United States · 2023
County measures, once you allow for city versus countryside
This site’s country data has one big trap. Nearly every measure moves with how rich a country is, so most links between measures just rediscover that rich countries differ from poor ones. County data has the same trap under a different name. Here, population density (how crowded a place is) plays the part of wealth. A link between two county measures, before adjusting for anything, usually just rediscovers that cities differ from the countryside.
So every pair here is adjusted. We compare places with the same population density, median household income and share of adults with a bachelor’s degree, so those differences can’t explain the link. Then we do it again comparing counties only with other counties in the same state, and again giving bigger places more weight. And each place is compared with the fifty counties most like it, not with the whole nation.
Counties3,222plus 52 states and DC
Measures205plus 69 published for states only
Pairs tested20,91016,080 clean enough to rank
Still clustered on the map20,794of 20,910 pairs; middle Moran’s I +0.35
Most county data in America is estimated by models, not measured
No survey is big enough to measure diabetes rates or median income in a county as small as Loving County, Texas. So federal agencies build statistical models. They use national survey data plus known facts about each county to predict the county’s figure. 50 of the 205 measures here are predictions of that kind. CDC PLACES says so in its own notes: its inputs are BRFSS health survey interviews, Census population estimates and American Community Survey data. SAIPE mixes the American Community Survey with tax and benefit-programme records.
That creates a problem you cannot see in the numbers. 2,312 of the 20,910 pairs here put a prediction next to one of the inputs its own model used. Another 1,098 put two predictions side by side that were built from the same county facts. Both kinds would match to some degree even if the real-world things did not. Only 16,080 pairs — 77% — are free of that tangle and of the count and arithmetic problems below. Those are the ones ranked here. The rest can still be explored, with a warning attached.
Four ways county data differs from country data
Counts mostly measure size
Here a link is measured by r: how closely two measures rise and fall together, from −1 (opposite) to +1 (in step); 0 means no link. All 1,128 pairs of raw counts have a middle link of +0.97 even after adjusting. Adding population to the adjustments changes that by only 0.003. Every count follows density with an r between +0.69 and +0.89. Every rate and median falls between +0.01 and +0.74. The two ranges do not overlap. Count pairs are marked and kept out of the ranking.
Giving big places more weight shrinks the sample
People are packed into so few counties that, once bigger places get more weight, a county link behaves as if it rested on about 266 counties, not 3,222. Both versions are shown for every pair. 3,929 pairs point in opposite directions in the two versions. In 45 of those, both directions are unlikely to be chance, even allowing for counties in the same state resembling each other. So when the average county and the average person disagree, it is usually that the link fades away, not that it truly flips. But not always.
Neighbours look alike
20,794 of 20,910pairs leave leftovers that still cluster on the map. The leftovers are the part the adjustments do not explain. This is measured with Moran’s I, which shows how much nearby places look alike, across each county’s 8 nearest neighbours. A p-value is the chance of seeing a link this strong by luck alone. One that treats 3,144 counties as 3,144 separate, unrelated cases is too hopeful. So every pair also gets a p-value that allows for counties in the same state resembling each other.
The unit changes the answer
The same pair measured across counties, metro areas and states can differ in strength, and sometimes in direction. Statisticians call this the modifiable areal unit problem. Each pair shows its own results for each level. Where a level is missing, the measure cannot be combined up to it: you cannot average county medians (middle values) into a metro median.
Strongest links that remain after adjusting
Ranked by the size of the link after adjusting. Only pairs with a q-value of 0.05 or less are shown. That means they are unlikely to be chance, even allowing for the 20910 tests run at once. Some pairs are left out: pairs built from the same numbers, pairs tangled together through a model, pairs measuring the same thing at two agencies, and pairs of raw counts. That is 346 built from the same numbers, 3,410 linked through models, 44 measuring the same thing twice, and 1,128 count pairs.
| Pair | raw r | adjusted r | adjusted, same state | weighted | state by state |
|---|---|---|---|---|---|
| Median age × Share aged 65 and overWhere Median age is higher, Share aged 65 and over tends to be higher: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Median age by US county · Share aged 65 and over by US county | +0.92 | +0.93 | +0.93 | +0.94 | +0.92 |
| Self-care disability × Material security scoreWhere Self-care disability is higher, Material security score tends to be lower: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Self-care disability by US county · Material security score by US county | -0.92 | -0.88 | -0.84 | -0.87 | -0.90 |
| Stroke × Material security scoreWhere Stroke is higher, Material security score tends to be lower: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Stroke by US county · Material security score by US county | -0.91 | -0.86 | -0.84 | -0.76 | -0.90 |
| Food insecurity × Returns claiming the earned income creditWhere Food insecurity is higher, Returns claiming the earned income credit tends to be higher: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Food insecurity by US county · Returns claiming the earned income credit by US county | +0.91 | +0.84 | +0.77 | +0.80 | +0.88 |
| Lack of reliable transportation × Returns claiming the earned income creditWhere Lack of reliable transportation is higher, Returns claiming the earned income credit tends to be higher: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Lack of reliable transportation by US county · Returns claiming the earned income credit by US county | +0.90 | +0.84 | +0.78 | +0.81 | +0.88 |
| Vision disability × Material security scoreWhere Vision disability is higher, Material security score tends to be lower: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Vision disability by US county · Material security score by US county | -0.89 | -0.83 | -0.80 | -0.77 | -0.89 |
| Housing insecurity × Returns claiming the earned income creditWhere Housing insecurity is higher, Returns claiming the earned income credit tends to be higher: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Housing insecurity by US county · Returns claiming the earned income credit by US county | +0.91 | +0.83 | +0.76 | +0.76 | +0.87 |
| Received food stamps × Material security scoreWhere Received food stamps is higher, Material security score tends to be lower: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Received food stamps by US county · Material security score by US county | -0.89 | -0.82 | -0.90 | -0.81 | -0.92 |
| Self-care disability × Returns claiming the earned income creditWhere Self-care disability is higher, Returns claiming the earned income credit tends to be higher: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Self-care disability by US county · Returns claiming the earned income credit by US county | +0.91 | +0.82 | +0.75 | +0.80 | +0.87 |
| Vision disability × Returns claiming the earned income creditWhere Vision disability is higher, Returns claiming the earned income credit tends to be higher: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Vision disability by US county · Returns claiming the earned income credit by US county | +0.88 | +0.81 | +0.75 | +0.82 | +0.87 |
| Fair or poor self-rated health × Material security scoreWhere Fair or poor self-rated health is higher, Material security score tends to be lower: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Fair or poor self-rated health by US county · Material security score by US county | -0.86 | -0.80 | -0.80 | -0.75 | -0.87 |
| Independent-living disability × Material security scoreWhere Independent-living disability is higher, Material security score tends to be lower: a strong link that holds when only places with similar density, income and education are compared, and inside states as well. Independent-living disability by US county · Material security score by US county | -0.88 | -0.80 | -0.81 | -0.77 | -0.86 |
How the US county data works · Every pair worth a page · Explore every pair · US county correlation matrix · US county topic scores compared · Every measure, and how urban each one is · What the measures describe, and how they change · Local income taxes · Do people leave after an income tax rise? · Where the rich move · Top tax rates since 1995 · US Murder (Homicide) Rate: What Drives It? · All 3,222 places
Reading a place page
Ranking counties nationally on almost any measure gives the same list: wealthy suburbs at one end, the rural south at the other. So each place page scores its measures twice: once against all counties, and once against the fifty counties closest to it in density and median household income. The gap between the two is the part worth reading. PlaceStats shows links only. Nothing on these pages shows that one measure changes another.