What we got wrong, and when we fixed it

We corrected enrollment and levy counts, withdrew an unsupported denominator label, and clarified several comparison universes.

A fresh derivation from the canonical 516-district master found two enrollment display errors. The median district’s 1991–2025 change is −13.75%, which rounds to −13.8%, not −13.9%. Exactly 20 districts equal their highest recorded enrollment within the 1991–2025 dataset in 2025; the earlier count of 28 had treated several rounded near-peaks as ties.

The performance-levy analysis also mixed rounded and unrounded language. Among the 211 included districts below $3.43, 192 — 91% — levy within half a cent of their certified ceiling. The site previously said 194, or 92%, and described all of them as exactly at the legal maximum. The revised language states the half-cent rule directly and does not turn a rounded comparison into an exact equality.

DESE’s current-expenditure guide publishes variants using both average daily attendance and September membership. The downloaded MCDS field used by this site does not identify which variant it contains. The site still reproduces DESE’s published district value, but it no longer labels that value as necessarily divided by September membership. A separate limitation now explains that Special School District records its own special-education spending while students remain enrolled in 22 St. Louis County component districts.

Median enrollment change since 1991: −13.9% → −13.8% Districts at their own 2025 enrollment peak: 28 → 20 Below-benchmark districts within half a cent of their ceiling: 194 (92%) → 192 (91%) Median real teacher-pay decline: 7.6% → about 7.5%

Three interpretation changes did not alter underlying data. The Teacher Pay page now distinguishes MNEA’s statewide 274-of-518 Baseline Salary Grant summary from the matched map’s 272 users among 510 salary-schedule records. Funding retains the reproducible spending/APR correlation (r = +0.030) but removes a threshold-dependent count that changed with tie rules. Salary comparisons now say whether they refer to the top of the Master’s lane or the highest published cell across any lane.

District child-poverty estimates now use Census SAIPE 2024, and teacher-pay context now includes the NCES comparable-wage index.

The site updated the current district child-poverty layer from the 2023 to the 2024 Census SAIPE release. The new file covers all 516 mapped districts. The statewide estimate is 13.9% of children ages 5–17, while the median district estimate is 15.4%. Because SAIPE is model-based, annual estimates for small districts can move sharply. Eighty districts changed by more than five percentage points between the two vintages, so the site now warns readers not to treat a one-year movement as proof of a local trend.

The Teacher Pay page and district profiles also add the 2022 NCES ACS-CWIFT index for all 516 mapped districts. It estimates local wages for comparable college-educated workers. It does not measure housing costs, where teachers live or what a particular teacher should earn. The site therefore uses it only as labor-market context and retains the published standard errors in the project data.

No Title I allocation layer was published. Federal allocation figures remain on hold until their school-year and allocation-stage definitions can be reconciled without mixing preliminary, final and carryover amounts.

Gideon 37 joined the 2026–27 four-day-week cohort.

The provisional current-year list omitted Gideon 37 after the district switched to a four-day calendar for 2026–27. Gideon is now included in the adoption map, district profiles, mobile-app data and current statewide count.

Districts scheduled for a four-day week in 2026–27: 192 → 193 Rural districts in that current group: 183 → 184 New 2026–27 adopters: 8 → 9

The historical comparisons do not change. Gideon was not on a four-day week in the 2025–26 comparison year, so the page’s pay, spending, enrollment and academic results retain their documented 2025–26 or earlier universes. The current-year list remains provisional until DESE confirms district calendars in October.

Zero debt-service levies now read like zero, and hosted programs carry a spending warning.

The district profile used one generic sentence for every levy: it said the difference between the total and operating levy repaid building debt. For 189 mapped districts, including Pleasant View R-VI, the debt-service levy is $0.00 and the operating levy is the total levy. The underlying figures were right, but the sentence was wrong. Profiles now state the operating and debt-service amounts separately and say plainly when no debt-service tax is levied.

A Pleasant View R-VI beta reviewer also explained why that district’s high DESE current-expenditure-per-pupil figure needs context: the district hosts an early-childhood special-education cooperative serving five districts, and the program’s expenses are included while the participating students are not all included in Pleasant View’s enrollment denominator. The site retains DESE’s published value, labels the measure more precisely, and adds the documented district note.

The Funding page, 13-year spending calculator and Methods page now carry a broader caution for districts that host multi-district special-education cooperatives or regional career/technical centers. Some program expenses may be recorded by the host while some students remain counted in their home districts. No spending figure or calculator total was changed or estimated. The correction is to interpretation and disclosure.

The rural and suburban median district operating levies are effectively the same.

The site compared a $3.60 rural operating levy with a $3.40 suburban levy and said the median rural district set a higher rate. That comparison mixed two measures. The rural figure used the district’s operating levy, while the suburban figure was pulled toward the residential homeowner rate for St. Louis County districts that levy different rates on different property classes. Missouri law defines the district operating levy as those subclass rates weighted by the assessed value in each class.

Median district operating levy: $3.60 rural vs $3.40 suburban → $3.61 rural vs $3.61 suburban Interpretation: rural districts set a higher median rate → nearly identical median rates apply to unequal property-tax bases

The Rural vs. Suburban page now also places equal-district medians beside enrollment-weighted medians for the four measures challenged in an independent red-team review. The schedule-maximum gap narrows from $39,309 to $36,650 under enrollment weighting; the property-wealth and child-poverty contrasts widen; and the weighted levy comparison reverses. This corrects the levy claim while showing which broader findings survive a different weighting choice.

The district maps do not cover every Missouri educational setting.

The site identified its 516-district map universe, but it did not give readers one clear place to see what sits outside that boundary. That omission could make a district map look like a complete count of Missouri education even though public charter schools, full-time virtual programs, private schools, home and Family Paced Education, and MOScholars require different reporting universes.

A new Beyond District Schools page now distinguishes those categories instead of forcing them into one misleading pie chart. It reports dated public and private totals where defensible, labels home/FPE comparisons as data not publicly available, identifies the charter, MOCAP and MOScholars questions awaiting agency responses, and presents the major teacher-pay, MOScholars, charter and calendar provisions of SB 727 together.

The page also adds a cautious answer to “where did the students go?” In the matched 516-district panel, enrollment rose 2.1% from 1991 to 2025 while shifting away from city and rural districts toward suburbs and towns; it then fell by about 58,500 students from the panel’s 2000 peak. Those totals describe redistribution and decline, not individual children’s destinations.

No district-map value changed. This is a scope correction: it makes the boundary and unavailable comparisons visible, keeps overlapping categories from being added together, and records the evidence that cannot yet support a statewide comparison.

DESE’s accreditation summary contains one more district than its detailed tables.

The February 2026 classification PDF’s summary box reports 512 accredited districts, five provisionally accredited districts and 517 total. Its district-by-district tables contain 511 accredited, five provisional and 516 total. We extracted every six-digit county/district code, checked for duplicates and joined the detailed list to the site’s 516-district geographic master in both directions. Every code matches; the mapped panel is not missing a geographic district. Missouri’s two special school districts are outside this accreditation comparison.

DESE PDF summary: 512 accredited; 517 total → retained as a disclosed source inconsistency Detailed rows used by this site: 511 accredited; 5 provisionally accredited; 0 unaccredited; 516 total

No prior website count was corrected. This note records an inconsistency in the official source before accreditation status appeared on the Achievement page. The page uses the complete detailed list and links to the official PDF.

The Four-Day adoption map’s 2026 label named the wrong school year.

The map uses the school year’s ending year: a district coded as a 2026 adopter began the four-day schedule in 2025–26. The page caption and map key instead said that those districts were projected for 2026–27. That confused the existing 2025–26 cohort with the map’s separate New for 2026–27 category.

Map label `2026`: described as projected for 2026–27 → correctly identified as school year 2025–26 Projected cohort: blurred with the 2026 label → identified only by the separate `New for 2026–27` category

No district count changed. This was a year-label error in the explanation. The current 2026–27 count remains 192 scheduled districts, with the new cohort provisional until DESE confirms calendars in October.

We corrected stale poverty-method language and made the project’s independence explicit.

The Achievement page still said a Census poverty sensitivity test was work we intended to do. The test had already been completed: in the matched 2009–2019 data, the Census SAIPE measure alone accounts for about 21% of the district variation in achievement level, the historical free/reduced-lunch measure alone accounts for about 42%, and a model including both accounts for about 43.4%. The page and Methods now report that completed check and explain why the historical FRL series was retained.

The site also repeatedly identified Missouri State University beside the author’s name in page descriptions and footers. Although those passages called the project independent, their repetition could imply that the university published or sponsored it. Metadata and sitewide author lines now identify Show-Me Missouri Schools as the publisher and describe the project as independent. The full perspective statement identifies Jon Turner as a retired Missouri public school teacher and administrator, a Missouri State University professor, and a Missouri Association for Rural Education board member, while making clear that neither organization sponsors, endorses, oversees or publishes the work.

Census sensitivity test: described as future work → reported as completed, with results and limitations Site publisher: unintentionally blurred by repeated university affiliation → Show-Me Missouri Schools, an independent project by Jon Turner Achievement interpretation: the remaining variation could be read as school quality → explicitly described as a mix of unmodeled factors and measurement error, not a school-effect estimate

No underlying achievement calculation changed. This correction aligns the explanation with analysis already documented elsewhere and narrows what readers should infer from the 42% and 3% figures.

We narrowed the achievement claim and corrected three finance and staffing definitions.

The Achievement page said poverty “predicts how far behind students start” and repeated a 42% figure as though it described nearly every achievement map. The underlying analysis is observational, and the 42% belongs to one matched measure: pooled SEDA achievement level and average free/reduced-lunch enrollment for 2009–2019. In the same matched data, FRL accounts for about 3% of the variation in learning rate. The page now keeps the finding strong while using associative language and naming the measure, years and district counts.

Three definitions also needed correction or a tighter label. The Methods page described per-pupil spending as total district spending divided by students; DESE’s published measure is current PK–12 instruction and support expenditure after specified exclusions and deductions, divided by September membership. Another Methods note said every per-student figure on the site used DESE attendance even though the F-33 debt, capital and cash figures correctly use F-33 fall membership. Finally, the F-33 cash field was presented as though it were unrestricted operating reserves, and a new small-district comparison needed to state that MNEA reports certified-staff headcounts, not staff FTE or class size.

Achievement claim: poverty predicts where students start; 42% applied broadly → economic disadvantage tracks achievement level much more closely than learning rate; about 42% versus 3% in matched 2009–2019 data DESE per-pupil spending: total spending divided by students → specified current expenditures divided by September membership F-33 denominator: DESE attendance → F-33 fall membership F-33 cash label: cash reserves → cash and investments across sinking, bond and other funds; unrestricted share unknown

The newly published correlation table uses the site’s current district fields as a separate transparency check and labels their different vintages. It is not mixed into the matched 2009–2019 headline calculation.

The completed levy-election reconciliation found three more successful increases and resolved every district name.

The earlier correction to 136 was still three events short. A line-by-line partition of every voter-flagged school operating row in all seven State Auditor reports found 140 candidate district-year rows: 139 successful operating-levy increases and one excluded decrease (Strain-Japan R-XVI in 2021). We had wrongly set aside three real increases: Smithville R-II in 2020, Mehlville R-IX in 2023, and Maplewood-Richmond Heights in 2025. Each district’s own record confirms voter approval and the operating-levy increase or transfer.

Successful operating-levy increases: 136 → 139 Distinct winning districts: 119 → 121 Per-year corrections: 2020: 23; 2023: 20; 2025: 24 → 2020: 24; 2023: 21; 2025: 25 Winning districts still below $3.43 immediately after at least one increase: 23 → 22 Districts with no recorded success in these seven reports: 397 → 395

The below-$3.43 count also contained a separate classification error: Smithville’s 2025 printed row showed a $3.1927 base operating ceiling, but the district also carried a temporary operating component. Its full operating ceiling was $3.8169, so it was not below the performance levy after that increase.

Every event now has an unambiguous district match. The build joins the Auditor’s legal name and tax year to a stable DESE/Census district identifier, never a fuzzy name guess. That resolves Cass County R-V as today’s Archie R-V and Scott County R-V as Scott Co. Central. District profiles now show the verified seven-report history again. When no success appears, the profile says exactly that — never that voters did not ask, because Missouri still has no statewide record of failed proposals. The generator will refuse to publish a history file if any voter-flagged report row is left unmapped or unclassified.

We removed unreliable district-by-district levy-election labels while preserving the verified statewide totals.

The district lookup used an older field to say that a specific district had won an operating-levy election in a particular year, or that the site had no confirmed win for it. That field contained only 13 districts from the superseded extraction. After the statewide record was rebuilt to 136 increases across 119 districts, 10 of those 13 older district matches did not reconcile cleanly by name. The aggregate count is independently verified; the individualized labels are not.

District lookup election status: individual label based on an incomplete district-match field → no winner/non-winner label until all district names are reconciled Verified statewide record: unchanged at 136 increases across 119 districts in seven reported tax years

The lookup now explains the limitation at the point of use. A missing individualized label must not be read as evidence that a district never asked its voters, particularly because Missouri keeps no statewide record of unsuccessful levy proposals. The per-district field will return only after every match has been checked against the seven State Auditor reports.

Resolved later on 27 July 2026: the full identifier-based reconciliation described above mapped every event, so the guarded district-level history has returned.

Two assessed-value source labels named the wrong vintage, and one denominator was missing.

The district lookup and the Rural vs. Suburban page already displayed property-wealth figures calculated from the Missouri State Auditor’s tax year 2025 assessed values. Their small source labels still described those figures as DESE’s December 2024 vintage. The figures were right; the source and year labels were not. The lookup also said that 194 of the 211 districts below the performance levy were already at their ceiling without repeating that the analysis covers 513 districts after excluding three statewide virtual-school hosts.

Dime-test source label: December 2024 AV → State Auditor, tax year 2025 Performance-levy universe at point of use: 194 of 211, universe unstated → 194 of 211 within 513 districts Rural/suburban wealth medians: DESE, December 2024 → State Auditor, tax year 2025; 386 rural and 45 suburban districts

The two excluded rural districts are statewide virtual-school hosts, so their reported enrollment is not a local denominator for property wealth per student. No displayed number changed in this correction; the labels now identify the data actually used and state the comparison universe where the figures appear.

Several explanations had not caught up with corrections already recorded on this page.

The 20 July levy-election correction fixed the main count and what is now the Taxes & Debt page, but two passages elsewhere still told parts of the superseded story. The Methods page still said only “roughly two districts a year” won an operating-levy increase, even though the corrected record shows about 19 a year. A generated district-lookup explanation also said 63 districts levy exactly $2.75 while the rest of the site used 64.

Successful operating-levy increases per reported year: roughly 2 → roughly 19 Districts levying exactly $2.75, all 516 mapped records: 63 → 64 Reported tax years in the series: six → seven

The $2.75 figures were both reproducible but used different denominators: 64 across all 516 district records, and 63 within the 513-district wealth-analysis universe because Laquey R-V is one of three statewide virtual-school hosts excluded from per-student property-wealth comparisons. The site now states that distinction instead of switching numbers without explanation.

We also restored the original wording of earlier dated correction records and added visible forward notes rather than rewriting those records in place. The debt-service history now shows all three verified stages: the 12 July parser correction moved the median total levy to $4.07; the 13 July operating-levy rebuild moved it to $4.08; and the later 13 July St. Louis County debt reconciliation produced the final $4.12. No source data changed in this cleanup.

Superseded later on 27 July 2026: the completed event partition moved the levy-election average from roughly 19 to roughly 20 successful increases per reported year.