Methodology

The county universe

This site covers 3,144 counties and county-equivalents across the fifty states and the District of Columbia. Puerto Rico's 78 municipios and the island areas are deliberately out of scope: they have a different data cadence, with no annual American Community Survey outside Puerto Rico, and including them would mean pages where most fields are permanently empty.

The list itself is derived from the Census Bureau's Vintage 2025 population estimates file at summary level 050 and cross-checked against the 2025 cartographic boundary file and the 2025 Gazetteer. All three agree at 3,144. The build fails if they ever stop agreeing.

We deliberately do not use the Census Bureau's ANSI code-list page, which is the obvious source and is frozen at the 2020 vintage. It still contains Connecticut's eight retired counties and none of the nine planning regions that replaced them. Most third-party county datasets and packages inherit that error.

Where each figure comes from

What is computed rather than published

Two figures on this site are ours, not the Census Bureau's, and both are labelled as such wherever they appear.

Population density divides the Vintage 2025 population estimate by the 2025 Gazetteer land area. Those are two different Census products and combining them is a choice, not a published statistic. We say so on every page that shows a density. This matters: the most widely read general reference computes its county density against the 2020 census count while displaying the 2025 estimate directly above it, which puts its figures out by up to six percent.

Rank positions, within a state and nationally, are computed from the figures on the page. Where a county has no published value for a measure it is excluded from that ranking rather than treated as zero, which is why some rankings show a denominator below 3,144.

How the maps are made

Boundaries come from the Census Bureau's cartographic boundary files at 1:500,000, not from raw TIGER/Line. TIGER county polygons extend to the legal boundary, which runs out into territorial water: rendered as a map, TIGER fills in the whole of Chesapeake Bay and fuses Virginia's Eastern Shore to the mainland. The cartographic files are pre-clipped to the shoreline and are what a reader expects a county map to look like.

We also do not use the widely-used pre-built US atlas TopoJSON. It is derived from a 2017 edition and still carries eight Connecticut counties and 29 Alaska boroughs. For a site whose entire product is per-county pages keyed on FIPS, that would make every Connecticut page wrong.

Simplification is done per state, with an absolute distance interval computed from that state's own metres-per-pixel at the width the map actually renders. The target is half a pixel of maximum deviation, so the simplification is provably invisible. A single national percentage would instead produce one uniform threshold of roughly 494 metres, which is about half a pixel across Virginia but around 24 pixels across the District of Columbia, where it erases the Potomac frontage entirely.

Simplification runs on a shared-topology structure, so a border between two counties is thinned once and both sides receive the identical line. Simplifying each county independently opens gaps: measured across Ohio's 88 counties, it produced 114 sliver holes totalling 126 square kilometres.

The keep-shapes option is always on. Without it, 43 counties reduce to empty geometry at aggressive simplification levels, including San Francisco, Staten Island, Nantucket, Kalawao and 24 Virginia independent cities. A county map that renders San Francisco as nothing is not a map.

Handling missing data

A missing value is never rendered as a blank cell, as zero, or as the word "undefined". It renders as Not available with the reason attached, drawn from a per-field availability flag that distinguishes four different situations:

• Suppressed: the Census Bureau withholds it to protect privacy at that population size.
• Not collected: no authoritative source publishes it for that area.
• Below threshold: the area is under the 65,000 population floor for a 1-year American Community Survey estimate. Only 845 of 3,144 entities clear that bar.
• Pending: we have not loaded it yet.

Twenty-one entities have no county seat, and that is correct rather than missing: they are the twenty-one that have no county government. Fifty-four entities are their own seat, because the city and the county-equivalent are the same government.

328 counties have populations under 5,000, where American Community Survey margins of error can approach or exceed the estimate itself. Their pages carry a warning saying so. 36 entities have populations under 1,000; their pages are excluded from search indexing because there is not enough published data to make them genuinely useful.

How often this updates

The pipeline is run by hand and its output is committed to the repository, so every change to every figure is reviewable before it ships. It is not run during the site build: if the Census Bureau's servers are down, the site still deploys with the last reviewed data rather than failing or silently shipping something incomplete.

Population estimates are published each March, the American Community Survey each December, and the cartographic boundary files each spring. The full refresh schedule is here.

Corrections

If a figure here is wrong, we want to know. The fastest way to get it fixed is to say which page, which figure, and what you believe the right value is. Contact details are here. Corrections to source data go upstream to the Census Bureau; corrections to our processing get fixed and the page's rebuild date changes.

Pages on this site are generated from federal datasets by an automated pipeline and reviewed by The County Maps US editorial team. The prose on explainer pages is written and edited by hand. The figures on county, state, ranking and comparison pages are generated.