FSFP Disadvantaged Pupil Impact Aid

The squaring is the whole character of the program, and no DPIA total shows it # ↑ Contents

Aid scales with the square of relative poverty, so a district at twice the state’s rate scores four times the index rather than twice. $525m distributed on a convex curve concentrates far more sharply than a per-pupil rate would, and the convexity is a policy choice as substantial as the amount. verified

The two poverty counts are not nested # ↑ Contents

d1 is 65% of the FY2025 economically disadvantaged ADM plus 35% of the FY2026 directly certified ADM. The second is administrative — a child is directly certified through participation in another benefit program — and it finds a median 61% of what the disadvantaged count does, fewer than half in 146 districts. Putting 35% of the weight on the smaller measure lowers the funded count almost everywhere. verified

The weighted ADM is capped at the district’s own enrolled ADM # ↑ Contents

The blend can produce more disadvantaged pupils than pupils, and where it would the published figure is the enrollment instead. verified

Where the statute says it # ↑ Contents

Disadvantaged pupil impact aid is R.C. 3317.022(A)(4), inside the section that assembles core foundation funding, and the $422 per pupil is statutory there. The squaring is at R.C. 3317.02(I)(1)(a), which defines the index as the square of the quotient of the district’s percentage over the statewide one — not a departmental elaboration but the General Assembly’s own construction. verified

The 65/35 blend is in the act, not the code # ↑ Contents

The Revised Code is not all of the law. H.B. 96 of the 136th General Assembly sets the blend in a table:

economically disadvantaged student ADM        FY2026    FY2027
FY2025 students                                 75%       65%
students certified as categorically eligible    25%       35%

So the blend is uncodified statutory law — the same category as the four supplements — changeable by the next budget and not by the department. verified lsc-budget, crates/project/tests/what_the_act_says_and_the_code_does_not.rs

What this repository computed # ↑ Contents

Not what Ohio publishes. Figures derived here from committed fixtures, each one citing the test that reproduces it.

THE COUNT THIS IS PAID ON HAS A FOURTEEN-YEAR HISTORY, AND IT ROSE BY MORE THAN HALF. R.C. 3317.03(B)(21) hands the definition of “economically disadvantaged” to the department, constrained only by “shall not be categorically excluded … based on anything other than family income” — and the department’s operative test has been free and reduced-price lunch eligibility. The Office for Child Nutrition published that count every October, and over Public sponsors it runs: verified dispersion::mr81, pinned by test

1998  28.9%      2003  33.1%      2008  39.6%
1999  28.6%      2004  33.8%      2009  42.6%
2000  28.5%      2005  35.7%      2010  45.0%   (denominator redefined)
2001  29.5%      2006  36.8%      2011  46.3%
2002  31.2%      2007  36.6%

17.4 points in fourteen years, most of it before the 2008 recession rather than after. What the state pays per disadvantaged pupil is legislated — $422 — and how many pupils qualify is delegated and has been rising steeply. A weight on a count whose definition one party sets and whose size neither controls is a different fiscal instrument from a weight on a fixed population, and the corpus had no way to see that before this series.

TWO FIGURES IN THE TABLE THIS REPLACES WERE WRONG, AND THEY WERE WRONG IN OPPOSITE WAYS. 2001 read 27.7%: the file is the only comma-delimited one in the series, and two Cleveland City rows whose school names carry a comma had put site identifiers in the enrollment column, adding 119,603 pupils to a district of 73,562. 2005 read 29.9%, which is the figure with the published 342,332-pupil defect left in — the one thing implausible_sponsors exists to take out. The first was a reading error and the second was a quoting error, and both survived because a plausible percentage in a rising series looks like a measurement. verified

Three cautions travel with it. The denominator is renamed and redefined in 2010, AdmCount to CECount, so the last two points are not on the same base as the first twelve. Sponsors are not districts — “Public” includes community schools and county boards of developmental disabilities. And the applications-based series ends in 2011. verified

IT ENDS BECAUSE THE MEASURE ENDS, NOT BECAUSE THE ARCHIVE DOES. Three more Octobers are published and extracted, and from 2012 the report is three files of which only the Traditional one still counts applications: community-eligibility schools feed every child and collect no forms. The enrollment under such a sponsor goes 7.0% → 9.9% → 16.6% in three years, so adding the three files gives 33.3% for 2014 and would read as poverty collapsing. What the source supports for 2014 is 43.7% to 48.4% — the directly certified count as a floor, what those schools may claim for as a ceiling — and that band contains 2011’s 46.3%. The direction is what cannot be settled. verified dispersion::mr81::streams_by_year, pinned by test

Whether the department’s current economically-disadvantaged count still tracks free-lunch eligibility is now half-answered: after 2011 free-lunch eligibility stops being a single measure at all, so any series joining this to the present is joining two definitions. Which one the department uses today is not established here. open

The squaring holds for all 609 districts and in both directions: away from an index of one, the plain ratio and the published figure diverge. verified

Almost. Three districts have more directly certified children than economically disadvantaged ones — Beachwood 1.03x, Nordonia Hills 1.54x, Ottawa Hills 2.35x on 32.9 disadvantaged pupils. All three have poverty shares of 0.03 to 0.18 against a statewide median of 0.49. Small counts, where two differently-built measures can invert. Recorded because the corpus briefly asserted the blend could not exceed its larger input, which is false, and because either count is a tempting poverty proxy elsewhere. verified

Where the cap binds, and why # ↑ Contents

Edgerton Local’s published blend is 342.38 against 352.03 computed from the stated 65/35 weights. Its enrolled ADM is 342.3818, and the published figure is that number rounded to the two decimals the column carries.

Three independent selections over the panel return the same single district, which is what makes this an identification rather than a coincidence: the only district whose blend exceeds its enrollment, the only one whose weighted ADM equals its enrollment, and the only one at a disadvantaged percentage of ~100%. verified crates/project/tests/questions_the_corpus_left_open.rs

The cause: a prior-year count against a collapsed enrollment # ↑ Contents

The blend’s first term is the FY2025 count. Edgerton’s enrollment fell from 504.48 in FY2025 to 342.38 in FY2026 — a third of the district in one year. Its count of 497.21 is 98.6% of the enrollment it was measured against and 145% of the one it is applied to.

The three districts whose count most exceeds current enrollment make the point better than any argument:

district                 ED/FY2025   current/FY2025   ED/current
Edgerton Local             0.9856        0.6787          1.4522
Garfield Heights City      0.9878        0.8634          1.1441
Clearview Local            0.9857        0.9042          1.0901

All three sit at 98.6% against the year the act measures in. What separates them is entirely how much they shrank. A ratio that is identical in three unrelated districts is not a data error in any of them — the dispersion is in the denominator. verified crates/project/tests/what_the_act_says_and_the_code_does_not.rs

The second count is a year behind the weights, and the act says it should not be # ↑ Contents

The act names the vintage, which nothing else did: the first term is the FY2025 count, not the year the aid is paid in.

H.B. 96 fixes the first term at FY2025 and sets the second on “the ADM of students directly certified as economically disadvantaged for the fiscal year for which the DPIA payment is calculated”. The greenbook writes the FY2027 row out to match — (FY 2027 Directly certified ADM x 35%) — with the rule in prose beside it: “the formula for each fiscal year will use a district’s directly certified ADM for that year.” The department’s FY2027 workbook carries FY2026 against those FY2027 weights. Its DPIA sheet heads the column d1b FY26 Directly Certified ADM, and its Directions sheet names the collection the column came from: FY26 Nov #2. verified the workbook’s own header and vintage table

It is the simulation, not an error, and it is not an exception either. FY2027 direct certification has not been collected, so the model substitutes the latest year that has — and enrolled ADM, the categorical FTEs and preschool special education ADM are all FY26 (Nov #2) on the same sheet. Every count in the model is the most recent collection there was rather than the year being funded. What that costs a reader is specific: every DPIA figure the corpus carries is computed on a count the actual FY2027 payment will not use. verified

The sizes say the same thing without the workbook. The greenbook publishes both counts statewide for traditional districts in FY2025, and the term the corpus labels FY2025 lands on its figure while the term it labels FY2026 does not:

term                              greenbook, FY2025    the model carries
economically disadvantaged ADM      847,000 (est.)         856,236     +1.1%
directly certified ADM              557,000                474,197    -14.9%

A term on the greenbook’s vintage would sit inside the first row’s band. verified crates/project/tests/what_the_act_says_and_the_code_does_not.rs

The 82,800-pupil gap is not a measured one-year fall in direct certification. The two figures are a fiscal year and possibly a collection point apart — Nov #2 is a point in the year and the greenbook does not say what its own is — so the difference is not attributable. What it settles is only that d1b is not FY2025’s count, which together with the header is the whole question. verified

The blend improves the targeting and cuts the pot, and the cut is the larger term # ↑ Contents

The direct-certification term was recommended before it was enacted. A Fordham Institute commentary on the House-passed budget called the count “a mess” — community eligibility inflating the disadvantaged rate until it no longer distinguishes a poor district from one that feeds everybody — and named direct certification as the fix that would “significantly improve the allocation of funds”. Three months later the enacted act adopted it. verified Fordham Institute — "Ohio House puts the brakes on Cupp-Patterson"

The targeting claim is right. Holding the statewide dollar total fixed so that only the distribution can move, the 65/35 blend sends the eight urban districts +$16.07m against the pure disadvantaged count — +11.3%, +$96 a pupil — and 423 of 609 districts gain. The districts that pay for it are the ones the diagnosis names: Pickerington Local at −$3.08m is the largest single loser in the state, a Columbus suburb reporting 97.9% economically disadvantaged against 25.5% directly certified, and Cleveland Municipal at +$5.82m is the largest gainer. verified crates/project/tests/what_direct_certification_moved.rs

And it is a cut about twice the size of the improvement. DPIA for traditional districts falls $84.5m then $31.8m, $649.2m to $532.8m before phase-in. The eight hold 30.1% of enacted DPIA, so their share of that $116.4m is roughly −$35m against +$16m of better targeting — net about $19m worse off. inference the apportionment, which is pro-rata and the cut is not verified the share and the sign of the comparison A better measure of a smaller pot, and either half quoted alone describes a different policy than the one that passed.

The counterfactual rescales the index rather than recomputing it, and that is a declared limit. Under the prior count the statewide percentage below would itself have been larger, so holding it fixed while the counts rise 17.6% overstates the level badly. The denominator implied by the published 0.5334 is 1,354,592 — smaller than traditional enrolled ADM — so it divides a population this model does not carry and cannot be recomputed here. The run therefore holds the statewide total exactly fixed, which prices the redistribution and declines to price the level. The greenbook prices the level. open what the statewide percentage would have been

What the statewide share is # ↑ Contents

d3 divides by 0.533380310606710 — the statewide economically disadvantaged percentage. R.C. 3317.02(I)(1)(a)(i) defines it as a computation, not a constant: the statewide ADM it divides by is the sum of enrolled ADM for all city, local and exempted village districts, statewide community school enrollment, and statewide STEM school enrollment. So it is recomputed rather than carried forward, and its value moves when any of those three move. verified ohio-laws

It moved, and so did the weights above it. The corpus holds two years of the calculator — FY2027 from the department and FY2026 from the Internet Archive — and between them the statewide percentage goes 0.565990 to 0.533380, while the blend that feeds it goes 75/25 to 65/35. Both changes push the same way, and a third pushes with them: the directly certified count the weight moved toward was itself restated 16.80% lower at the median, under the same column label and from the same collection named on the Directions sheet. verified crates/project

Statewide the program pays $567,673,869 in the FY2026 model and $525,094,312 in the FY2027 one — a fall of 7.50%, in 562 of 609 districts. The economically disadvantaged count is the same input in both workbooks, identical in 593 districts of 609, so none of the fall is a change in how many poor children Ohio has. It is the weights and the certification count. verified

What is not established is why the direct-certification figure was restated, and the FY2026 workbook is the only other observation there is. open

What is delegated is narrower and more interesting. R.C. 3317.03(B)(21) hands the department the definition of “economically disadvantaged”, constrained only by “a student shall not be categorically excluded … based on anything other than family income.” The act then sets weights on two counts whose definitions it does not control. Two levels, not one. verified

The poverty measure is top-coded and this component is not # ↑ Contents

The report card’s economically disadvantaged share saturates at 100% under community eligibility, which is why 100 districts print exactly that. The count DPIA uses is an ADM rather than a share and does not saturate the same way — so the two poverty figures a district page can show are different measures, and the corpus holds both. verified