected official USDA indicators. The joined analysis download is generated in your browser after the current CDC records load.
Sources
- CDC PLACES: Local Data for Better Health, County Data, 2025 release.
- USDA Economic Research Service, Food Environment Atlas, current downloadable version updated July 30, 2025.
- USDA Food Access Research Atlas for definitions and the underlying low-access framework.
- Cooksey-Stowers, Schwartz and Brownell (2017), the earlier county-level food-swamp comparison that motivated the updated question.