Proposals

After writing the 13 class definitions from decision ontology, the full schema layer was read back and examined for what it implied but did not yet declare. Three tables were presented to the user.

IMPLIED EDGES PRESENTED (12):

doctrine -[tensions-with]-> doctrine doctrine -[grounded-in]-> legislation litigation -[challenges]-> program program -[funds]-> education-agency revenue-stream -[substitutes-for]-> revenue-stream fiscal-period -[part-of]-> fiscal-period litigation -[cites]-> litigation actor -[succeeds]-> actor education-agency -[merged-into]-> education-agency formula-component -[replaces]-> formula-component metric -[derived-from]-> metric scenario -[compares-to]-> scenario

The first six were characterized as structural fixes rather than enrichments, because each closes a hole that makes something in the confirmed scope inexpressible:

• doctrine as written had zero outgoing edges. Its instances would have been near-orphans, and the class could not reach the constitutional text it originates in.
• litigation could challenge only a funding-regime. The 2025 Franklin County EdChoice decision challenges a program, and vouchers are in confirmed scope.
• program had only diverts-from. Capital facilities assistance is a positive flow to a district, and the capital channel is in confirmed scope, so the entire channel was unrepresentable.
• Ohio repeatedly substitutes one revenue stream for another (tangible personal property tax replacement payments, casino revenue offsets). Without an edge the substitution reads as an unexplained gap in one series and an unexplained appearance in another.
• A biennium contains two fiscal years. fiscal-period.ont.yml states this in its own description while declaring only follows.

The remaining six were characterized as enrichments: cheap to declare now, expensive to retrofit once instances exist.

CONNECTORS PRESENTED (9): dew-foundation, tax-abstract, lsc-budget, ohio-laws, ohio-courts, ofcc-projects, census-f33, nces-ccd, bls-cpi.

CALCULATORS PRESENTED (9): foundation, millage, local-capacity, deduction, dispersion, deflate, regime-diff, project, scenario-delta.

Everything presented was approved as presented. No item was modified or discarded.

IMPLIED EDGES — all 12 approved. Wired in step 7 both as declared edge types in the relevant .ont.yml files and as links: entries on the specific instances they connect. Declaring them in the schema as well as the instances keeps the class definitions an accurate account of what edges exist; an instance link with no corresponding schema edge would be an undeclared relationship.

CONNECTORS — all 9 approved:

dew-foundation Dept. of Education and Workforce School Finance Payment Reports and Foundation Funding files. Feeds education-agency, revenue-stream, metric. The authoritative per-agency state aid series. tax-abstract Dept. of Taxation property tax abstracts. Feeds revenue-stream, parameter, metric. Assessed valuation, effective millage, HB 920 reduction factors, 20-mill floor status. lsc-budget Legislative Service Commission redbooks and Catalog of Budget Line Items. Feeds legislation, fiscal-period, program. The only continuous appropriation-line series and the main pre-2000 source. ohio-laws General Assembly / LSC bill archive. Feeds legislation. ohio-courts Supreme Court of Ohio opinion archive. Feeds litigation. ofcc-projects Ohio Facilities Construction Commission. Feeds program, education-agency. Facilities assistance projects and local shares. census-f33 Census Annual Survey of School System Finances. Feeds metric, education-agency. Cross-state comparability, roughly two-year lag. nces-ccd NCES Common Core of Data. Feeds education-agency. Enrollment, identifier crosswalks, agency boundary changes. bls-cpi BLS CPI-U and ECI. Feeds metric. Deflator series.

CALCULATORS — all 9 approved:

foundation Per-agency state aid under a named regime, broken out by component. millage Effective millage under HB 920 reduction factors; 20-mill floor status. local-capacity State share index and local wealth position. deduction Community school and scholarship deductions against a resident district. dispersion Equity statistics across agencies: coefficient of variation, McLoone and Verstegen indices, federal range ratio. deflate Nominal to constant-dollar normalization. regime-diff Component-level difference between two regimes for one agency-year. project Forward series for ADM, assessed valuation, and revenue. scenario-delta Baseline versus perturbed run, as a per-agency table.

Approval defines the target, not the delivery. Per the bootstrap protocol, connectors are invoked during seeding only where the demand threshold is met, and stubbed otherwise; calculators are run during step 7 only where the seeded corpus makes the result meaningful, and stubbed otherwise.

The user approved all three tables in full, declining the narrower options offered for each (six structural edges only; five state-primary connectors; four core calculators). This is consistent with the scoping choices recorded in decision ontology, where the maximum option was taken on every axis — 1851 to present, all four funding channels, district by fiscal-year grain, and simulation together with projection.

On the edges, the six structural fixes were not really optional. A class with no outgoing edges cannot satisfy the corpus rule that every node carry at least one outgoing link, so doctrine had to gain one; and two confirmed scope channels — vouchers and capital — had no expressible connection to the classes they act on. Approving the six enrichment edges alongside them costs nothing at schema time and avoids a later revision that would have to touch every affected instance.

On the connectors, the nine divide into three groups that serve different purposes and are hard to trade off against each other. Five are Ohio primary sources and carry the numbers and the law. Two are federal (census-f33, nces-ccd) and exist for comparability and for identifier stability across agency mergers, which a 170-year panel needs. Two are infrastructure: ofcc-projects for the capital channel, bls-cpi for deflation. Dropping bls-cpi in particular would have quietly undermined the stated goal, since a corpus spanning 1851 to the present cannot compare any two nominal figures honestly without it.

On the calculators, project was the only genuinely contestable item, since projection is a weaker epistemic act than simulation. It was retained because the user had already chosen “both simulation and projection” for what prediction should compute, and dropping it would have narrowed that decision after the fact rather than implementing it.

Nothing was discarded, so there are no rejected items to record. The risk this creates is scope: nine connectors and nine calculators is a large domain computer, and most will exist as stubs for some time. That is acknowledged rather than mitigated — the stubs record intent, and the phase that needs one builds it.