Every ballot page on this site lists the counties inside each congressional district. Those lists are computed here from Census boundary files rather than typed in, which removes one class of error and introduces another. Two of them are worth writing down, because a reader in Connecticut or Alaska will notice both.
Connecticut's counties have not existed since 1960
Our Connecticut pages name Fairfield, Hartford, Litchfield, Middlesex, New Haven, New London, Tolland and Windham. Connecticut abolished county government in 1960. Those eight names survived for another sixty years purely as statistical containers, and on 6 June 2022 the Census Bureau retired them too, adopting the state's nine planning regions as county equivalents in their place: Capitol, Greater Bridgeport, Lower Connecticut River Valley, Naugatuck Valley, Northeastern, Northwest Hills, South Central, Southeastern and Western.
There is no clean crosswalk between the eight and the nine. They are not merges or splits of one another; the planning regions were drawn around councils of governments and follow town lines that cut across the old boundaries. So the fix is not a rename, and until the boundary file we ship is replaced with a post-2022 vintage, a Connecticut reader is being shown the geography of their grandparents.
The old names are still widely used in conversation and still appear in plenty of federal datasets, so they are not meaningless. They are just not administrative units, and have not been for sixty-six years.
Alaska is missing Aleutians West
Alaska has one at-large district. It therefore contains all twenty-nine of the state's boroughs and census areas, by definition, with no possible exception. Until 4 September 2026 our data listed twenty-eight.
The missing one was Aleutians West, and the cause is the 180th meridian. Aleutians West runs from longitude 172.46 east to 166.09 west, across the antimeridian, out to Attu. The containment test used to build these lists is spherical, and asking it whether the Alaska district polygon contains the Aleutians West centroid at 173.64 west returns false. Asking the same test about Anchorage returns true. The polygon is fine and the census area is fine; the wrap between them was not handled.
This one is now fixed, and not by patching the geometry. An at-large district is the state, so it contains every county equivalent in that state with no test required. The six at-large states are now filled in from the state's county list directly, and a check in the build asserts that invariant so a regenerated boundary file cannot quietly drop one again. Alaska lists twenty-nine.
A third one, smaller, that explains five blank lists
Five districts show no counties at all: CA-15, CA-36, CA-42, FL-14 and FL-27. All five are coastal, all five are smaller than the county they sit in, and all five have a geometric centroid that lands in water. FL-14's falls in Old Tampa Bay, CA-15's in San Francisco Bay, CA-36's in Santa Monica Bay just west of Torrance.
The fallback test, whether the district sits wholly inside one county, fails for a duller reason. The county file and the district file are simplified independently, so their coastlines disagree by a few hundred metres. Of CA-36's 115 boundary vertices, 96 fall inside Los Angeles County and 19 fall just outside it, into water that the other file calls land. A strict containment test reads that as a district crossing a county line.
None of these three affect a race page's dates, offices or candidate names, which come from official sources rather than from geometry. They affect the county lists, which is a smaller claim than it sounds like and still wrong.
If you find another one, the contact page reaches one person and I read it.