Half of America's county childcare prices are not county prices
The Labor Department publishes childcare prices for 2,995 counties. In 1,322 of them the number is identical, to the cent and across every year, to another county in the same state.
Cite as: FACTANKER, https://factanker.com/record/county-childcare-prices-that-are-not-county-prices
The National Database of Childcare Prices is the closest thing the United States has to a local price index for childcare. It is published by the Department of Labor, it reaches down to the county, and it is used to argue about affordability in places where no other number exists. The latest release covers 2008 to 2022.
We loaded all of it: 281,510 prices across 2,995 counties. Then we compared each county's price series with every other county in the same state. Of the 2,532 counties with at least eight years on record, 1,322 — 52.2 % — have a series that is identical to another county's. Not similar. Identical, to the cent, in every year.
In Georgia, 99 counties share one straight line
The largest single group is in Georgia. Ninety-nine of its counties carry the same weekly price for infant care at a licensed centre, year for year:
| Consecutive years | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|
| Weekly price | $92.75 | $95.00 | $97.25 | $99.50 | $101.75 | $104.00 | $106.25 |
The step is $2.25. Every year, in all ninety-nine counties, for as long as the series runs. Across Georgia as a whole, 99 % of its 159 counties share their series with another county. Arkansas is at 99 %, South Carolina at 98 %, Louisiana and Washington at 95 %, Texas at 95 % of 254 counties.
| State | Counties with a price series | Share whose series is shared with another county |
|---|---|---|
| Georgia | 159 | 99 % |
| Arkansas | 68 | 99 % |
| South Carolina | 46 | 98 % |
| Louisiana | 64 | 95 % |
| Texas | 254 | 95 % |
| Washington | 39 | 95 % |
| Kansas | 105 | 94 % |
| Alabama | 67 | 94 % |
| Tennessee | 95 | 92 % |
| Kentucky | 120 | 90 % |
This is not a leak. The publisher says so.
The department's own technical report, dated September 2024, describes the method plainly: county values are estimated from state figures with a regression on census characteristics of the county, weighted by households. It reports a fit of R² = 0.84 for the 50th-percentile model. Prices reported by providers as hourly, daily, monthly or yearly are first standardised into a single weekly figure.
So the method is documented. What is not documented is which rows it produced. Each price carries a quality flag with the value 1, 2 or 3 — and the technical report does not say what the three levels mean. In this release almost every price has one: 169 of 281,510 do not, and for school-age care the source publishes no flag column at all.
The series is smoother than a market
The shared series are not the only artefact. Across the whole infant-centre dataset, 14,471 of 32,449 year-over-year steps — 45 % — are exact arithmetic continuations of the step before: the same difference, repeated. That is the signature of filling in the years between state surveys, not of measuring them.
Where it shows most plainly is where the line runs the wrong way. In 82 counties the nominal weekly price falls over more than a decade. Owyhee County, Idaho drops from $228.60 to $120.38 a week between 2012 and 2022 — 47 % down, in dollars that were not adjusted for inflation, through a period in which American childcare became a national affordability debate.
Two more Idaho counties, Blaine and Teton, fall from $272.15 to $208.00 — the same two numbers in the same years, in two separate counties. Rawlins and Franklin County, Kansas run from $134.40 to $95.60, also identically, although they sit at opposite ends of the state.
What the data still tells you
This is not an argument that the database is worthless. At state level it is informative, and where a county genuinely was surveyed the figure is the best local number that exists. The 2022 spread it shows is real and large: San Francisco County at $606.61 a week — $31,544 a year at 52 weeks — against counties in the low hundreds.
What it cannot support is the comparison it most invites. Two counties in the same state, held up against each other, may be two copies of one number. Before using a county figure as a local fact, check whether its neighbours carry the same one.
Our own pages carry this data
FACTANKER publishes these prices, county by county. Since 27 September every one of those values is labelled estimated rather than observed, shows the weekly figure as the source published it beside the annual figure we calculate from it, and states in its first sentence that it is a modelled county estimate. The method flag is passed through unchanged, because we do not know what it means either.
That labelling is the point. A number that was modelled can still be the best available; a number that was modelled and presented as measured is a different thing entirely.
How this was measured
Source: NDCP2022.xlsx, U.S. Department of Labor, Women's Bureau, retrieved 27 September 2026, SHA-256 6722a364…. Series compared by concatenating each county's weekly prices in chronological order and grouping identical strings within a state; counties with fewer than eight years excluded. Arithmetic continuation means the second difference of the weekly price is below one cent. Every figure above is reproducible from the full result table, which carries the weekly source value, our annual derivation, the method flag and the fact ID for each row.