Reference

What this site does not cover

Every gap below is already stated on the profiles it affects. None of them was possible to see in aggregate, which meant a reader deciding whether to trust the site had to assemble the picture one page at a time. This is that picture.

The figures here are counted when the page loads rather than written into it. That is not fussiness: two numbers elsewhere on this site had drifted by a factor of four before anyone checked them, and a page about the limits of the data is the worst place for a stale number.

School results are from before the pandemic, and 1 state publishes none at all

Proficiency figures are from the 2018–19 school year, the last full year of state testing before the pandemic. The school directory beside them — enrollment, teacher counts, which schools exist — is current, from 2024–25. The two are labeled separately wherever they appear, and the Schools signal combines them: four fifths of it is how students tested then, one fifth is how the schools are staffed now.

1 state or territory reports no proficiency data to the federal collection at all: VT. Nothing in a schools ranking can include it, which is why the best-schools ranking contains no city from that state. That is an absence in the federal file, not a judgment about those schools.

The reported figures also rest on a narrow base: a school with no tested grades has none, and any school whose published proficiency range is wider than twenty percentage points, or with fewer than ten students tested, is dropped rather than guessed at. The guide works through what that leaves a reader able to say.

Puerto Rico has 85 profiles and 81 Area Scores

Several federal collections stop at the fifty states. The Population Estimates Program, the CDC mortality files, the Bureau of Economic Analysis price parities, the Federal Housing Finance Agency house price index and the IRS migration files all publish nothing for Puerto Rico. FEMA hazard data does reach it, so do the Department of Education’s 2018–19 school assessments, and the American Community Survey covers it fully. The Census Building Permits Survey publishes Puerto Rico separately from the fifty states, in its own Island Areas release rather than the standard place and county files — easy to miss, and missed here for a while, but real once found.

Some measures are absent; each profile names which and why.

3,029 profiles have no air-quality reading

The EPA publishes daily air quality from physical monitors, and most counties do not have one. Of 9,533 published profiles, 6,504 carry an air-quality figure and 3,029 do not. Where a monitor exists the reading is measured rather than modeled, which is the reason it is worth having and the reason it cannot be extended to places without one.

2,449 cities show their county’s unemployment rate

The Bureau of Labor Statistics publishes city-level unemployment only for places of 25,000 or more. AreaProfile publishes cities from about ten thousand, so 2,449 of them show the rate for the county containing them, labeled as such on the profile. It is the honest figure available rather than an estimate of the city’s own rate — but it means two neighboring towns in one county show an identical unemployment rate, and that unemployment feeds the Opportunity signal and therefore the Area Score.

Home price change is never a town’s own

The Federal Housing Finance Agency publishes its house price index for states, metropolitan areas and, for the largest metros, metropolitan divisions — never for a single town. Of 4,226 published cities, 2,362 show their metropolitan area’s figure, 1,295 show their division’s (New York, Los Angeles, Chicago, Dallas, Atlanta, Miami and other large metros are published only that way), and 496 outside every metropolitan area show their state’s non-metropolitan figure. The remaining 73 show none: Puerto Rico, which the metropolitan series do not reach, and places the agency has no series for or that straddle two divisions. The profile names which area a figure belongs to, and the index covers mortgages bought or guaranteed by Fannie Mae and Freddie Mac, so it says little about cash purchases.

4,101 cities show their county’s migration figures

The Internal Revenue Service publishes where tax filers moved, by county. A city shows the figures for the county that contains it, labeled as such, so two cities in one county show the same net migration and the same incomes of arrivals and departures. The data counts people who file tax returns and sees a move only when the address on a return changes, so it misses people who do not file, and it leaves out moves from abroad on both sides.

Reported crime is not from the same year everywhere

Reporting to the FBI is voluntary, and a department that reports one year may not report the next. Published profiles currently carry crime figures from 4 different periods: 2025 calendar year (2,887), 2024 calendar year (37), 2022 calendar year (23) and 2023 calendar year (13). Two cities compared side by side can therefore be showing years that are some way apart, and a city missing entirely is a department that did not report rather than a place without crime.

Survey figures carry a margin, and for small places it is wide

The American Community Survey samples rather than counts, and publishes a margin of error with every estimate. AreaProfile stores it and shows it beside the figure. For a city the margin on median household income is typically about a tenth of the estimate; for a ZIP area of a few thousand people it can be a quarter of it.

That is why a comparison page will sometimes decline to name a winner, and why a ranking will say that its own order is not separable. It is also the limit of what the site can check: the ACS is the only source here that publishes a margin at all. Figures from FEMA, the FBI, the CDC, NOAA, the FAA and the Department of Education arrive with no stated uncertainty, and some of them — school proficiency on a small cohort, a crime rate in a small town — are noisier than the survey figures that do carry one.

Where a single figure is missing from a single profile, that profile says so and says why. Data sources carries each agency’s coverage, cadence and terms; the glossary defines every measure; the methodology covers how they are combined. If something here looks wrong, tell us — corrections are published.