Census geographies
The level selector on the city maps presents its five options as a ladder from coarse to fine. That holds for three of them (counties, tracts and block groups), which nest inside one another, and not for the other two (PUMAs and ZCTAs), which cross the lines above them. The difference decides what a map at that level can claim.
The standard hierarchy
The Census Bureau publishes this as a diagram, and its shape is the argument: a spine of levels that strictly nest, with every other geography branching off the tier it attaches to. Where a branch leaves the spine tells you what it can be aggregated to. School districts and legislative districts leave from the state. PUMAs leave from tracts, because they are built from whole ones. ZIP Code Tabulation Areas and urban areas leave from the very bottom. They are assembled from blocks and nest inside nothing above them.
The spine itself is the reason the resolution argument on these maps works. A block group is part of a specific tract, inside a specific county. Aggregate every block group in a tract and you get the tract exactly.
It is also why suppression rises as you descend. A block group holds a few hundred households, and the five-year survey interviews a few dozen of them, against a hundred or more in a tract, so its margin of error widens until the Census Bureau withholds it. Resolution is bought with reliability. Marked levels are the ones this site draws; the rest are here because a reference that omits them makes the mapped ones look like the whole system.
What does not nest
ZIP Code Tabulation Areas and PUMAs sit on the same selector and behave differently. A ZCTA approximates a USPS delivery route, and mail routes were drawn for carriers rather than for statistics, so they cross county lines and state lines and change shape between vintages. A PUMA is built from whole tracts and holds at least 100,000 people, which in practice keeps it inside one county in a dense metro but guarantees nothing.
The consequence is practical rather than theoretical. Selecting the ZCTAs “in” a city by ZIP prefix, the obvious approach and the one this site used first, kept 58 Orange County ZCTAs for Los Angeles and missed 32 Cook County suburbs for Chicago. Both are now selected by testing their geometry against the counties being mapped.
School districts are three geographies
“School district” is not a census geography. It is three: unified districts that run all grades, elementary districts, and secondary districts. The last two overlap each other, covering the same ground for different ages.
Illinois is where this bites. It has 386 unified districts, 373 elementary and 106 secondary, verified against the 2023 ACS. A choropleth of Cook County school districts therefore forces an editorial choice before a single polygon is drawn: unified only, which leaves holes; unified plus elementary, which is coherent for K–8; or unified plus secondary, coherent for 9–12. No option is the map, and a map that does not say which it chose is not answering the question.
New York City is the opposite problem. The Department of Education covers all five boroughs as a single unified district, so a school-district map of the flagship city on this site would be one polygon. Its 32 community school districts are administrative subdivisions rather than census geography and carry no ACS data at all.
Area counts: 2020 census geography; the 51 state-level entities include DC. Counties use the 2023 vintage, because Connecticut replaced its eight counties with nine planning regions in 2022. School district counts: ACS 2023 5-year, Illinois, queried per summary level.