Five-year forecast filings — the department's republished submissions

Source. Ohio Department of Education and Workforce, School District Five-year Forecasts. Districts file under R.C. 5705.391; the department republishes every filing as one flat file per submission window. Type. Primary source, with a qualification that matters — see What it is not below. Location. https://public.education.ohio.gov/School%20District%20Five-year%20Forecasts/, an open directory listing. Roughly 5–6 MB per tab-separated file.

What it contains. One row per district per forecast line code, about sixty lines each, covering the general fund only. Each row carries eight values: three prior fiscal years as audited actuals, the filing year’s own forecast, and four future forecast years. 660 reporting bodies — Ohio’s 609 traditional districts plus joint vocational school districts.

The lines this corpus extracts:

Code	Line	Why
1.010	General property tax (real estate)	The levy yield actually collected, after H.B. 920
1.030	School district income tax	Where one is levied
1.035	Unrestricted grants-in-aid	State foundation money as the district books it
1.040	Restricted grants-in-aid	State money it may not spend freely
1.050	Property tax allocation	The state reimbursing rollback and homestead
1.070	Total revenue	The district’s own receipts
2.080	Total revenue and other financing sources	The only one that closes the cash identity
5.050	Total expenditures and other financing uses
7.010	Beginning cash balance, 1 July	Carry-over
7.020	Ending cash balance, 30 June	Cash on hand

Why it matters here. Every other per-district figure in this repository is modeled — what a formula computes, what a district spent per pupil on the department’s definitions, what its pupils achieved. This is the only record of money that changed hands, and the only measure of what a district holds rather than what it is given.

It also reaches two things that were otherwise blocked. FY2020 — the year the temporary transitional aid guarantee holds districts at — becomes an observation rather than something inferred from the guarantee itself, which had made every check of the guarantee circular. And 1.010 is the local side as revenue, which tax-abstract would have supplied as valuation and still has not.

Which filings are pinned, and why those. Two, three years apart, so their actual windows tile without overlapping: the FY2023 required spring update (actuals FY2020–FY2022) and the FY2026 required spring update (actuals FY2023–FY2025). Six consecutive closed years. The required spring window is chosen over the optional fall and final ones because every district files it.

Every figure in it is nominal. The span it covers contains the sharpest price change in forty years — CPI-U June rose 25.1% from FY2020 to FY2025 — so a nominal statement about this panel can have the wrong sign, not merely the wrong magnitude. Statewide district cash ends the span 8% above where it started in nominal dollars and 14% below in constant ones. Anything drawn from this source must say which basis it is in, and crates/deflator is what converts it.

Caveats:

• Actual and forecast are different claims and this corpus extracts only the actuals. The forecast columns are a treasurer’s projection made under incentives — a board arguing for a levy and a board defending a balance want different numbers on the same line. Treating one as a measurement is the characteristic error with this source.
• Booked aid is not formula output. 1.035 is what lands in the general fund; the FY2027 calculator’s “total state support” is a different construction. A ratio between them carries that gap, and any claim resting on one must be stated against a control measured the same way.
• General fund only. Capital, food service, and most federal programs sit in other funds. An expenditure figure here is not the district’s total spending.
• FY2021–FY2024 are the pandemic relief years. ESSER money was booked in the general fund by some districts and separately by others, so a balance rising across that span is not evidence about a district’s own position. project::finances::Finances::pandemic_years names the window in code.
• The two filings can disagree about the instant where they meet. FY2022’s closing balance and FY2023’s opening balance are the same moment reported by filings three years apart. The median gap is $2 and the 90th percentile $9,054, but nine districts restate by more than a million and the largest by $13.3M (Paint Valley Local). These are reclassifications, not errors, and the panel is not a single continuously audited series.
• A filing may carry a district’s year and not every line of it. Toronto City (IRN 044917) is absent from lines 5.050, 7.010 and 7.020 of the FY2023 file, so it has no expenditure and no cash balance for FY2020–FY2022. The extract writes those cells empty, not 0: read as zero, the district published $0 of spending against $9.86M of revenue in the live feed for three years, and appeared as a tenth $1M-plus seam restatement above. Not recoverable by re-fetching — the lines are not in the published file. Pinned in project::finances::INCOMPLETE, which is what PARTIAL cannot see: PARTIAL counts rows, and this district has all six.
• One district is short. Green Local (IRN 049619, Scioto County) is in the FY2027 funding model and in the FY2023 filing, and absent from the FY2026 required spring update — so it carries FY2020–FY2022 only. Pinned in project::finances::PARTIAL.
• Three published layouts. Files from FY2022 onward carry a header row. The FY2013–FY2021 files are headerless with a different column order, and the 2008–2012 files use zero-padded fixed-width amounts and a different line-code format. The parser is header-driven and refuses a headerless file rather than reading it positionally. Extending the panel before FY2020 means writing a per-era reader, the same shape of problem as census-f33.

The committed extract is crates/project/fixtures/district-finances.csv: one row per district per closed fiscal year, 3,957 rows. Both source files are digest-pinned in crates/connect/source-digests.txt.