Skip to content

Audit summary: how this investigation was produced

Status: independently reviewed, October 8, 2026 (document-based analytical review; corrections applied the same day). The review is described in "Independent review and response" below. It is not a financial-statement audit, and the report still has known open items (listed below and in each page).

Origin and method

The investigation began from a direct question: was it true that Houston ISD's per-student reimbursement from the state of Texas had fallen? It was prepared with an AI assistant (Claude) working from public records under a person's direction, between 2026-10-06 and 2026-10-07.

  1. State formula data. For the first edition, the person directing the work downloaded three "Summary of Finances" reports by hand from TEA's public report dashboard (a JavaScript application with no automated data export) and supplied the PDFs; they are reproduced in full in sources/. For the second edition, the remaining years were retrieved from the same dashboard with a site-restricted browser tool (see "Second edition" below).
  2. Four parallel research tracks, each run as a separate research pass and cross-checked against the others afterward:
  3. HISD's audited financial statements (ACFR), FY2022-FY2025.
  4. HISD's adopted budgets, FY2025-FY2027, and in-year amendments.
  5. TEA/Comptroller data: PEIMS actuals (district and by campus), TAPR staffing data, School FIRST ratings, and peer-district comparisons.
  6. Federal funds (single audits), bonds and debt, and other policy-driven money flows (TIRZ, recapture history, NES funding, vouchers, charters).
  7. Spot-checks against source PDFs. Headline figures from each research pass were independently re-verified by searching the extracted PDF text (or, for image-only pages, OCR output) and confirming the exact figure, before being written up. Several figures were corrected this way; see below.
  8. A reproducible extraction script (scripts/sof_extract.py) parses the three TEA Summary of Finances PDFs into data/sof_key_figures.csv and checks that the totals reconcile internally (state aid equals the sum of its components, in each of the three reports) before writing the output. It exits non-zero if a check fails.
  9. Browser automation, added partway through the investigation (Playwright, restricted by configuration to TEA's own report domain, using the system browser), was used to pull one additional school year's recapture figure directly from TEA's live dashboard, as a primary-source cross-check against the audited figures.

Corrections made during the investigation

  1. The explanation for HISD's rising state aid was wrong in an early draft, and was corrected. The first explanation offered was that falling enrollment had reduced HISD's local tax revenue below the threshold for recapture. That is backwards: a falling entitlement from fewer students should, if anything, increase recapture risk, not reduce it. The actual cause, confirmed directly from TEA's own compression worksheet, is the 2023 property tax relief law (Senate Bill 2, 88th Legislature, 2nd Called Session): TEA's own calculation shows HISD would have owed $156.3 million in recapture for SY2026-27 without the additional 10.7-cent tax-rate compression that law enacted. The correction, with the worksheet citation, is in data/hisd_sof_trend.md.
  2. An earlier assumption that HISD does not pay recapture was wrong. HISD paid $49.1 million (TEA's figure) or $56.9 million (the audited figure; see the discrepancy below) in recapture for SY2024-25/FY2025, before the 2023 compression law brought it close to zero for the following two years.
  3. The ACFR's own "cost per pupil" statistical series was found to be computed inconsistently across years: for FY2020-2022, HISD's reported operating expenditures included bond-refunding escrow payments that should have been excluded, inflating those years' figures by the exact amount of the escrow payment. Recomputed, consistent figures are given in notes/acfr.md alongside the as-published ones.
  4. Twenty campus names in the TEA campus-actuals data were found to be mislabeled (the campus ID numbers are correct; the names attached to them are not — for example, campus 101912310 is labeled "Horn El" but is actually Houston MSTC). This was caught and documented, but not corrected campus-by-campus in the underlying data file; anyone using data/hisd_campus_actuals.csv by name rather than by ID should check against sources/tea_peims/campus/allcamp_actual_2025_101912.txt.

Known unreconciled inconsistencies (not corrected; flagged for a future pass)

These appear across the report pages with full citations; they are collected here for visibility.

  • HISD's FY2025 recapture payment is reported as five different figures in five different documents. Second edition: three match specific TEA report runs, the audited $56.9M is within $1.4M of the run current at year-end close but matches none exactly, and HISD's own $44.5M budget figure matches no TEA run (see below).
  • HISD's tax year 2025 tax rate is reported as both $0.8783 and $0.8489 (a May 2026 bond official statement). Mostly explained in the Oct. 12 recheck: $0.8489 was the rate recommended on Sept. 11, 2025, and the $0.0294 difference is exactly the one-year "disaster pennies" the board added on Oct. 15. TEA's 2025-26 Summary of Finances uses the higher rate. Why the May 2026 disclosure shows the September rate is still open.
  • The ACFR's own statistical tables disagree with its own notes on the FY2025 tax rate (one table shows 0.7116/0.8783; the notes and levy table show 0.7016/0.8683) — the table row appears to have been mistakenly populated with the following year's rate.
  • A subsequent-event note in the FY2025 ACFR describes a $20.555 million tax-note issuance, which conflicts with the $114.39 million par value in that issuance's own official statement; the ACFR note appears to have been copied from an earlier, smaller issuance.

Second edition (2026-10-08): history back to FY2019

What was added. Every series was extended back to school year 2018-19 (HISD's FY2019): TEA's Summary of Finances for every school year from 2018-19 through 2026-27, plus TEA's full list of report runs for each year (390 runs); HISD's audited ACFRs for FY2019-FY2021; adopted budgets for FY2019-FY2024; TEA PEIMS actuals (district, statewide, peers and campuses), TAPR staffing, FIRST ratings and Comptroller property values from 2018-19; and federal single-audit data from FY2018. A new report section (Section 9) presents the seven-year view.

How the TEA reports were obtained. The Summary of Finances dashboard is a JavaScript application. It was driven with a browser automation tool configured to reach only TEA's report domain (tealprod.tea.state.tx.us), using the system browser. The report files were downloaded from inside the browser session, because the report URL returns an empty viewer page outside a dashboard session; the server also needs about ten seconds to generate each report. Each year's run list was read from the dashboard page and parsed by scripts/sof_runs_from_snapshot.py.

Checks. - scripts/sof_extract.py now parses all nine years. Its reconciliation checks pass for every year (state aid equals the sum of its funds; Available School Fund equals the per-capita distribution), and for each year from 2019-20 the recapture printed in the report equals TEA's dashboard figure for the same run. - The research pass that extended the TEA PEIMS, TAPR, FIRST and campus data changed several scripts. A regression check against the first edition's committed data files found every previously published value unchanged; the files only gained rows, years or columns. That check was repeated independently (an initial repeat reported false differences because it keyed on too few columns; corrected, it found zero changed values). The headline 2024-25 figures still reproduce: NES $14,288 vs non-NES $9,193 per student; HISD all-funds operating spending $14,147 and revenue $12,666 per student. - Headline audited FY2019 figures were re-verified against the ACFR text: General Fund revenue $2,200,600,360, expenditures $1,991,206,129, recapture expense $265,231,840.

Findings that resolved first-edition open items. - HISD's audited recapture expense tracks TEA's Near-Final estimate current at fiscal-year close (September 30), not TEA's final settlement, which can come years later. In six of seven years (FY2019-FY2025) the two are within $2.4M; FY2020 is the exception. This explains why the ACFR's expense line and its own statistical table disagree. - Of the five FY2025 recapture figures: about $65M matches TEA's preliminary runs of Sept. 2024–Feb. 2025; $55.5M matches the Near-Final runs of Sept. 19 and Oct. 17, 2025; the audited $56.9M is within $1.4M of the run current at Sept. 30, 2025; $49.1M is the final settlement. HISD's $44.5M budget figure matches no TEA run. - HISD's FY2024 budget carried $326.5M of recapture because it was adopted in June 2023, when TEA's first projection for 2023-24 was $335.2M; the November 2023 tax-compression law reduced it to $0.

Corrections made in the second edition. 1. SY2026-27 figures now use TEA's Oct. 5, 2026 run (47318), which superseded the Sept. 14 run used in the first edition: total state aid $281.9M (was $274.3M). The per-student entitlement ($9,899) and the $156.3M counterfactual recapture are unchanged. 2. The summary page said the General Fund had run deficits "every year since the state's 2023 takeover" while showing FY2023, which ended one month after the takeover began. It now says "since FY2023" and notes that FY2024 and FY2025 are the first full years under the state-appointed board. 3. During the write-up, a working note had FY2022's audited recapture expense as $186.5M (a figure from a different ACFR table); the expense line is $184.5M. Corrected before publication. 4. Six scripts (build_peer_comparison.py, build_tapr_peers.py, parse_campus_actuals.py, analyze_campus_nes.py, fetch_first.py, fetch_tapr_dd.py) had the author's private working-folder path hard-coded, so they would not run from a clone of this repository. They now resolve paths relative to the script. Re-running the four offline builders produced byte-identical data files.

Update (2026-10-08): budgeted and projected years

What was added. Every report table now runs through FY2027. FY2025 is the latest audited year, so FY2026 and FY2027 are filled from HISD's adopted and amended budgets and its in-year forecasts, TEA's Summary of Finances estimates, and TEA's enrollment counts. Each figure carries a label saying which (see the note on the summary page). New sources:

  • TEA's "Budgeted Financial Data" for HISD and the state total, 2019-20 to 2025-26: the adopted budgets districts file with TEA.
  • TEA's 2025-26 campus budget report for HISD, which also carries TEA's fall 2025 enrollment counts.

HISD's budget page and tax-rate page were re-checked and had nothing new since 2026-10-06. The references page was reorganized by fiscal year, with a status for every document.

Checks. - scripts/parse_tea_budget_reports.py runs 58 checks, all passing. Fifty-six check that revenue sources sum to total operating revenue, and spending by function to spending by object, for HISD and the state in every year. Two check that HISD's 2025-26 filing ties to its own adopted budget: revenue exactly, spending within $198. - scripts/build_gf_fy2019_2027.py holds every budget figure with its citation. It checks that each budget's revenue, spending and other sources add up to its change in fund balance, and that the beginning balance plus that change equals the ending balance. Audited and budgeted function lines must sum to their totals, and revenue sources to total revenue. All pass. - scripts/parse_campus_budget.py reuses the campus-actuals parser and NES exclusions unchanged. Every campus parsed with a total. Campus counts sum to 167,786 against TEA's district count of 167,793. - The new scripts were run from the public repository copy and produced byte-identical data files. - A TEA budget report is not an actual. Its "all funds" columns cover only the adopted funds (no federal grants), so the report pages compare it only with General Fund actuals.

Choices a reviewer may question. - FY2026 per-student figures divide by TEA's fall count (167,793), while FY2019-FY2025 use HISD's ACFR enrollment, which ran 0.3-0.4% higher in FY2023-FY2025. The difference is noted under the tables. - The FY2026 column uses the final amended budget (June 11, 2026) rather than the adopted one. It is HISD's latest full statement of the year, and it matches HISD's own May 2026 year-end forecast ($730.0M ending balance). The adopted figures are shown alongside in Section 5. - HISD's FY2027 average-salary projection ($83,338) uses HISD's own measure, which gives $68,527 for 2022-23 where TEA's TAPR gives $64,954. It is shown in a separate column, not as a continuation of the TAPR series.

Correction made in this update. Section 3's staffing table had placed HISD's FY2027 salary projection in the FY2026 column. It now has its own FY2027 column.

Independent review and response (2026-10-08)

The review. A separate reviewer, commissioned by the person directing this work, examined the report and its added FY2026-FY2027 material on October 8, 2026. The reviewer re-extracted the audited history, PEIMS totals, campus data, all nine Summary of Finances reports and the new budget files directly from the source PDFs. Its script ran 6,805 arithmetic and data comparisons, all passing. The reviewer also checked official web sources. The review states that it is an analytical review of documents, not an audit of transactions. It confirmed the core findings: the audited General Fund deterioration, the reserve decline, the federal-relief cliff, leadership-spending growth, and a large descriptive NES spending gap. It proposed 32 corrections.

The response. Every proposed correction was checked against the source documents before being accepted. The response addresses 33 items (one of the 32 was split in two): 21 were accepted as proposed and 12 in part; none was rejected outright. Where a correction was accepted only in part, it was usually because the published pages had already been corrected and the error survived only in the private working notes. The reserve-policy item is the exception, below. The most important corrections:

  1. Formula funding per attending student was understated. The earlier figures ($9,225, $9,964, $9,899) added Tier One's full entitlement to only Tier Two's state aid, leaving out the part of Tier Two paid by HISD's own enrichment taxes. Corrected: $9,894 (2024-25), $11,064 (2025-26), $10,988 (2026-27 preliminary). scripts/sof_extract.py now extracts the full Tier Two entitlement for all nine years and checks it against Tier Two state aid; it reproduces the reviewer's figures exactly. The correction strengthens the finding that per-student formula funding did not fall. Part of the 2025-26 rise, however, came from a one-year local tax (see "Recheck" below).
  2. $29.5 million of 2025-26 state aid was misclassified as tax-relief backfill. It is aid under TEC §48.257(b-1), which the Summary of Finances computes from the Teacher Retention Allotment. Tax-relief aid is about $113 million of the $141.6 million increase. The claim that the increase is "not new money" was withdrawn.
  3. General Fund instruction growth now appears beside all-funds instruction. General Fund instruction rose 21.7% from FY2023 to FY2025, but all-funds instruction rose 0.4% as grant-funded instruction fell. Leadership spending rose across all funds (school leadership 43.9%, instructional leadership 91.8%).
  4. NES comparisons were separated from HISD's online school. Texas Connections Academy, counted with non-NES campuses, grew 31% in 2025-26. The corrected same-campus enrollment change is −9.7% for NES and −4.0% for other brick-and-mortar campuses, not −1.1%. An attribution of part of the NES spending gap to NES itself was removed.
  5. The FY2027 debt-service comparison mixed funds. The $438.5 million is $418.2 million of I&S-tax debt and $20.3 million of maintenance-tax notes. The adopted Debt Service Fund budget uses $14.0 million of fund balance.
  6. Control findings were overstated. The FY2025 equipment finding covers assets "unaccounted for" or excluded from the counts, with no questioned costs. The 2024 ESSER audit found a documentation gap, with "no indication" of missing the 20% learning-loss requirement.
  7. Smaller corrections:
    • a double subtraction of recapture in the peer-revenue notes;
    • the "local plus state" figure, mislabeled;
    • the federal scope of TEA budget data;
    • spliced NES budget totals;
    • overlapping FY2027 cut categories;
    • property-sale sensitivities for FY2026 and FY2027;
    • the $15 million projected closure saving;
    • contracted-charter amounts that are budgeted, not actual;
    • variable-rate debt and rating wording;
    • four statute numbers parsed as dollar amounts in data/sof_key_figures.csv.

Accepted in part: the reserve policy. The reviewer showed that board policy CE(LOCAL) sets a goal, not a requirement, measured on the adopted budget. The report's 2.48 months is not that measure; the policy text is now in sources/budget/HISD_board_agenda_2025-09-11.pdf (packet pp.278-279). The report still says the June 2025 unassigned balance ($464.8 million) fell short of the goal, because it does on every basis available: HISD's own workshop figure ($548 million), the earlier workshop figure (~$525 million), and the policy formula applied to the adopted budget (~$499 million). The shortfall is now given as a range, with each basis named.

Pages with substantive corrections carry a dated correction note.

Press copies removed. Since the first edition, the public repository had included text extractions of five news articles used to compile the NES campus list (sources/tea_peims/nes/). That contradicted this project's practice of citing press by URL only. They were removed on 2026-10-08 and are kept in the private working archive; earlier copies remain in the repository's history.

Recheck (2026-10-12)

The whole report was rechecked against its sources on October 12, 2026, four days after the review. The recheck itself was done offline; the downloads it called for were made later the same day (see "Sources added" at the end of this section). It found these errors, all corrected:

  1. The 2025-26 Tier Two increase was misexplained. Section 4 attributed it to the state's falling compressed rate and called HISD's October 2025 "disaster pennies" one cent. The disaster rate was $0.0294 per $100 (HISD's spokesperson, quoted by KPRC on Oct. 15, 2025). It counts as Tier Two enrichment, which is why the enrichment rate rose from 5 to 8 cents (2025-26 Summary of Finances, line 10). By law it lasts one year. Most of the 2025-26 Tier Two rise is therefore a one-year item.
  2. New information: HISD adopted its tax-year 2026 rate on October 8, 2026. The rate is $0.8421, with M&O $0.6754 (HISD Legistar file 26-346, status Passed). The report had called it proposed. The adopted rate drops the disaster pennies, while TEA's 2026-27 preliminary report assumed they would stay (M&O $0.7048). The 2026-27 formula entitlement will therefore come in below TEA's preliminary figure; a proportional estimate gives roughly $10,500 per attending student, labeled as our approximation. TEA's newest run (47363, October 9, the day after adoption) still assumes $0.7048. The report now uses that run: $10,992 per attending student and $283.2 million of state aid, against $10,988 and $281.9 million in run 47318.
  3. Tax-year 2025 rate disagreement mostly explained (Section 8): the bond statement's $0.8489 is the rate recommended before the disaster pennies were added.
  4. Suburban spending comparison: the summary said HISD spends "21-26%" more per student than Houston-area suburban districts. Katy ISD was omitted; the range is 15-26%.
  5. Staffing table labels: total employees for FY2023-FY2025 come from the ACFR statistical section, not TEA's TAPR as the table implied.
  6. TIRZ: Sections 3 and 7 gave two different sets of TIRZ figures without saying why. Both are now labeled ($75.5M/$56.1M General Fund expenditure versus $48.4M/$37.9M of taxes paid into the zones). The pair is listed in Section 8 as unreconciled.
  7. Smaller corrections:
    • The FY2027 enrollment loss is a $50 million revenue reduction in HISD's budget, not specifically state aid.
    • Formula funding stayed flat for six years, not five.
    • Leaving out the online school, enrollment at never-NES campuses fell 10%, not 9%, from 2018-19 to 2024-25.
    • The summary still described four of the five FY2025 recapture figures as TEA snapshots, after Section 8 had been narrowed to three.

Sources added on October 12.

  • The tax-rate adoption record and its attachments: the resolution, the taxpayer impact statement and the board presentation (sources/budget/legistar_26-346_*).
  • The presentation attributes the $0.0362 M&O cut to "tax compression and disaster pennies".
  • It also dates the FY2027 budget adoption to June 11, 2026, which conflicts with HISD's own adopted-budget posting (June 25). That conflict is listed in Section 8.
  • TEA's newest 2026-27 Summary of Finances: run 47363, October 9. It still assumes the old M&O rate. The report now uses it: $10,992 per attending student and $283.2 million of state aid.
  • CPI-U data: BLS figures via FRED, for the inflation comparisons.

Not yet checked: HISD's EMMA filings after May 2026. EMMA's lists need a browser, and this project's browser tool is limited to TEA's site.

How to re-verify

What this investigation does not cover

  • HISD budget books for FY2020 and FY2024 (not found on HISD's site or in the Internet Archive; the one-page adopted schedules were used).
  • TEA's separate per-year "Cost of Recapture" reports, and every run's PDF other than the latest of each year (only the run list was recorded).

  • How HISD settled the $80.1M by which its FY2019 recapture expense exceeded TEA's final settlement.

  • FY2026 audited actuals (not yet published by HISD as of 2026-10-08) and 2025-26/2026-27 PEIMS actuals (not yet published by TEA). FY2026 and FY2027 are shown from budgets, HISD forecasts and TEA estimates, labeled as such.

  • The FY2027 adopted budget book's object-level and school-by-school detail (HISD had not yet published it; only the function-level schedule and board workshop decks were available).
  • Independent verification of HISD's own-reported attendance (ADA/WADA) data against a source outside TEA.
  • A reconciliation of the recapture discrepancies listed above, and of the two TIRZ measures (Section 8).
  • Transaction-level evidence the reviewer recommended obtaining:
    • HISD's general ledger and monthly FY2026 close;
    • position-level payroll histories showing which grant-funded positions moved to the General Fund;
    • the recapture payable/receivable roll-forward;
    • management's certified reserve-policy calculation;
    • inventory reconciliations;
    • executed financing documents for the CTE center.