The worst food deserts in America

How this ranking is computed

The USDA Economic Research Service Food Access Research Atlas classifies every U.S. census tract against a single distance-to-supermarket rule: more than one mile from the nearest large grocery store in urban tracts, or more than ten miles in rural tracts. A county's "low food access percentage" is the share of its residents who live inside any tract that crosses that threshold. The ranking below sorts every U.S. county by that percentage from highest to lowest and shows the 100 counties at the top of that list. Because every figure is computed from the same underlying tract-level table, a county's position can be reproduced row-for-row by reading the USDA file directly – a property food-policy researchers and county health departments rely on when citing a specific number.

Three structural patterns surface in the top of this list. The first is sparse-rural geography: counties in Alaska, Nevada, parts of the Mountain West, and the Great Plains often place near the top because their populations are small and supermarkets are concentrated in one or two regional hubs, which leaves most residents outside the ten-mile rural threshold. The second is small-town consolidation: counties whose historic main-street grocers closed in the last two decades and were not replaced by a regional chain show high shares even when total population is moderate. The third is high-poverty urban tracts: in dense metros, the rural-distance rule does not apply, but the urban one-mile threshold flags neighborhoods where the closest supermarket is past a transit barrier – an interstate, a river, or a low-frequency bus corridor.

Distance alone is not the same thing as a USDA-designated "food desert". A tract becomes a food desert only when low-access overlaps with low-income status, defined by the Census Bureau American Community Survey 5-Year Estimates. That is why the table also surfaces SNAP participation and the share of households without a vehicle: SNAP participation acts as a practical low-income proxy, and no-vehicle households convert any measured distance to the supermarket into a real obstacle on foot, by transit, or by relying on a neighbor's car. A county near the top of the low-access ranking that also reports above-median SNAP and above-median no-vehicle share is the canonical food desert profile.

The ranking refreshes with each USDA Atlas vintage. Between vintages, individual tract scores can move because of supermarket openings and closures, redistricting of census tract boundaries, or revised Census ACS denominators. Counties that appear and disappear across vintages are often near the threshold rather than deep in the food desert tail. For a stable cross-vintage view, food-policy analysts typically read both the absolute low-access share and the count of low-access tracts as a fraction of total tracts in the county, both of which are surfaced on the per-county detail pages linked from the table below. Source URLs for the underlying files are listed in the data citation at the bottom of this page.

Two follow-on questions almost always come up after reading this ranking. The first is how a specific county at the top of the list compares to its state median: the per-county detail pages answer that by showing the county figure alongside the state-rolled-up average, which lets a reader see whether the county is the worst in an otherwise low-access state or the worst in an already-stressed region. The second is how the ranking would shift if SNAP participation or no-vehicle households were used as the sort instead of distance: the sibling rankings linked at the bottom of this page reorder the same county set against those alternative criteria, and many counties at the top of one list also appear near the top of the others – a useful sanity check that the underlying poverty signal is real rather than an artifact of a single indicator.

A common misuse of food-desert ranking data is to treat the headline figure as a fixed attribute of a county rather than as a snapshot of a specific USDA vintage. The data file refreshes on a multi-year cycle, and counties near a classification threshold can move when a single supermarket opens or closes, when a census tract boundary is revised, or when the ACS denominator is updated. For policy work that needs to track change over time, the recommended approach is to pin the comparison to a specific Atlas vintage and to use the absolute count of low-access tracts rather than the population-weighted share, because the absolute count is less sensitive to ACS denominator revisions. For descriptive reporting, the headline share is usually fine, with the caveat that any single-year reading is one observation in a noisy time series rather than a permanent label. Each county's detail page surfaces both the share and the absolute tract count so a reader can choose the framing that best fits the question being asked.