Census geography connector

The site can group districts by county because the department attributes each district to exactly one, and the corpus inherits that attribution. Legislative districts have no such attribution and cannot: 339 of Ohio’s 609 school districts have population in two or more of the 99 House districts, and Columbus has population in eleven. Nothing in Ohio’s funding system uses a legislative district as a unit of account, so nobody publishes the mapping.

It is nevertheless derivable, entirely from files other people publish:

Census 2020 block assignment block -> unified school district, 276,428 Ohio blocks Census 2024 SLD block equivalency block -> House district, current map Census 2020 PL 94-171 block -> population, and population 18 and over NCES CCD LEA directory Census school district code -> Ohio IRN

The vintage of the second file is the whole reason it is a separate source. The block assignment archive ships its own SLDL file, which is the obvious thing to reach for and is the 2020-cycle map. 66.3% of Ohio’s blocks changed House district between that map and the one now in use. A page built on the convenient file would have been wrong for two-thirds of the state while looking entirely plausible.

The fourth file closes something older. census-f33 has carried “the per-district join needs an NCESID-to-IRN crosswalk this repository does not hold” since it was wired, and nces-ccd has been Declared since genesis. The CCD directory’s ST_LEAID column is the IRN behind an OH- prefix, and all 609 panel districts join through it with no losses. The crosswalk was one column in a file the corpus had already named.

Declare census-geography as a thirteenth connector, status Wired, feeding education-agency and actor, with three sources: the 2020 block assignment file, the 2024 state legislative block equivalency file, and the 2020 PL 94-171 redistricting file.

Move nces-ccd from Declared to Wired and give it the LEA directory as a source. Its recorded blocker was about deriving an identifier-change series across decades, which remains unbuilt; a single year of the directory was never blocked and was never tried.

Apportion by under-18 population rather than total population. Total population would weight a House district of retirees the same as one of young families, and the quantity being divided is school funding. Under-18 is still a proxy — it counts children in community schools, private schools and none — and the page says so.

The site gains 99 House district pages carrying figures no one else publishes, including the department. That is a different kind of number from everything else here, which is either the department’s own or a reproduction of it, and it is labeled as such on every page that shows one.

One property is exact and is the only accuracy claim made: each school district’s shares sum to one, so the 99 House districts sum to the statewide total to the cent. Asserted in project::legislative_district and again over the rendered page.

Two staleness risks the corpus now owns. The apportionment uses 2020 population against a FY2027 model, so a district that has grown or emptied since the census is weighted as it was. And Ohio redistricts; the crosswalk is pinned to the 2024 map and will need regenerating when that changes, which the fixture’s header assertion will not catch because the header will not change.

The NCESID-to-IRN crosswalk being solved also unblocks the per-district F-33 join, which census-f33 records as its own remaining gap. That is not done here and is now a retrieval task rather than an identifier one.

The Senate followed, and the decision extends to it unchanged. sldu24.zip is a fourth source under the same connector, and the crosswalk now carries a chamber column: 1,085 House rows and 862 Senate rows over the same 609 districts.

One thing the Senate added that the House could not. Ohio’s constitution requires each Senate district to be exactly three whole House districts, and the block data confirms it with no exceptions — so the Senate crosswalk was derivable from the House one, exactly rather than approximately. It is read from the Bureau’s own file anyway, on two grounds: the composition is not sequential, so a derivation needs the mapping regardless, and a rule worth relying on is worth checking against the file that would break it if a future redistricting did.

The Senate view is also less of an estimate, which the alternatives above did not anticipate. Seats three times larger mean 392 of 609 school districts lie wholly inside one against 270 for the House. That does not change the decision to apportion, but it does mean the two chambers’ pages carry different amounts of derivation and each says which.

Assign each school district to one House district. Rejected: it is what the county pages do and it works there because the department’s attribution is total. Here it would state something false about 339 of 609 districts, and the failure would be invisible — a page saying “Columbus City is in House District 26” looks exactly like a page saying something true.

Publish membership without apportioned totals. Considered seriously, and it is the honest minimum: list which districts overlap a seat and stop. Rejected because a member asking what a funding change does to their district needs a magnitude, and refusing to estimate one does not stop the question being asked — it moves the estimate somewhere with no methodology attached. The apportionment is published with its weights, its vintage and its caveats stated.

Weight by enrolled ADM instead of population. Rejected: ADM is known per district, which is the thing being split, not per block. Any block-level pupil count would itself be an apportionment, so this would add a step without adding information.