What is a county seat, and why do 21 places have none?

A county seat is the town or city where a county’s government physically sits: the courthouse, and the offices that keep the county’s records. That is the whole definition. Most of the 3,144 county-equivalents in the 50 states and the District of Columbia have that field filled in with the name of a real place. For 54 of them the answer is the entity itself. For 21 it cannot be filled in at all, because there is no county government to have a seat.

What is actually in the building

Strip away the architecture and a county seat is a filing cabinet with a jurisdiction attached. The county clerk keeps the county’s own records. The recorder keeps deeds, mortgages, liens and plats, which is why every property transaction in the county eventually ends up in that office. The assessor values the land for tax. The sheriff runs the jail. The division of duties varies by state, and some states fold several of these offices into one, but the principle holds: the seat is where the county’s paper lives. The trial court sits in the same building or next door.

That explains why the seat mattered so much. If you wanted to prove you owned your farm, marry, sue, serve on a jury, probate your father’s will or pay your taxes, you went to one specific town, and only that town would do. For most of the nineteenth century you got there on horseback or in a wagon.

Why the courthouse decided where towns grew

The standard justification for county size in the settlement era was that a farmer should be able to reach the courthouse, do his business and get home inside a day. That is one reason counties tend to be smaller as you move east and older, and larger as you move west into country surveyed after the railroad arrived. Texas has 254 counties, more than any other state, and Georgia 159, the next highest; Delaware has 3 and Rhode Island 5. The travel argument is the traditional explanation for the high counts, though it was never the only force at work. In Georgia the old county unit system gave each county weight in statewide primaries, which made creating more of them worth doing for reasons that had nothing to do with horses.

Whichever town won the seat got the lawyers, the newspaper, the hotel, the land office trade and the crowd that came in on court day. The town that lost often got nothing, and in a fair number of cases simply stopped existing. The contests were correspondingly ugly. Elections were held and then disputed and rerun. Records were moved out of one courthouse at night and into a rival town’s building before anyone could get an injunction. In Kansas the seat contests of the settlement era turned violent more than once; the Gray County fight between Cimarron and Ingalls is the example usually cited, and accounts of it disagree on nearly every detail except that the county records were taken by force. The general pattern is not in dispute: the seat was worth fighting over, and people fought.

The 21 with no seat at all

The 21 blanks are not missing data. They are the county-equivalents that have no county government, so there is nothing for a seat to be the seat of.

Group	Count	Why there is no seat
Alaska census areas	11	Census Bureau subdivisions of the Unorganized Borough, created for statistics, never governments
Connecticut planning regions	9	Replaced the eight legacy counties in the 2022 Census vintage; county government in the state ended in 1960
Kalawao County, Hawaii	1	Population 82; the mayor is by law the Director of the Hawaii Department of Health

The Alaska eleven are Aleutians West, Bethel, Chugach, Copper River, Dillingham, Hoonah-Angoon, Kusilvak, Nome, Prince of Wales-Hyder, Southeast Fairbanks and Yukon-Koyukuk. Alaska has 30 county-equivalents in total, and the other nineteen (13 boroughs, 4 cities and boroughs, 2 municipalities) do have governments and do have seats. The census areas cover the middle of the state that has never been organized into boroughs, and their status as statistical constructs shows in how freely they change. Chugach (02063) and Copper River (02066) replaced the Valdez-Cordova Census Area (02261) effective 2 January 2019, and Kusilvak (02158) replaced Wade Hampton (02270) on 1 July 2015. When Petersburg Census Area became Petersburg Borough it kept GEOID 02195 and acquired a government, and a seat, on the way. Yukon-Koyukuk alone covers 145,767.49 square miles of land, which dwarfs San Bernardino County in California, the largest entity outside Alaska at 20,068.42.

Connecticut’s nine planning regions are groups of towns, and they are not coterminous with the eight counties they replaced. That is why no county to region crosswalk exists and only a town-level one works, and why the 2020 Decennial Census products, still published on the eight legacy counties, cannot be joined to the regions at all. If you are working with Connecticut data across time, that distinction will bite you long before the county seat field does. There is more on it in why Connecticut replaced its counties.

Note that 35 entities have no county government while only 21 have no seat. The fourteen in between are Rhode Island’s 5 counties, which lost their governments in 1846, and the 9 Massachusetts counties abolished between 1997 and 2000. Those fourteen still carry a nominal seat, though the dataset does not record why the name survived the government there and not elsewhere. Counties without county government sets out the full 35.

The 54 that are their own seat

The second category is the opposite problem. In 54 cases the seat and the entity are the same government, so the field is a tautology rather than a blank. These are the 40 independent cities, 38 of them in Virginia plus Baltimore city in Maryland and St. Louis city in Missouri, together with the consolidated city-counties and unified governments where a city and a county merged into one body. Carson City, Nevada belongs with them in practice, although the Census files give it no type word at all: its name is literally Carson City.

Filling the field in with the entity’s own name is correct, and it is also why joining a county table to a city table on name alone produces nonsense in these rows. Virginia is the hard case. It has 133 county-equivalents, 95 counties and 38 independent cities, and four names exist as both a county and a city, so Fairfax County (51059) and Fairfax city (51600) are different governments in different rows, as are the two Franklins, the two Richmonds and the two Roanokes. See what an independent city is for how that split happened and what it does to your joins.

The two rows most datasets drop

Arlington and Henrico in Virginia are ordinary counties with ordinary county governments, but their seats are unincorporated places rather than incorporated towns. They go missing from most bulk county seat files. The likely reason is that such files are built by matching the seat name against a table of incorporated places, so these two match nothing and fall out without an error; the files themselves rarely say. If a dataset comes up two seats short of the total, this is the first thing to check.

A blank is not a bug

Situation	Entities	County seat field
Seat is a separate town or city	All the rest	Name of that place, unincorporated in the Arlington and Henrico cases
Entity is its own seat	54	The entity’s own name
No county government	21	Empty
Total, 50 states and DC	3,144

County seat is a field that cannot be filled for every row, and the reason is not sloppy collection. It is that American county-equivalents are not one kind of thing. They are counties, parishes, boroughs, census areas, planning regions, independent cities and a handful of one-off arrangements, and some of them are statistical containers with no government inside. A dataset that prints an empty cell for Yukon-Koyukuk or the Capitol Planning Region is telling you the truth. One that prints a plausible town name there has invented it. If you want the underlying taxonomy first, start with what a county is and the seat question mostly answers itself.