Sizing the five units beside the district

The method page said, in the row headed Community schools and scholarships, that “the deduction that routes money from a resident district to a community school or a scholarship is not modeled here”. The corpus has since read R.C. 3317.022 and established that there is no such deduction to model: the community and STEM school unit and the four scholarship units are funding units beside the district unit, computed and paid directly. The row described a mechanism that does not exist and said nothing about the one that does — the whole of what this site had to say about private schooling, and it was about the wrong regime. Issue #474 is the correction.

Correcting the sentence meant sizing the channel, and sizing it raised two questions that a sentence would have answered by accident.

The community and STEM school unit has two candidate publishers and they are not two views of one number. The department publishes a community and STEM school simulator — the same office, the same formula, the release this repository already reads for the FY2027 traditional calculator — and the Census Bureau’s Annual Survey of School System Finances reports what those schools received. One is the paying side of a model, the other the receiving side of a survey, and f33-has-two-publishers already records that the panel changes hands inside its own window.

That change of hands settles it. Of the 357 non-comparable Ohio agencies the panel carries at FY2022, from NCES’s file, 298 are schools in the department’s FY2027 model. At FY2023 the panel switches to the Bureau’s own elsecNN.xlsx and carries 46, of which none is in that model. The receiving side does not merely disagree with the paying side at the ends of the window; at the current end it holds no community school at all, and a figure read off it would be reporting the disappearance of a file rather than of a payment.

The six units have no year in common, and four of them cannot be moved onto one. The district unit is the FY2027 department model, which is what every other figure on this site is computed from. The community and STEM unit has a model for FY2027 and for FY2025 and — per a-supplement-paid-to-a-population-the-panel-does-not-hold — will never have one for FY2026, because the department publishes one such workbook at a time and replaces it in place. The four scholarship units are sized by the 2025 Scholarship Annual Report, which covers the school year 2024-25 and is the only committed source that sizes them from the administrator rather than from statute. LSC’s redbook does publish all four on an FY2027 basis — Table 5, $1,251.1m — but those are estimates in a budget analysis, and putting an estimate beside a model to buy a shared year label trades a stated year for an unstated basis.

The community and STEM school unit is sized from the department’s own model, at the vintage the district unit is shown at. dispersion::community_school_funding::total_state_support on Vintage::Fy2027: $1,511,789,771.13 across 355 schools and 126,714.26 enrolled ADM. Same publisher, same statutory formula, same fiscal year and the same kind of claim as the district figure it sits beside — a model’s output, not a survey’s observation.

Every unit carries its own fiscal year and its own basis, and the page renders both. The feed’s funding_units block gives each unit a fiscal_year, a basis — model or report — and the source it was read from, and series_years gains a row per basis so the chips on the new cards are derived like every other year label on this site. That is “Every figure says its year” applied to a table whose rows are genuinely on different years: its rule is that where rows are on more than one reckoning, every row says which, and half-labeling is worse than none.

The scholarship channel is published as two figures, not one. $991,191,150.12 is what the four programs that publish an expenditure paid; $103,944,388.15 is Jon Peterson’s, which the report displays as six disability-category figures and never totals. The feed carries the second with derived: true beside it and the page says so. The sum, $1,095,135,538.27 over 166,587 students, is the quotable number and it is the one that is only partly quoted — the distinction the_scholarship_channel_by_the_numbers already holds in the crates, carried through to the page rather than flattened at the feed boundary.

The derived figure lives in project::scholarship::report as a named constant, not in scholarship-programs.csv. That fixture’s expenditure column is quoted figures and the blank cell is the claim that the report publishes none; jon_peterson_publishes_participation_but_not_a_total exists to keep it blank. A constant beside the reader, documented as derived and bound as a figure, puts the number where the feed can reach it without changing what the extract asserts.

Category 3 nonpublic support travels beside the six units and is not one of them. Auxiliary Services, the Nonpublic Administrative Cost Reimbursement and the reimbursement fund move under R.C. 3317.024, 3317.06, 3317.062, 3317.063 and 3317.064 — not under R.C. 3317.022 — so the feed puts them in their own block with the sections that carry them. A reader who meets the scholarships and not these has met about three-quarters of what the state spends on nonpublic schooling and been given no sign of it.

The site can state the mechanism in force for the first time. The method row now says what R.C. 3317.022 does — six units, paid directly — instead of denying a model of a deduction that was abolished. The deduct-era mechanism is a separate and still-open question, and saying so is cheaper than the sentence that was there.

The statewide page shows a total it previously implied was the whole of state support. realized_aid_total across 609 districts is the district unit alone. Beside the community and STEM unit and the four scholarship units it is one of six, and a reader who took it for “what Ohio spends on schools” was reading a figure this feed can now contradict — the same failure casino was added to the feed to stop.

Four of the five new figures are on a year no lever can move. The scenario machinery prices eight levers over 609 districts and these populations intersect that panel in nothing, which is the standing reason the equity supplement is not a lever. Nothing here changes that, and the method page records the absence of a scholarship lever as a limit rather than working around it.

f33-has-two-publishers gains a second consequence. It was recorded as a break in Ohio’s own state column at FY2016 and as a coverage difference at the panel’s far end; it is also the reason the receiving side cannot answer a question about community schools at all after FY2022. The population, not the column, is what moved.

Size the community and STEM unit from the Census receiving side. Rejected on the file. At FY2022 it would have given $1,348.58m over 134,885 pupils against the model’s $1,511.79m over 126,714.26 — a gap that is part year, part basis and part population, and decomposable into none of the three. At FY2023 and FY2024 it gives nothing: the Bureau’s Ohio file carries 46 non-comparable agencies and not one of them is a school in the department’s model. A figure that exists for one year of a fifteen-year panel is not a series, and a reader who saw it fall to zero would be watching a publisher change.

Put all six units on FY2027 using LSC’s Table 5. Rejected on basis. Those four figures are a budget analysis’s projections, marked estimated in scholarship-bounds.tsv for exactly this reason, and the district figure beside them is a department model. Matching the year label would have cost the thing the label is for: a reader could no longer tell which numbers are what somebody paid and which are what somebody expects to pay. The bounds are held and are reachable; they are not what sizes a unit.

Fill Jon Peterson’s expenditure into scholarship-programs.csv and publish one channel figure. Rejected, and the repository had already rejected it once in a test comment: “a derived total in a column of quoted ones is the kind of figure that gets quoted back as published”. Filling the cell would also have made project/nonpublic-support-against-the-scholarship-channel silently stop being the upper bound its label says it is — 24.48% would have become 22.16% with nothing in the corpus saying the denominator had grown.

A scenario lever for scholarship amounts. Rejected and recorded on the method page rather than in code. Scholarships are not per district, the lever machinery is, and the expansion’s award is a continuous function of family income under R.C. 3310.08 — so the distribution a lever would have to move is one this corpus does not hold. See “A supplement paid to a population the panel does not hold”, which settled the same question for the community school equity supplement on the same ground.