The JVSD rewrite that changed no percentage

Issue #369 recorded the JVSD base cost method as outside the lever set because it “needs a population the 609-district panel does not contain”. Issue #371 argued the second half was too weak a statement — the population is identified, its finances are committed, and the new method is written out in a committed greenbook — and that what was actually missing was narrower: each district’s base cost and base cost enrolled ADM, “as the department computes them”. It also recorded, from three probes, that the department publishes no JVSD calculator where it publishes its others.

The probes were right and the conclusion from them was wrong. There is no page at .../State-Funding-For-Schools/Joint-Vocational-School-Districts, and there is no JVSD calculator anywhere. But the department publishes payment reports for this population — what it paid, per district, after the year closed — and files them under Career-Technical Funding. No path on the department’s site contains the words “Joint Vocational” at all, so every guess at one returns a 404 that reads exactly like an absence. The archive runs FY2014 to FY2027.

For FY2022 through FY2026 a payment report is the better artifact than a calculator would have been, not a lesser one: it is an actual rather than a projection. Six years are now extracted, spanning the amendment — FY2022-FY2025 under prior law and FY2026-FY2027 under item 7.

And what was “actually missing” was never the hard part. Base cost, base cost enrolled ADM, the two valuations, the charged one, per-pupil base cost, per-pupil local capacity and the state share percentage are all published per district, and the Base Cost sheet carries the whole 76-column build-up #371 called unreconstructable.

And the extraction found what this record exists for: item 7 does not change the state share percentage. Base cost per pupil is aggregate base cost over the same base cost enrolled ADM that divides the half-mill charge, so the ADM cancels and the per-pupil form computes the number prior law computed. That is not an argument from algebra alone: 1 − (charged valuation × 0.0005) / aggregate base cost reproduces the published percentage on all 49 districts in all six years, on both sides of the amendment, to a part in a million.

What it changes is the ADM the percentage multiplies — from the three-year average built into aggregate base cost to the current year’s enrolled ADM. Prior law was a subtraction with no enrollment term anywhere; item 7 is a multiplication over a live count. Worth $10.2m in FY2026, 36 of 49 districts better off.

Neither the act’s own list nor the greenbook’s statement of the formula says this. Both describe the per-pupil restatement, which is the half that cancels.

Reported, not levered — the same disposition as “A supplement paid to a population the panel does not hold”, and for the same structural reason, with a second reason this population adds.

The structural reason: Policy is eight fields over one population. Every lever is priced through project::policy::apply over the 609 districts of foundation’s panel. A JVSD has no IRN among them. Unlike a community school it does have taxing authority, a valuation and a local share — so its formula is far closer in shape to a district’s, and the absence of a join is a fact about the panel rather than about the formula. A lever would still need a third Outcome channel carrying a total over a second population, which is what that record rejected and rejects here.

The second reason is specific to this population: the lever it would most obviously take is dead over most of its range. A MinimumStateShare lever is the natural one, since item 7 explicitly maintains the 10% minimum and project::policy already carries that field for districts. No joint vocational district has been on that floor in any of the six years — the lowest share observed is 0.10923, in FY2027. Dialing the floor down therefore moves exactly zero dollars, and dialing it up moves money only once it passes a district’s actual share. The corpus would be shipping a control whose null region it had measured and not mentioned.

This is not an argument that no lever over JVSDs could do anything — a base cost scale would, and not merely proportionally, since scaling aggregate base cost moves the state share percentage as well as the amount it multiplies. It is an argument that the lever this population’s own amendment points at is the one with nothing behind it, which is a poor reason to open a second population inside Outcome.

The thing that is worth reporting is the one no lever would have shown: that a rewrite described as a change of method changes no percentage, and that its real effect is a change of enrollment vintage which neither published account names.

Six years rather than two, because a before and an after inside one fixture is what makes that statement checkable. dispersion::jvsd_funding reproduces the department’s own payments under prior law’s subtraction on FY2022-FY2025 and under item 7’s multiplication on FY2026-FY2027, to the cent, on every district.

FY2026 is the only year item 7 is priced in. FY2027’s report is a September payment whose enrolled ADM is FY2026’s on 43 of 49 districts, because FY2027 enrollment does not exist in September 2026. Under prior law that would have been harmless; under item 7 the provisional count reaches the money. item_7_incidence returns None for every year but FY2026 rather than returning a number that would mostly measure a placeholder.

One connector, not a new one. dew-foundation, six new sources. Same office, same release model, same epistemic status, a different population — the argument a-supplement-paid-to-a-population-the-panel-does-not-hold made for the community school simulator, unchanged.

A ninth Policy lever with a third Outcome channel. Rejected on the population and now also on the payoff: see the decision. The population argument alone would have left this a close call, since a JVSD’s formula has the valuation and local share a community school’s lacks. The measured null region on the minimum state share is what makes it not close.

Take only FY2026 and FY2027, the two years item 7 governs. Rejected. It is the cheaper extraction and it cannot state the finding: “the rewrite computes the same percentage” is a claim about both sides of an amendment, and with only the after side the identity reads as a restatement of the new formula rather than as a measurement across the change. The prior-law years are what make it falsifiable.

Take the whole archive, FY2014-FY2027. Rejected as scope without a question. FY2014-FY2021 precede the Fair School Funding Plan and are a different formula with different columns, and nothing in the corpus currently asks them anything. The span here is the plan’s own era.

Read the bracketed column tags, as a reader of one year’s file naturally would. Rejected, and recorded because it is the plausible mistake rather than a hypothetical one. The tags are not stable across the series: [I] is Total State Support in FY2024 and FY2025 and the Base Funding Supplement in FY2026 and FY2027 — a $559m column and a $1.4m one under one letter — and [A1] is Aggregate Base Cost in FY2024-25 and Base Cost Per-Pupil in FY2026-27, dollars against dollars-per-pupil. This is the same shift dew-sfpr-line-by-line records for the traditional district reports, made by the same act. Columns are found by the label text after the tag, and an ambiguous match is an error rather than a first-wins.

Take base cost enrolled ADM from Detailed SFPR, where the rest of the state share is. Rejected: it is the wrong column of two that are spelled identically. Detailed SFPR [i] and Base Cost [b] are both Base Cost Enrolled ADM and they disagree on five of the forty-nine FY2026 districts; the department’s own local capacity is computed from the second. Using the first reproduces 44 of 49 and misses the rest by 1-5% — close enough to read as rounding, and it is not.

Take per-pupil base cost at full precision from the Base Cost sheet. Rejected, and this is the subtlest of the three. Base Cost [F] carries 9612.6462793433 where Detailed SFPR [A1] carries 9612.65, and the department pays on the rounded one: the full-precision copy misses Apollo’s FY2026 base cost payment by $3.42. The two agree to a cent, so a check written at cent tolerance passes while discarding the only thing that distinguishes them.

Carry local capacity as a column of the fixture. Rejected as a second copy of a determined number: it is the charged valuation times 0.0005 exactly, on all 49 districts in all six years, against both the dollar column prior law publishes and the per-pupil column item 7 publishes. The identity lives in the builder, where a changed millage is a named failure rather than a column nobody re-derives.

Price item 7 as base_cost_aid_under_item_7 less base_cost_aid_under_prior_law. Rejected as the noisier of two correct framings. Those two expressions are rounded differently — one off a per-pupil figure the department rounds to the cent, the other off a full-precision aggregate — so their difference carries a residue of several dollars on a district the amendment does not move at all. Holding the share and the per-pupil cost fixed and varying only the ADM isolates the mechanism exactly, and makes a district enrolling its own average move zero rather than nearly zero.