Recognized valuation

crates/regime-diff ran the charge-off counterfactual against total taxable value and carried a caveat saying so: the charge-off’s real base was recognized valuation, “an H.B. 920- adjusted figure deliberately lower than total taxable value wherever reduction factors bind”, which the corpus did not hold. The recorded blocker was that tax-abstract had been expected to supply it and turned out not to.

Both parts were wrong, and the definition was the more serious of the two.

Recognized valuation is not an H.B. 920 adjustment. The Legislative Service Commission defines it arithmetically: valuation “recognizes” a district’s inflationary increase in carryover real property evenly over three years instead of all at once — two thirds deferred in the reappraisal or update year, one third the year after, nothing by the third. H.B. 920 holds revenue flat by cutting the rate; recognized valuation holds state aid flat by deferring the base. LSC discusses them under one heading — “Provisions that Soften the Effect of H.B. 920 Tax Reduction Factors” — which is the likeliest way the conflation happened.

The error named the wrong districts, not just the wrong amount. Under the recorded reading the overcharged districts were those with long-bitten reduction factors: a wealth-and-tax-history story. Under the real rule they are the ones whose county reappraised recently. Those sets are close to unrelated.

And the figure was never going to arrive by retrieval. Nobody publishes recognized valuation per district; it was an internal step in a formula retired in 2021. Waiting for a connector to carry it would have waited forever.

Reconstruct recognized valuation rather than retrieve it, from three things the corpus can hold:

• the rule, transcribed from LSC with its worked example reproduced as a test;
• the Department of Taxation’s county reappraisal calendar, carried as constants in regime_diff::recognized_valuation::CYCLE and cataloged;
• Table SD-1 extended from two tax years to four, TY2021–TY2024.

Estimate each district’s inflationary increase as its event-year growth in excess of its own ordinary growth, measured from its two non-event years. Per district, not against a statewide average, because a growing suburb and a shrinking rural district have different ordinary rates.

Make the valuation base a required argument to at_fy2027 rather than a default. ChargeOffBase::{TotalTaxable, Recognized} forces every call site to say which it means, which is the mechanical guard against the class of error that produced this decision: the corpus got the base wrong by never having to choose one.

Carry the correction across sources as a ratio, not a level. The profile report’s FY2023 assessed valuation per pupil and SD-1’s TY2024 total taxable value are different vintages on different denominators; the recognized share transplants soundly where a dollar figure would not.

The site’s headline regime comparison reversed. The corpus published that the median district is $289 per pupil better off under the Fair School Funding Plan than under the charge-off. On the corrected base it is $45 per pupil worse off, and 316 districts would have done better under the charge-off against 290 under the plan — where the published figures were 193 against 413.

The incidence claim moved further. The corpus published that every wealth quintile does better under the plan, most of all the richest. Only the top quintile still gains; the bottom four lose by $118 to $365 per pupil. The shape survives — the plan is genuinely more generous to property-rich districts — but “every quintile gains” was an artifact of overstating the charge-off by about 8% for everyone.

Statewide the deferral is 8.20% of taxable value at TY2024, $34.5bn, $793m of charge-off at 23 mills, and it falls entirely by calendar: 12.97% for the 184 districts whose county revalued in TY2024, 7.04% for the 304 in TY2023, nothing for the 123 in TY2022. Two districts identical in wealth, effort and need are charged a tenth differently because of which year their county auditor was scheduled.

TY2024 is a high point of the cycle and the pages say so. LSC put the long-run effect nearer 2% of valuation and $125m a year. This cycle’s revaluations were roughly twice the size LSC’s own worked example assumes, so the correction is larger now than it would be in an ordinary year. The direction is not cyclical: total taxable value always overstates the charge-off.

Four tax years broke four things that assumed two, none of which would have failed loudly. millage_analysis in the bundle, the statewide rateFell derivation in feed.ts, the tax change card in tax.ts, and a crate test each took first() and last() of the series. With two years those are adjacent; with four they span a reappraisal, and every one would have gone on rendering a plausible number. Each now slices the last two and says why. The crate test was worse than wrong — its [before, after] pattern stopped matching at all, so it silently exercised nothing until a count assertion caught it.

Keep the caveat and change nothing. Rejected once the definition turned out to be wrong: a caveat that misdescribes its own subject is worse than no caveat, because it tells a reader the error is somewhere it is not. This was the status quo for fourteen phases.

Wait for a source that publishes recognized valuation per district. Rejected because there is no such source. It was a working step inside a retired formula, and the years it governed are in ODE payment reports that dew-payment-reports records as unindexed and partly missing. The reconstruction needs no retrieval at all beyond one more year of a table already wired.

Estimate the inflationary increase against a statewide background rate instead of each district’s own. Rejected: per-district background growth has a median of 0.96% and a 10th-to-90th range of 0.12% to 3.28%, so a single rate would misattribute a fast-growing district’s new construction to its reappraisal. The staggered calendar makes the per-district control free, which is the argument for the four-year window over a three-year one.

Apply the phase-in to total value rather than to real property alone. Rejected as simply not the rule — LSC’s increase is in carryover real property, and public utility tangible personal property does not reappraise on the county cycle. The deferral is computed on real property and subtracted from total value, which is what the mechanism does.

Model the exempt property adjustment too. Not rejected, unbuilt. LSC records a separate adjustment for about 13 districts with large state-owned exempt property — $836.4m in FY2008, $19.2m of local share, roughly 0.3% of statewide valuation. It needs the department’s list of those districts, which the corpus does not hold, and it is recorded as [open] on the parameter node rather than approximated.