7 rural US counties lost their last FDIC-insured bank branch between 2016 and 2026; 3 have no depository branch of any kind left

Summary

Using FDIC branch-level deposit data (Summary of Deposits) joined to USDA ERS's 2023 Rural-Urban Continuum Codes, 7 nonmetro (rural) counties in the 50 states went from having at least one FDIC-insured bank branch in 2016 to having zero in 2026: Coosa County AL, Sierra County CA, Lincoln County ID, Keweenaw County MI, Blaine County NE, Jones County NC, and Tyrrell County NC. (An eighth unit, the Eastern District of American Samoa, meets the same criterion but is a territorial district, not a county in the 50-state sense, and is reported separately.) Of those 7, 4 now have at least one NCUA-insured credit union branch that gives residents a physical depository-institution alternative (Coosa AL, Lincoln ID, Jones NC, Tyrrell NC). The remaining 3 — Sierra County CA, Keweenaw County MI, and Blaine County NE — have no bank or credit union branch at all, leaving residents dependent on online/mobile banking, ATMs belonging to institutions headquartered elsewhere, or travel to a neighboring county for any in-person depository service.

Method

Rurality: geo.rural_urban_continuum (USDA ERS RUCC, 2023 vintage, based on 2020 Census population), filtered to metro_nonmetro = 'Nonmetro' (RUCC codes 4-9). This is a single, decade-spanning classification (RUCC updates only every ~10 years) applied across the whole window, consistent with how the source itself is published.

Branch presence: banking.sod (FDIC Summary of Deposits, an annual June-30 branch census), counting distinct uninum (unique office numbers) per county in 2016 and in 2026, the two endpoints of the trailing ten-year window this table actually has loaded (data run through 2026-09-28; both years have full national coverage, ~76-92K branch-year rows).

Disclosed exclusion: the query applied a county_fips IS NOT NULL filter (banking.sod carries a null county_fips only for foreign/non-US offices, since a domestic branch always resolves to a county). Checked directly: this filter excluded/removed 0 rows in 2016 and 0 rows in 2026 — every domestic branch row in both years already carried a real county_fips, so the exclusion changed nothing about the 7-county result; it is reported here for full transparency about every predicate the query used, not because it altered the count.

Definition used: any county with >=1 branch in 2016 and 0 branches in 2026 — not restricted to counties that had exactly 1 branch to start. A county that had two branches and lost both has just as certainly lost its last branch as one that had exactly one; restricting to exactly-1-to-zero would have undercounted the true population of "lost the last branch" counties.

Interim-year check: for all 7 counties, branch counts were also pulled at 2018/2020/2022/2024 to confirm the decline to zero was a genuine multi-year drawdown (not a one-year data blip that bounced back) — in every case branches stepped down monotonically (or held, then dropped) toward zero and stayed there through 2026.

Replacement check: banking.ncua_branch_locations (NCUA credit-union office snapshot, current) filtered by county name and state for each of the 7 counties, to see whether a credit union branch — the most common formal substitute for a closed bank branch — is present today.

The 7 counties, in detail

For the 3 counties with no formal replacement, the population base is very small (431-3,236 people), which is itself part of why no bank or credit union has found it commercially viable to maintain a branch. What likely substitutes for a branch in practice, based on general industry patterns (not directly measurable in this corpus): online/mobile banking through the resident's existing account, out-of-county ATMs, USPS money orders for basic cash transactions, and periodic trips to the nearest branch in an adjacent county.

How this compares to the published literature

Independent research on this topic uses different metrics than a literal zero-branch count, so figures are not directly comparable. A widely cited Federal Reserve analysis found that 44 "deeply affected" counties nationally had 10 or fewer branches in 2012 and lost half or more of them by 2017, 39 of them rural — a percentage-decline threshold, not a to-zero threshold, and over an earlier and shorter (2012-2017) window (Fed Communities, "The Last Bank Branch Standing"). That same period saw rural counties lose branches at roughly 14% versus about 9% for urban areas. More recent Fed and industry reporting (2025-2026) describes an accelerating pattern in Appalachia and the rural Midwest, with some residents now traveling 20+ miles for in-person banking, and documents rural mayors organizing in 2025 to push back on unchecked closures (Banking Dive, "Bank branch closures leave deeper void in rural counties, Fed says"; Banking Dive, "Keeping the banking desert at bay in rural America"). The Philadelphia Fed and academic work define a rural "banking desert" as a census tract more than 10 miles from any bank branch — a distance-based definition, again distinct from the county-level, branch-count-to-zero measure used here (Philadelphia Fed). A farmdoc daily analysis specifically tracking rural branch counts over time corroborates a sustained multi-year decline in rural commercial bank branch counts nationally (farmdoc daily, "Changes in the Number of Commercial Bank Branches in Rural Areas"). None of these sources publish a literal "counties that hit zero branches" list, which is why this report computed that figure directly from FDIC's own branch-level data rather than relaying a secondary count.

Caveats

• Small n. 7 counties is a small population; each is a single, named case rather than a statistical sample, and the finding should be read as an enumerated list, not a rate with a confidence interval.
• Grain and definition choices matter. A different rurality classification (e.g., tract-level RUCA rather than county-level RUCC), a different window (e.g., 2014-2024 or 2015-2025), or FDIC's institution-level rather than branch-level table could each shift the exact list by one or two counties at the margin — this is inherent to any small-count, threshold-based measure, not a defect in the data used.
• NCUA snapshot is current, not historical. The credit-union presence check reflects today's NCUA branch registry, not a year-matched historical snapshot, so it answers "is there a credit union branch there now" rather than "did a credit union open there specifically because the bank left." No causal claim about timing or motivation is made.
• The American Samoa district (Eastern District, 60010) met the identical >=1-to-0 criterion (3 branches in 2016 to 0 in 2026) but is a territorial district rather than one of the 3,142 counties/county-equivalents in the 50 states, so it is reported separately rather than folded into the headline count of 7.
• "Replaced" is defined narrowly as a formal depository-institution branch (bank or credit union). It does not capture non-branch access some residents may use — online banking, fintech apps, retail check-cashing, or a bank's own ATM network without a staffed branch — none of which is measurable at county grain in this corpus.

Every query behind this report

7 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT MIN("year"), MAX("year") FROM banking.sod

SELECT "year", COUNT(*) AS n FROM banking.sod WHERE "year" IN ('2015','2016','2023','2024','2025','2026') GROUP BY "year" ORDER BY "year"

WITH b2016 AS (
  SELECT county_fips, COUNT(DISTINCT uninum) AS branches_2016
  FROM banking.sod
  WHERE "year" = '2016' AND county_fips IS NOT NULL
  GROUP BY county_fips
),
b2026 AS (
  SELECT county_fips, COUNT(DISTINCT uninum) AS branches_2026
  FROM banking.sod
  WHERE "year" = '2026' AND county_fips IS NOT NULL
  GROUP BY county_fips
),
rural AS (
  SELECT county_fips, county_name, state_fips, population, rucc_code, rucc_description
  FROM geo.rural_urban_continuum
  WHERE metro_nonmetro = 'Nonmetro'
)
SELECT r.county_fips, r.county_name, r.state_fips, r.population, r.rucc_code,
       COALESCE(b16.branches_2016,0) AS branches_2016,
       COALESCE(b26.branches_2026,0) AS branches_2026
FROM rural r
LEFT JOIN b2016 b16 ON r.county_fips = b16.county_fips
LEFT JOIN b2026 b26 ON r.county_fips = b26.county_fips
WHERE COALESCE(b16.branches_2016,0) >= 1 AND COALESCE(b26.branches_2026,0) = 0
ORDER BY r.state_fips, r.county_name

SELECT county_fips, "year", COUNT(DISTINCT uninum) AS branches
FROM banking.sod
WHERE county_fips IN ('01037','06091','16063','26083','31009','37103','37177','60010')
  AND "year" IN ('2016','2018','2020','2022','2024','2026')
GROUP BY county_fips, "year"
ORDER BY county_fips, "year"

SELECT cu_name, site_name, county_name, state_abbr, city
FROM banking.ncua_branch_locations
WHERE (state_abbr='AL' AND county_name ILIKE '%Coosa%')
   OR (state_abbr='CA' AND county_name ILIKE '%Sierra%')
   OR (state_abbr='ID' AND county_name ILIKE '%Lincoln%')
   OR (state_abbr='MI' AND county_name ILIKE '%Keweenaw%')
   OR (state_abbr='NE' AND county_name ILIKE '%Blaine%')
   OR (state_abbr='NC' AND county_name ILIKE '%Jones%')
   OR (state_abbr='NC' AND county_name ILIKE '%Tyrrell%')
ORDER BY state_abbr, county_name

SELECT DISTINCT cu_name, county_name, state_abbr FROM banking.ncua_branch_locations
WHERE (state_abbr='CA' AND county_name ILIKE '%Sierra%')
   OR (state_abbr='MI' AND county_name ILIKE '%Keweenaw%')
   OR (state_abbr='NE' AND county_name ILIKE '%Blaine%')

SELECT "year", COUNT(*) AS null_county_rows FROM banking.sod WHERE county_fips IS NULL AND "year" IN ('2016','2026') GROUP BY "year"