Why did Connecticut replace its counties?

Connecticut did not abolish its counties in 2022. It abolished county government in 1960. What changed in 2022 was narrower and, for anyone working with data, far more disruptive: the Census Bureau agreed to stop using the eight old counties as Connecticut’s county-equivalent geography and to use nine planning regions instead, effective with the 2022 vintage. The counties had been lines on a map since 1960. Now they are not even that.

Both halves of the story get mangled. People who know about 1960 often assume the counties vanished then, which is why they are baffled to see Hartford County on a 2015 map. People who know about 2022 often assume the planning regions are the old counties under new names, which is worse, because it leads them to build a crosswalk that cannot exist.

What 1960 actually did

Connecticut runs local government through towns. Every town in the state belongs to exactly one planning region today and belonged to exactly one county before, and it is the town, not the county, that has been the working unit. In 1960 the legislature dissolved county government outright, and its functions moved up to the state or down to the towns. The county boundaries themselves survived, but only as statistical geography: lines that the Census Bureau and others drew on maps and hung data from.

That put Connecticut in the same category as Rhode Island, which abolished county government in 1846, and later Massachusetts, which dissolved nine of its fourteen county governments between 1997 and 2000. Nationally, 35 county-equivalents have no county government at all and 21 have no county seat. Connecticut’s nine planning regions are counted there, along with Alaska’s eleven census areas, all five Rhode Island counties and Kalawao County, Hawaii.

What 2022 actually did

Connecticut has had regional planning organisations for decades, built up by towns choosing which regional council to join. Those councils, not the counties, are the bodies through which the state’s towns handle regional business. The state asked the Census Bureau to recognise them as the official county-equivalent geography. The Bureau’s notice was published in the Federal Register on 6 June 2022, and the regions took effect as county-equivalents with the 2022 vintage of Census products.

The eight retired counties were Fairfield (09001), Hartford (09003), Litchfield (09005), Middlesex (09007), New Haven (09009), New London (09011), Tolland (09013) and Windham (09015). Those codes are retired, and nothing in a current county file carries them.

The nine planning regions

GEOID	Planning region	Population	Land area (sq mi)
09110	Capitol	994,115	1,027.35
09120	Greater Bridgeport	337,697	140.19
09130	Lower Connecticut River Valley	177,311	424.12
09140	Naugatuck Valley	463,349	412.78
09150	Northeastern Connecticut	98,096	553.90
09160	Northwest Hills	114,690	786.64
09170	South Central Connecticut	578,741	367.25
09180	Southeastern Connecticut	284,015	598.14
09190	Western Connecticut	640,482	532.15

Population figures are Vintage 2025 estimates from the Population Estimates Program, released on 26 March 2026; land areas come from the 2025 Gazetteer. Every region’s full legal name ends in “Planning Region”, so the first row is formally Capitol Planning Region. That phrase is now a legal type in its own right in Census files, with nine members, sitting alongside 2,999 counties, 64 parishes, 40 independent cities and 13 boroughs.

The codes carry a hint. County codes are assigned alphabetically within a state, which is why a rename usually forces a new one: Shannon County, South Dakota became Oglala Lakota County in May 2015 and moved from 46113 to 46102. Connecticut’s regions are also in alphabetical order, but they run from 110 to 190 in steps of ten, outside the 001 to 015 block the counties used. Nothing in the numbering implies a relationship between a region and a former county, because there is none.

Why no county to region crosswalk exists

This is the point that costs people the most time. The planning regions are groups of towns. The old counties were also groups of towns, but the two groupings were drawn for different purposes at different times, and they are not coterminous. A town that sat in one county can sit in a region most of whose other towns came from a different county. Because the boundaries cut across each other, there is no assignment of counties to regions, and no assignment of regions to counties, that is correct.

The only crosswalk that works is at town level. Every town belongs to exactly one region and belonged to exactly one county, so town-level data can be aggregated either way. If your source is already summarised to the eight counties, you cannot recover the regions from it. You have to go back to town, tract or block level and re-aggregate.

Northwest Hills Planning Region is the usual trap. At 786.64 square miles it covers roughly the ground people picture as Litchfield County, and it is the region most often described as Litchfield renamed. It is not. No planning region is a former county under a new name, and a region that is nearly a county is more dangerous than one that is obviously nothing like it, because treating the two as a single unit produces a quietly wrong number rather than a visible error.

The decennial data problem

The change took effect with the 2022 vintage, which came after the 2020 Census. The 2020 Decennial Census products for Connecticut are therefore published on the eight legacy counties and have not been reissued on planning regions. Meanwhile the Population Estimates Program, the American Community Survey and current cartographic boundary files all use the nine regions.

Connecticut’s decennial data and its estimates data consequently sit on incompatible geographies, and the incompatibility is silent. Nothing errors. You get a join that drops every Connecticut row, or worse, a partial join after someone has helpfully renamed things by hand. Anyone comparing a 2020 Census count against a later estimate for a Connecticut area has to work from towns, tracts or blocks, which nest inside both geographies. This is one case of a general rule: read a dataset’s vintage before trusting its geography.

The survey data is the easier half. The American Community Survey publishes 1-year estimates only for areas of 65,000 people or more, which nationally is 845 of 3,144 entities, leaving 2,299 with 5-year estimates alone. Connecticut is unaffected by that cut: the smallest planning region, Northeastern Connecticut, has 98,096 people, so all nine clear the threshold.

Why so many maps still show eight counties

Shapefiles get cached. A project that pulled a county boundary file in 2019 and committed it to a repository will keep drawing eight Connecticut counties forever. Non-Census sources also run on their own update cycles, so postal data, property listings, weather products and state agency files each moved, or did not move, on their own schedule, and some may still key to the old counties for reasons of their own. And a great deal of historical work is correctly on the old counties, because that is the geography the data was collected on.

An eight-county Connecticut map is not automatically wrong, then. It is wrong only when it claims to be current. The quickest check is the identifiers. If you see 09001 through 09015, the file predates the 2022 vintage. If you see 09110 through 09190, it is current. If the file carries no codes, count the shapes. A stated national total can also be suggestive rather than conclusive: older material tends to follow the Census Bureau’s own phrasing of 3,143 counties and the District of Columbia, current material 3,144 county-equivalents. The Connecticut pages here are built on the nine regions, from the 2025 cartographic boundary files.

One habit is worth carrying beyond this case. When a geography changes, check the identifier as well as the name. COUNTYNS, the eight-digit GNIS identifier, survives renames in a way FIPS codes do not, and it is the safest field to join on when you suspect a boundary or naming change has happened underneath your data.