What foundation funding means

This repository publishes a foundation aid figure in two places, and until now only one of them had been reasoned about.

project::appropriations sums line items from the committed enacted series, FY1999 to FY2027. Its FOUNDATION_LINES is 200501, 200550, 200612 — the GRF and Lottery halves of the formula’s own line, plus the pre-FY2006 predecessor carried at $0.00 through its overlap. AppropriationYear::foundation_funding is built from it, reaches the feed, and the site renders foundation_funding / enacted as a percentage.

project::budget_analysis reads LSC’s published Foundation Aid Appropriations table directly, whose bottom line Total foundation aid is $11,230,057,557 for FY2026 as enacted. That table has five lines, and two corpus nodes quote LSC’s description of them [verified]: GRF 200550, GRF 200502 Pupil Transportation, Fund 5VS0 200604, Fund 5YO0 200491, and Lottery 200612, which “are used to collectively support state foundation aid payments”.

So the corpus said five and the crate summed two, and nothing reconciled them.

The obvious reconciliation does not work. Summing those five line items out of the committed series gives $11,547,417,388 for FY2026 — $317m above LSC’s own total, not equal to it. LSC splits 200502 into Earmarks and Remainder – Foundation aid and counts only the foundation half; budget_analysis::PUPIL_TRANSPORTATION already carries that split and says why. “Adopt LSC’s five lines” was therefore never available as a line-number sum. It would have produced a third number agreeing with neither.

One of the three exclusions is different from the other two, and is not definitional. ALI 200604 is titled Foundation Funding - All Students from FY2022, in the same fund and over the same span as the two included lines, at $500m to $600m a year — 5.9% to 7.2% of what was published. 200502 and 200491 are a real question about what the word means. 200604 is not: it is foundation funding under that name in the enacting act.

The cause is legible and is the part worth recording. FOUNDATION_LINES’ own doc comment cites 200604 as its cautionary example — “an appropriation line item is not a stable identifier across this period — 200604 names three different programs across three funds — so any series built by line number needs its succession established before it means anything.” That is right. But the succession was established for 200501 and used to exclude 200604 rather than to resolve it, and its third program is the one carrying the foundation-funding title.

Two numbers, two spans, both published, each naming the other.

budget_analysis::FOUNDATION_AID is the authority wherever LSC prints a table. It is read rather than summed, it includes the pupil-transportation earmark split that a line sum cannot reproduce, and it is what a reader comparing against LSC should be given.

appropriations::foundation_history is the only thing that reaches FY1999, because no such table exists for most of that span. It is summed by line and it is not LSC’s number. It is not an approximation of LSC’s number either — the difference is structural, not error.

Resolve 200604’s succession rather than excluding it, and resolve it by title. A line counts when it is titled Foundation Funding - All Students. The title is the evidence the succession is established from; a year is the same fact with the evidence discarded. It also gets the boundary right without being told: H.B. 110 retitles the line, so the FY2021 actual it restates reads the new title while the FY2021 enacted row — H.B. 166’s, which is what this series is built from — still reads Student Wellness and Success. A year-based rule set to FY2021 would have been wrong by $400m and would have looked correct.

Leave 200502 and 200491 out, as a decision. Pupil transportation and nonpublic education support are paid alongside the formula, and a series that includes them answers “what does Ohio appropriate that LSC groups with foundation aid” rather than “what does the formula cost”. Where the first question is wanted, LSC’s own total already answers it and is already computed.

Carry the widening in the type, not in a footnote. Basis::AsEnacted is what the act appropriated under that name; Basis::FormulaLines is the comparable subset. This follows dispersion::survey_basis::Basis, which exists for the same shape — a published column whose meaning changes partway along — and which resolves it by giving the comparable quantity its own basis rather than by dropping the year the change happens.

The published series rises in six years and in no others.

FY     was              now              share of enacted
2022   $8,202,198,712   $8,702,198,712   65.8% -> 69.8%
2023   $8,322,348,712   $8,922,348,712   65.8% -> 70.5%
2024   $9,242,195,000   $9,842,195,000   58.9% -> 62.7%
2025   $9,593,442,000  $10,193,442,000   66.9% -> 71.1%
2026   $9,894,181,974  $10,494,181,974   66.5% -> 70.6%
2027  $10,131,392,875  $10,731,392,875   66.4% -> 70.3%

FY2002 through FY2021 are untouched, which is the check that the title rule caught the program and not the line.

No bound figure moved, and that is the design rather than luck. Every summary statistic this workspace publishes off the series — the noise floor in what_the_series_cannot_settle, the mean and spread of annual movements, the FY2012 and FY2013 readings the casino null result rests on — differences two years, so all of them run on FormulaLines and none of them changed. crates/figures.json is byte-identical across this change.

Had they run on AsEnacted instead, FY2022’s movement would have been $677.7m where $500.0m of it is the channel arriving, and the median of exactly those magnitudes is the casino argument’s detection floor. The argument would have been weakened by a definition change.

The corpus’s five-line quotation stops being in tension with the crate. It was [verified] and correct; what was missing was that LSC’s five is a description of LSC’s table, which the repository computes elsewhere and cannot reproduce by addition.

Adopt LSC’s five as a line sum. Rejected because it does not work: $11,547,417,388 against LSC’s $11,230,057,557 for FY2026. The gap is the pupil-transportation earmark split, so the only way to get LSC’s number is to read LSC’s table — which budget_analysis already does. Adopting the five would have replaced a number that is honestly not LSC’s with one that looks like LSC’s and is off by $317m, which is the worse of the two failures.

Keep the narrow set and document the exclusion. The issue’s own “not attempted” position, and defensible for 200502 and 200491. Rejected for 200604: excluding a line that the enacting act titles Foundation Funding - All Students, in the same fund and span as the lines that are included, is not a definition of foundation funding that can be stated without embarrassment.

Include 200604 from FY2021. Rejected, and it is the trap this decision exists to mark. The line reads Foundation Funding - All Students in FY2021 in the actual column, because H.B. 110 restates the prior year under the new title. The enacted row for FY2021 is H.B. 166’s and says Student Wellness and Success, which is what it was: money paid alongside a frozen foundation amount rather than through it. A year cutoff would have been wrong by $400m in the direction that looks right.

One series, switching to LSC’s table where it exists. Considered. Rejected because it makes the published series change definition partway along its own length with nothing in the type to say so — the exact defect survey_basis was written to correct, reintroduced deliberately.

Make foundation_movements take a basis. Rejected. Offering a basis implies both are reasonable for a comparison and one is not; a caller who passed AsEnacted would get a number that is a definition change wearing the shape of a funding change. The function takes no basis and says why in its doc comment.