Chartered nonpublic enrollment by building, 1977–2025

Source. Ohio Department of Education and Workforce, Enrollment Data, under Frequently Requested Data. Twenty-three attachments on one page: eighteen annual files, one an October, and five compilations carrying the years before the annual series began. Type. Primary source — the department’s own October headcount of chartered nonpublic schools. Location. Enrollment Data, every file a getattachment URL under that path. The filenames are inconsistent to the point of being unusable as a pattern — nonpub.xls, nonpub2015.xls, Revised-FY16-nonpub.xls, web_nonpub_2011.xls, oct_08_fy09_adm_web_nonpub.xls — so the twenty-three are enumerated from the page’s anchors, not composed. #257 recorded that this page reaches 1978 and no connector owned it; this entry is the consequence.

What it contains. One row a chartered nonpublic school, or in the modern era one row a school and grade: IRN, school name, and a count of pupils by grade and sex. The historical files add county, school type and a parallel count by race. From October 2015 a separate sheet carries the sector’s true statewide total. Out-of-state pupils are counted separately throughout the modern era.

Why it matters. R.C. 3317.024(E)(2)(d) divides the appropriation for auxiliary services (R.C. 3317.06) and for nonpublic administrative cost reimbursement (R.C. 3317.062) by the membership in chartered nonpublic schools “as determined as of the last day of October of each school year”. This page is that determination. Before this connector the repository held the denominator for one year — 711 schools and 173,156 pupils, from the 2023-24 Education Landscape fact sheet — and every per-year question about the nonpublic aid lines was unanswerable. It is also the only long series of the sector at all, which is what the scholarship participation series has to be read against if participation is to mean anything.

Every count under ten is masked, and it is not a small effect. A cell printed <10 stands for some number from one to nine. Ohio’s nonpublic sector is mostly small schools, and the modern layout is one row a school and grade, so most cells in most schools are masked: in October 2023, 5,113 building-level cells are masked against a published in-state total of 171,830, and the unmasked cells sum to 146,078. The building rows bound that October at [146,078, 192,095] and no arithmetic over them recovers the published figure. Every quantity in the extract is therefore a floor, a ceiling and a count of censored cells. Where a publisher states a true total it is carried separately, in a column that says which publisher said it.

The two totals a reader will reach for are different quantities. This is the trap in the source and it is worth stating twice.

• The historical compilations’ State Total row is an Excel SUM over cells that had already been overwritten with the mask text, and SUM ignores text. It equals the sum of the unmasked cells exactly, on all twenty-eight sheets that carry one — the last three sheets of the 2000-2008 compilation carry no such row. It is a floor wearing the label of a total: read as Ohio’s nonpublic enrollment, 1977-78 comes out as 243,489 when the sector was between 260,867 and 266,736.
• The annual files’ state-totals sheet is a true pre-masking total, computed before the building rows were censored. October 2023’s says 171,830 where its own building rows floor at 146,078. From October 2015 every annual file carries one; before that, none does.

Two blocks measure the same pupils and they do not agree. The historical layout carries enrollment twice — across twenty-six grade-and-sex columns and across five race columns. Same population, so each bounds it, and the race block, being a fifth as wide, masks a fifth as many cells and bounds it far better: 1977-78 is [243,489, 266,736] on grades and [260,867, 270,209] on race. They are not merely differently precise. To 1982 a number of schools reported race and left every grade cell zero, so the grade block is absent for them rather than masked; from 1983-84 the two agree on every building with no masked cell in either. Both are carried side by side and neither is reconciled into the other. A caller wanting one number a building should prefer the race block where it exists.

nonpub_fy17.xls is October 2015’s building data under October 2016’s name. Its building sheet is Revised-FY16-nonpub.xls’s: 5,883 of 5,887 keyed cells are in both and not one of the shared cells differs. Matched against the 2014-2019 compilation, which states a per-school total for each of FY2014 through FY2019, the shared sheet agrees exactly with FY2016 on 86 unmasked schools and with no other year on more than six. So the rows are FY2016’s, the state-totals sheet beside them is genuinely FY2017’s, and October 2016 has no building-level data of its own. A pairwise sweep of the whole series found no other repost. Taking the sheet at its filename publishes 2015’s schools as 2016’s, silently.

Revised-FY16-nonpub.xls’s own state-totals sheet is separately wrong: it reads 144,512, below its own building rows’ floor of 147,309, so it cannot be a total of them. The compilation’s 171,395 is the figure for that October.

Two publishers describe six Octobers and disagree. The 2014-2019 compilation restates FY2014–FY2019, which the annual files already cover, and the two do not line up. October 2017 reconciles exactly — 167,558 in-state plus 1,533 out-of-state is the compilation’s 169,091. The others do not: October 2016’s annual total of 171,426 is the compilation’s total, not its in-state figure, and October 2018 reads 165,346 against the compilation’s 165,511, which is neither. The extract keys the sector panel on October and basis, so both publications get a row and the disagreement is data rather than a resolution taken in a reader.

The _ADM in the column names is not ADM. Every annual file’s notes sheet says the counts are a headcount taken during the first full week of October and are not equivalent to average daily membership — while the columns those counts sit in are named BOYS_ADM, MALE_ADM and so on. The notes sheet is right.

Layout eras. Six, all resolvable by column name:

1977-78 … 2006-07   one row a school; grades as Boys/Girls pairs, K through 12; race block
                    beside them. 2001-02 inserts a Total column and shifts the race block;
                    the sheets from 2002-03 are named "… ethnic data unavailable" and carry
                    no race block at all
2007-08 …           one row a school AND grade. IRN, ORG_NM, grade, male, female, and from
                    2008 an out-of-state pair. Grade codes are "Kindergarten"/"1st Grade"
                    to October 2013 and "KG"/"01" after
October 2009        building totals only: IRN, name, one number, 787 schools
October 2015 …      a state-totals sheet joins the annual file, two rows wide
FY2014 … FY2019     the compilation's By School sheet: FY, IRN, school, PS-12 total
FY2014 … FY2019     the compilation's State Totals sheet, two grade spans an October

The sex columns are BOYS_ADM/GIRLS_ADM to October 2018 and MALE_ADM/FEMALE_ADM after; three files spell the second one FAMALE. The preschool pair, where it exists, is school-level and repeated identically on every one of a school’s grade rows — summing it multiplies a school’s preschool count by its number of grades. No grade block in any era carries a preschool row, so the annual files are K-12 and the compilation’s per-school totals are PS-12.

Status. Wired. Forty-nine Octobers, 1977-78 through 2025-26, in crates/project/fixtures/nonpublic-building-panel.csv and crates/project/fixtures/nonpublic-sector-panel.csv. Both are built from one read of the twenty-three files, so a failure skips both.

What the extraction established, beyond the rows.

• The statutory denominator is the in-state count. The FY2025 auxiliary services appropriation net of the College Credit Plus earmark, $164,077,754, over October 2024’s 179,693 in-state pupils is $913.10, and LSC publishes $913. Including the 1,478 out-of-state pupils gives $906. The per-pupil figure for the nonpublic aid lines is reproducible from this series and was not reproducible from anything the repository held before.
• The single-year fact sheet reproduces exactly. October 2023’s state-totals sheet gives 86,232 male and 85,598 female in state, 742 and 584 out of it — 173,156 — and its building sheet carries 711 distinct IRNs. Both are what the 2023-24 Education Landscape states, from a different office and a different publication.
• 711 was never short of anything. The LSC redbook’s 747 schools and greenbook’s 725 read as a defect in the fact sheet’s 711 until the series exists; on the series they are simply later Octobers — 711 in October 2023, 722 in October 2024, 749 in October 2025. The redbook was written against a vintage the fact sheet predates.
• The sector’s arc is not monotone. It falls from a floor of 243,489 in 1977-78 to 176,811 in 2006-07, with a local peak of 218,542 in 1996-97 on the way, and then falls again to 161,664 in October 2020. It has risen in each of the three Octobers since universal EdChoice eligibility: 166,013, then 168,111, 171,830, 179,693 and 184,069.

Other caveats:

• The district a building sits in is not published anywhere in this source. Auxiliary services flows to a nonpublic school through the district its building is located in, so a per-district question about the nonpublic sector needs that column, and no file in any era carries it. County and school type exist for 1977-78 through 2006-07 and then stop. The Chartered Nonpublic School Information page publishes no directory, and edchoice-designated-2627.csv is public buildings only. The extract is keyed on building IRN alone for that reason, and the connector records the gap as still blocked rather than papering over it with a name match — district names are not unique is a settled finding in this repository and school names are worse.
• A building count is not a school count in the way a reader expects. The panel counts distinct IRNs in a file, and an IRN is a chartered building: a school operating two buildings appears twice.
• October 2009 contributes a bound and nothing else. It is published as one number a school, so that year has no grade detail, no sex split and no out-of-state count.
• October 2020 and October 2021 are pandemic Octobers. The count moved for reasons the file does not record, and neither year has a second publisher to check it against.
• The out-of-state columns begin in October 2008 and are masked like everything else. They are small — 1,300 to 1,700 a year — and the masking is proportionally much heavier: October 2023’s out-of-state floor is under a tenth of its published 1,326.

Also read by #

• crates/connect/src/fixtures/nonpublic.rs — the two panels and the masking arithmetic over them.
• Auxiliary services, whose per-pupil figure and sector denominator are this series.