Historical district verification and title linkage

October 4 correction: workbook dates and report wording

Three month descriptions in earlier report versions were wrong. The extracted CSV dates were correct. A direct check of all 3,362 workbook rows now verifies each title, district, status and date against its source cell. The workbook defines numeric dates as approximate months; the first day in an Excel serial is not an exact restriction date.

The sample classifications remain 13 supported school-topic records, four other records and 19 unresolved. No model-ready records or new restriction onsets are added. Independent review has been prepared but has not been completed.

Earlier pass: original article recovered and Davis vote checked

Version 12. The unchanged 36-record sample now contains 13 supported school-topic records, four other records and 19 unresolved records. Nine support an original district candidate. One primary reinstatement decision is added; no new restriction onset or duration is accepted.

What changed in this follow-up

All 20 unresolved records received an access or corroboration follow-up. One original body was recovered: CNN’s Davis School District story. Nineteen selected records remain unresolved. Readable coverage elsewhere was retained as corroboration, without assuming it was identical to an inaccessible selected article.

Davis, Utah: a documented reinstatement decision

The district’s June 20, 2023 minutes record a 7–0 vote to reinstate the King James Bible at all school levels where it had been removed. The minutes describe eight copies, not eight ban instances. Public comments included residents and a school librarian. This establishes a board decision alongside public participation; it does not establish that opposition caused the decision.

The posted minutes are version 1 with blank signature fields; later approval was not separately checked. The record does not establish the initial removal date or when each copy returned to circulation. No duration is calculated. PEN row 3196 lists The Bible in Davis with an April 2023 month; CNN’s publication date is June 21. These dates describe different fields and remain separate.

District minutes, page 2 · CNN article. The official minutes and checksum are included in the research ZIP. The article text is not redistributed.

Earlier pass: fixed 36-record validation sample

How the sample was selected

1. Freeze the current 1,054-record review ledger and record its SHA-256 checksum.
2. Group records by the month of their GDELT timestamp. This timestamp is not assumed to be the publication or restriction date.
3. Order each group by SHA-256 of a fixed seed and record ID, then select the first three records. Retain previous reviews, duplicate-looking stories and unavailable pages. Do not substitute easier records.
4. Read accessible article bodies, check geography against the official 2022–2023 NCES files, and classify each proposed district link separately.
5. Compare selected named titles with the preserved PEN index. Record differences in date, stage and scope without resolving them by assumption.

The fixed seed is banned-books-month-validation-v11-20261002. Monthly frame counts range from 29 to 247, so the equal-per-month sample is not proportional to news volume. The record-level CSV contains inclusion fractions and design weights, but this report gives sample counts only. No national accuracy estimate, recall estimate or confidence interval is presented. Independent human review remains pending.

Errors that a district-name search can make

What the PEN comparison established

Thirty-two selected title probes produced twelve title–district correspondences. They are not twelve newly verified restriction events. Dearborn and Central Bucks reporting concerns stages that differ from the PEN entries. Portage lacks a verified onset date. Madison County, Mississippi, and Calvert County are absent from this preserved index; that absence cannot be recorded as zero bans.

For Indian River, the article reports a March removal, while PEN records February. The Florida Department of Education’s annual report lists the same graphic adaptation and author under Indian River’s four removals, but supplies no month. The title and district have corroboration; the date conflict remains unresolved. The article about The Hill We Climb refers to an April committee meeting, while its PEN entry records February. Those dates also remain separate.

Florida 2022–2023 objection report, page 9. The full official PDF, its checksum, and the row-level PEN comparisons are included in the research package.

Monthly review coverage

The complete 36-row review, source links and decisions are in GDELT_Validation_Sample.html, included with this report in the ZIP.

Earlier pass: Newtown and Ludlow outcomes verified

Newtown: retention, with a process for individual concerns

The June 1, 2023 minutes record a motion for Flamer and Blankets to remain in the high-school library while the administration develops a process for individual parent or guardian concerns. The motion passes unanimously. The discussion clarifies that the books would remain available even before that process existed.

The downloaded copy is marked pending approval. Its attendance list is internally inconsistent: the secretary is marked absent but appears in the discussion and seconds the motion; another name appears twice. The review therefore records unanimous, without converting the roster into a 5–0 tally. A later official record indexes approval of June 1 minutes, but a final corrected June 1 copy was not acquired. Earlier access interruptions remain a separate question.

Ludlow: distinguish the two motions

The signed June 13 minutes record a procedural motion about legal concerns, with two yes and three no responses. The later motion to adopt the proposed IJL Library Resources policy received no second, so no adoption vote occurred. The June 27 minutes approve the earlier minutes in a 5–0 consent-calendar vote. These are three different actions; neither numeric tally is a vote to ban books.

PEN lookup and the remaining date discrepancy

The preserved 2022–2023 index contains no Connecticut rows. Its only Massachusetts district is Abington, not Ludlow. No Newtown title–district link or Ludlow title-specific policy link is accepted. Missing index coverage is not a zero-ban observation.

Hillsborough's March 28 board decision still differs from PEN's April 2023 field. The preserved workbook has no row-specific source hyperlink or comments explaining the entry. Reporting month, later implementation or an index error are possible explanations, not findings. Both dates and both scope labels remain unchanged; no timing estimate uses this provisional link.

Steps completed

1. Follow the public district archive links into the relevant school-year and meeting folders.
2. Download the Newtown June 1 and Ludlow June 13/27 PDFs; record original hashes, acquisition times and the navigation path.
3. Read the minutes, use OCR for scanned Ludlow pages, and inspect the decision and approval pages visually.
4. Keep draft status, attendance discrepancies, procedural votes and later approval separate.
5. Check the preserved PEN index by state and district before searching titles. Retain missing matches and conflicts explicitly.
6. Regenerate the current 1,054-row ledger, six-case outcome table and 24-record chronology offline.

What this adds to RQ3

What happened when a district reviewed a challenged book with input from librarians and local residents?

The selected cases show different outcomes following review or public discussion: retention, rejection of a proposed policy, a proposal that never reached an adoption vote, and removal. These records help define outcome categories. They cannot show how common any outcome is or whether public opposition caused it. The cases were selected for review and have different institutions, proposals and evidence types.

Current files and next evidence

Use current_record_review_v10.csv for the latest all-record ledger and selected_case_outcomes_v10.csv for the six cases. The chronology, PEN lookup, source register, access log and reusable SVG accompany them. Older versioned tables remain audit history. The PDF excerpts preserve decision context while omitting unnecessary public-comment details; their transformations and original hashes are recorded.

Next, obtain PEN's underlying source for the Hillsborough April field and a corrected/final Newtown minutes copy, then extend the review to a defined larger sample. Independent checking and known access dates are still needed before duration or timing models. No new paid query, model, forecast or website publication occurred.

• newtown_june1_minutes_v10 — Original pages 1-2: motion and unanimous outcome.
• ludlow_june13_minutes_v10 — Original page 3, POLICY FILE: IJL LIBRARY RESOURCES.
• ludlow_june27_approval_v10 — Original page 1, CONSENT CALENDAR.

Earlier pass: Hillsborough vote verified; PEN date conflict found

The board webcast's opening identifies March 28, 2023. At the end of the meeting, the final motion seeks removal of This Book Is Gay from all Hillsborough middle schools and passes 4–3. Public comment includes a media specialist advocating retention. The Florida Department of Education's 2022–2023 report independently lists Hillsborough, the same title and Juno Dawson, with three objections and one removal.

What matched, and what did not

The preserved PEN workbook has one Hillsborough entry: This Book Is Gay, author recorded as Dawson, Juno, on row 658 of Sorted by Author & Title. The explicit author-name equivalence and reviewed district identity support the title and district identification. The date cell is formatted mmmm yyyy and supports April 2023 only. This differs from the March 28 board decision. Whether PEN recorded a later implementation, a reporting month or another date convention is unresolved. Converting its Excel serial does not supply an exact ban day, and April must not be replaced with March without evidence.

What stays unresolved

• PEN labels the entry Banned from Libraries and Classrooms; the transcript supports a middle-school removal order. Both source labels are retained. No high-school removal is inferred.
• The transcript has conflicting header metadata: its HTML title references December 2022 and its body heading gives 4 p.m. The spoken opening, linked meeting page and prior district notice agree on March 28, 2023. No exact decision time is assigned.
• This evidence includes caption text, not approved minutes or an independently reviewed video. Independent human review remains pending.
• The physical date when access first ended, any later reversal and restriction duration remain unknown. PEN's origin field remains Unclear.

How to reproduce this step

1. Run the earlier workbook-to-CSV check and verify the unchanged source hashes.
2. Filter the 3,362 PEN rows to Florida and the reviewed Hillsborough district name; this returns one row.
3. Compare the title and the explicitly reviewed author-name forms; inspect the original date-cell number format.
4. Compare the resulting April interval with the March 28 board decision day. Flag the discrepancy and keep the event link provisional. Preserve both scope labels.
5. Export the link, source register, separate date fields and current six-case outcome table. Leave model eligibility off.

Relevance to the research question

RQ3: What happened when a district reviewed a challenged book with input from librarians and local residents?

This case adds a removal outcome to the three previously confirmed retention or policy-rejection examples. The presence of public objections and professional input does not by itself predict the outcome. These selected cases are not a representative comparison and cannot estimate the effect of resistance.

Files and evidence

The latest 1,054-row review table is hillsborough_record_review.csv. Current selected outcomes are in selected_primary_outcomes_current.csv; Newtown and Ludlow still need decision verification. This pass also provides hillsborough_pen_link.csv, hillsborough_chronology.csv, hillsborough_source_register.csv, hillsborough_evidence_chronology.svg and hillsborough_review_summary.json. Earlier versioned fields remain as audit history.

• hillsborough_webcast_transcript_v9: Opening call to order; final main motion and announced result immediately before closing discussion.
• hillsborough_webcast_meeting_v9: Meeting heading, agenda item 1.01 and Transcript link.
• fldoe_objections_2022_23_v9: PDF page 1: report scope; PDF page 9: Hillsborough row, This Book is Gay / Juno Dawson, 3 objections and 1 removal.

The ZIP includes the official state PDF and short transcript/archive review notes with original-file hashes. Re-running the package reproduces the frozen judgments; it does not independently adjudicate them. No new BigQuery query, forecast, causal model or website publication was performed.

Earlier pass: selected-case PEN lookup and chronology

Old Rochester retained Felix Ever After and Push; Forsyth retained Endlessly Ever After; Mead declined the proposed library-policy revision. A news item about a challenge or retention can be relevant to RQ3 without being a PEN ban instance.

What the lookup found

The district-row counts describe this index only. They are not estimates of all bans, and missing matches must not become zero-event observations. Five national entries for Felix Ever After and 20 for Push belong to other state/district combinations. Joining on title alone would assign those events to the wrong place.

Dates and evidence

The chronology keeps three clocks separate: the official decision day, the date displayed on the selected news page, and the GDELT record timestamp. Where a decision and article share a calendar date, their order within that day is unknown. The reviewed articles are not established as the first coverage of these cases. News-page bodies were retrieved retrospectively; the original text available on the displayed publication date has not been verified.

Old Rochester's September 6, 2023 approval of June minutes falls outside the July 2022–June 2023 study window and is included only as later documentary corroboration. Mead's September 26 approval likewise confirms the earlier minutes, not a second policy decision. No approval date is used as a restriction date.

Reproducible steps

1. Use the three confirmed cases from version 7; leave the other three outcomes unresolved.
2. Validate the preserved workbook and all 3,362 derivative rows using the earlier workbook check in the pipeline.
3. Compare state and explicitly reviewed district names before comparing titles. Normalize only Unicode, case and whitespace; do not silently fuzzy-match.
4. Record every partial match rejected for another district or another title. No candidate reached author/episode acceptance in this pass.
5. Build the dated evidence table from confirmed decision dates, reviewed publication labels and preserved GDELT timestamps. Keep unknown access dates blank.

selected_case_pen_scope.csv records the search scope; selected_decision_pen_links.csv records three title searches and one policy case; selected_pen_candidate_exclusions.csv records 29 rejected partial matches; selected_case_timeline.csv contains 11 chronology entries; selected_case_chronologies.svg is the reusable figure. The latest all-record table is case_timeline_record_review.csv.

RQ3: What happened when a district reviewed a challenged book with input from librarians and local residents?

These selected examples document procedures and outcomes. They cannot estimate whether librarian involvement or public objections caused retention, or compare treated and untreated districts. The study still needs additional outcome records, independent review, and matched observations before such comparisons. All model-eligibility flags remain no. No new BigQuery query or website publication occurred.

Earlier pass: primary records for six selected outcomes

RQ3: What happened when a district reviewed a challenged book with input from librarians and local residents?

Decision evidence

Corrections and boundaries

Mead's decision date is now September 12, 2022, replacing the earlier month-only precision in the primary fields. Its minutes record four opposing votes and one abstention on the library-policy revision. Old Rochester's minutes record two separate 7–1 retention votes on June 21, 2023. Forsyth's district register identifies April 18, 2023 as the retention date for Endlessly Ever After, but does not verify the news account's participation or attendance details.

Old Rochester's June minutes were approved September 6; Mead's September 12 minutes were approved September 26. The Old Rochester file linked under June 21 is an agenda packet with older minutes, so the actual June decision was checked in the September packet. The distinction matters: a proposed action is not evidence of an adopted decision.

For Newtown, indexed draft-minute text and a later approval record were located, but the full decision text was not acquired. A vote approving minutes must not be confused with the vote on the books. Hillsborough's district notice confirms a scheduled appeal meeting, not its outcome. Ludlow's corrected news account still describes a proposal that was not put to a vote; the official archive has not yet corroborated it.

These records do not establish when a book first became unavailable, how long a restriction lasted, or whether public participation changed the decision. A retention decision does not itself prove that an implemented ban was reversed. Current retrospective sources cannot be treated as information available to a historical forecasting model.

How to reproduce and inspect this pass

1. Read the six frozen reviews in primary_outcome_review_v7.json.
2. Check the source URLs, original PDF page numbers and SHA-256 values in primary_source_register.csv. The package includes selected official minute pages and short webpage review notes. Notes are reviewer summaries, not full source copies.
3. Read primary_access_log_v7.json for unsuccessful and partial retrievals. A missing document is an evidence gap, not proof that no action occurred.
4. Run python run_all.py from the extracted package. The final script is review/build_primary_review.py.
5. Use primary_record_review.csv as the latest combined table. Earlier reported fields remain unchanged for audit history. The new primary fields apply only to the six selected records; linked story copies were not automatically promoted.

The package still contains 1,054 candidate records. The 382 uncached records remain substantively unreviewed, and no record is eligible for a national timing model. Independent human review remains pending. No BigQuery job, paid extraction, website publication, causal estimate or Ohio forecast was added.

Remaining work

Retrieve the outstanding decision records, independently check the three confirmed cases, and then link their titles and stages to PEN records. Descriptive case timelines can follow that linkage; they must not treat policy proposals or retained books as newly imposed bans.

Earlier pass: district identities and selected news outcomes

The 162 resolved article-to-entity links cover 90 distinct school systems. 4 articles discuss more than one district, so the link total exceeds the article count. Earlier content classifications remain intact: identifying the district does not independently validate the article, date, or outcome.

Names that needed care

Hempfield Area and Hempfield are separate Pennsylvania districts. Maine RSU 24 is distinct from RSU 40/MSAD 40. Orange Unified in California is distinct from Orange County schools in Florida. The Spring Lake High School and Old Rochester Regional High School names were checked against the NCES school directory. School systems, administrative offices and charter agencies were kept separate.

Four records retain unresolved identities: Madison County Schools, Marion County, Brookfield and the caption-only Hempfield record. Similar names, reporter biographies and unsupported state assumptions were not accepted as identity evidence.

What happened after review: six selected reports

These cases were chosen to examine different procedural outcomes relevant to RQ3. They are not a representative sample and cannot establish whether public opposition caused a decision. Current cached news text is the source; primary district documents have not yet been checked for these six outcomes.

Correction: Ludlow declined to vote on the proposal. An earlier headline described rejection, but the article explicitly corrects it. Retaining a challenged book is also different from reversing an implemented ban. Hillsborough's reported board order does not establish the day books physically became unavailable.

Dates are separated by precision. Mead has only a month-level contextual date. Newtown and Hillsborough dates are inferred from weekday language anchored to the displayed publication date; they are not exact primary-record dates. First inaccessibility and duration are blank for every new case. No GDELT timestamp is substituted for a restriction date.

Repeated reporting

Four pairs share the same publisher's article identifier despite different headline slugs: Central Bucks/WFMZ, Denton, Old Rochester and the Lafayette public-library commentary. Three St. Lucie iHeart records share the byline, date and exactly matching extracted article body. A Dearborn roundup shares a segment with a standalone article; the whole roundup is not treated as a duplicate. Every original record is retained.

Next evidence step

The resolved identities can guide full-context article review and retrieval of district minutes or decisions. The six selected outcomes need primary-source corroboration. The 382 uncached records remain substantively unreviewed. These results do not support a national timing model, a causal effect of opposition, or an Ohio forecast.

RQ3: What happened when a district reviewed a challenged book with input from librarians and local residents?

Use identity_record_review.csv for the current combined table; identity_article_district_links.csv for one-to-many identity links; selected_reported_outcomes.csv for dates and outcome limits; and repeat_relationships.csv for story relationships. identity_validation_worklist.csv lists the remaining work for all 160 leads. Earlier report sections and output files are retained as an audit history.

The package reproduces these tables from frozen assistant decisions. Independent human review remains pending. No additional BigQuery job, paid extraction or public website update occurred in this pass.

Earlier pass: all remaining cached records assessed

The earlier 206 supported district-topic records remain separate from 160 new passage-level leads. This pass creates no confirmed NCES links and no new ban counts. News about a proposal, lawsuit, retention decision or public meeting is relevant to the topic without establishing a restriction.

Results of this pass

What the evidence can and cannot show

Several passages describe rejected proposals or books retained after review, including Mead, Forsyth County, Newtown, Ludlow and Old Rochester. Others describe requests, campaign promises or committee recommendations. These stages need separate coding. They cannot all be counted as bans.

Current pages for Hillsborough, Newtown, Orange Unified and other districts include outcomes or updates that differ from the earlier captured headline. The GDELT timestamp is a news-record timestamp. It does not establish when student access changed, or that the current page text was available then. Models must use only information available before a forecast period.

Caption-only matches remain weaker leads. District names such as Madison, Hempfield and Rochester require exact geographic verification. A library located in a city is not automatically part of the city's school system. No person’s beliefs or motives are inferred from district characteristics.

Method and audit trail

The assistant read selected passages from 278 cached records. A deterministic extractor selected unique long lines and filtered common boilerplate; it can still select captions or miss relevant text. Per-record notes, source URLs, cache hashes, passage lengths and passage hashes are frozen in the input JSON. The report does not claim a complete read of every article. Two passages visibly contained encoded text despite the earlier quality heuristic missing them; both remain unresolved. The 81 other encoded pages were assessed for access only and were not decoded.

There are 6 groups with identical selected passages. These are possible repeats to check against full article provenance; they are not newly verified story families. No records were collapsed. The previous 28 verified/comparison-supported families remain limited to their earlier review batch.

What remains

382 records still have no usable cached text and have not received substantive review. Their recorded access outcomes are retained. The new 160 district-topic leads need full-context review, exact NCES matching, repetition checks and defensible dates. Unresolved excerpts need accessible source evidence; missing evidence must not become zero bans.

RQ3: What happened when a district reviewed a challenged book with input from librarians and local residents?

District matching will use the NCES directory and manual review where necessary. Every event must have a school year; months require source support or a precision flag. Exact and probable duplicates will be documented before combination. Statewide restrictions remain separate. Counts must reconcile with PEN definitions, and missingness must be reported by state and school year. The evidence assembled so far does not estimate the effect of public opposition.

Files and reproduction

Use cache_triage_record_review.csv for all 1,054 current dispositions, cache_triage_decisions.csv for this pass, and validation_worklist.csv for next actions. Earlier files beginning latest_ represent the preceding full-content pass and are retained for comparison. Run python run_all.py to reproduce the package offline. Frozen judgments are inputs, not independent validation. Full third-party article text is excluded.

No additional BigQuery job or paid extraction was run. The public website has not been changed by this review.

Earlier pass: content review and repeated coverage

Review actor: AI assistant. Independent human coding remains pending. The older input field named human_review_status is retained for reproducibility; its name must not be read as evidence that an independent person reviewed those records. The latest table uses content_review_status.

What was checked

One representative was read for each of 29 story profiles. There are 122 additional records with exactly matching extracted bodies after whitespace normalization. A further 21 changed versions were compared, and one unreadable version remained unresolved. Hashes, extraction boundaries and the decision basis are preserved. A matching headline alone never transfers a content decision. Full third-party article text is excluded from this package.

Repeated reports and geographic errors

The NPR Forsyth County story has 82 retrieved copies: the representative, 80 exact-body copies, and one copy with two sponsorship insertions. These are distribution records for a single story. The two CNN Gorman versions share a broader source family. The 28 reviewed families include 22 with supported district-topic coverage and cover only this selected batch; they are not a count of independent stories in the annual dataset.

The Knox County headline describes Maine RSU 40/MSAD 40 and Medomak Valley High School. It must not be assigned to a Tennessee or Kentucky district with a similar county name. Other stories mention places where people were interviewed or where publishers operate. Those places are not automatically restriction locations. Additional districts mentioned in an article remain separate review leads.

Why publication timing still needs care

• The Conroe page was captured as a preview on April 17, 2023 but now reports a decision in an April 19 update.
• The Elmbrook reprint displays November 28, 2022, after its November 8 GDELT record.
• The St. Lucie page now reports a board outcome although its GDELT headline described the removal request.
• The Keller page includes an update after the original August 16 publication.

Current article bodies cannot be assumed to have existed at the first GDELT timestamp. No forecast or lead-lag model uses these records. Publication fields, update notes and restriction stages are retained separately. The previous 20-row selected episode audit remains an exploratory source comparison, not a panel of known restriction onset dates.

Remaining work

359 unreviewed records have cached text, while 382 have no cached text. The latter already have access outcomes in the URL audit. An encoding-pattern check flags 82 cached pages for further access review; it does not decode restricted bodies or establish relevance. District enrollment is now joined for 51 reviewed identities, while agency eligibility remains pending.

The next content review should cover the remaining accessible stories and seek public primary sources for unresolved cases. The selected batch does not support a precision estimate, a national news-attention measure, or an estimate of the effect of public opposition. No new BigQuery job, paid service or website update was performed.

RQ3: What happened when a district reviewed a challenged book with input from librarians and local residents?

Read and reproduce

Start with latest_record_review.csv, reviewed_story_families.csv and remaining_content_review_queue.csv. Run python run_all.py in the unzipped package to rebuild outputs offline. The frozen judgments are inputs: reproduction does not turn them into independently validated evidence. Optional cached-page extraction requires separately obtained page text.

Reviewed representatives

Earlier pass: URL access audit and selected episodes

The annual source remains 15,767 GDELT records. The alias screen retains 1,054 candidate records. The latest manual batch reviewed 13 articles: 12 supported district-topic links and one concerned backpacks rather than book content. One correct district, Madison County in Mississippi, was missing from the mechanical matches because the NCES name uses abbreviated words. The manual correction is recorded separately; source data and all incorrect candidate pairs remain traceable.

Access and substantive review are different checks

HTML retrieval means that a page was returned with enough text for possible review. It can still contain a subscription notice, unrelated content or navigation. The collector did not log in, solve access challenges or treat blocked pages as irrelevant. Robots failures and unavailable rules are conservatively grouped with disallowed requests. Full article text is not distributed in the package; URLs, response hashes and access decisions are included.

The 749 exact-headline groups are a review aid. They do not establish independent reporting units. The largest group contains 92 records. Similar titles may hide changes in a story, and the same story may carry different titles.

What the episode checks establish

The 20-row selected episode table links 16 distinct PEN rows and preserves four unmatched title cases. It is a purposive case audit. Dates of publication, challenge, recommendation, board action and first inaccessibility remain distinct. Every first-inaccessibility field is still blank and every model-eligibility flag remains no.

RQ3 retained: What happened when a district reviewed a challenged book with input from librarians and local residents? Conway and Newtown illustrate different outcomes after review; they do not identify a causal effect of participation.

Reproduction and next work

The enrollment file now covers 41 reviewed district identities. The full agency inventory still needs eligibility checks. Run python run_all.py to rebuild the package outputs offline. The optional check_public_access.py collector is not run by that command. Original PDF minutes, source hashes, decision ledgers and all code are included; no credentials are required.

Next: complete substantive review and independent coding, reconcile primary title-level restriction records, and finalize district eligibility and coverage before estimating a national timing model. No additional BigQuery job, paid service, deployment or website update was performed.

Sources for the new case review

• Keller ISD to remove challenged books | The Texas Tribune — 2022-08-16; reviewed 2 October 2026.
• Spring Branch ISD book ban: More than $30K of taxpayer's money and 220 staff hours used for single ban, documents show — 2023-03-29; reviewed 2 October 2026.
• Michigan GOP front and center as Arab Americans push for removal of LGBTQ books in Dearborn schools — 2022-10-14; reviewed 2 October 2026.
• READ: Texas Teacher's Speech on Banned Books — 2022-07-26; reviewed 2 October 2026.
• South River, NJ, bans traditional school backpacks and book bags — 2022-08-02; reviewed 2 October 2026.
• Granbury ISD faces federal investigation over LGBTQ book removals — 2022-12-20; reviewed 2 October 2026.
• Somebody wants the Bible removed from Davis County school libraries — 2023-03-20; reviewed 2 October 2026.
• Nixa School Board to review multiple books that parents want banned or violate state law — 2023-06-19; reviewed 2 October 2026.
• Madison County Schools places 10 books in restricted circulation — 2022-08-02; reviewed 2 October 2026.
• Conway School Board passes bathroom ban, culls books with transgender themes — 2022-10-11; reviewed 2 October 2026.
• Newtown School Board Votes Unanimously Against Banning 2 Books — 2023-06-02; reviewed 2 October 2026.
• Nixa School Board bans four books after special meeting — 2023-06-20; reviewed 2 October 2026.
• Wilson County School board votes to remove two books for 'obscenity' — 2022-12-07; reviewed 2 October 2026.
• CONWAY_MINUTES_20221011 — district minutes.
• SBISD_AGENDA_20230124 — district agenda.
• NEWTOWN_PACKET_UNAVAILABLE — district packet attempt.
• SBISD_FOLLOWUP_20230201 — supplemental news outside annual candidate rows.
• DAVIS_RELEASE_REPRINT_20230620 — district statement reprinted by news outlet.

Earlier pass: expanded screening and enrollment

The expanded screen produced 2,065 record/entity pairs across 1,054 records, including 940 records missed by the original literal-name screen. These are leads, not ban events. All 114 original records were recovered.

The first purposive review batch attempted 15 new records: six supported district-topic coverage and nine could not be read. The remaining 925 have not been attempted. This is not a random validation sample and does not estimate matching accuracy. The full pending queue is included.

Why article-level review matters

Name matching also returned Wake County for “in wake of,” San Antonio ISD for a story about North East ISD, and districts in other states with similar names. These incorrect candidate pairs are explicitly rejected in the adjudication file. Two PEN aliases, Keller and Conway, are linked only after manual geography and historical-name checks. Confirmed article links are deduplicated by record and resolved ID; duplicate alias pairs are not separate articles.

Historical enrollment

The official NCES 2022–2023 LEA membership release provides two retained total rows for each of 18,601 agencies. The analysis denominator excludes adult education; original totals and source flags remain available for comparison. All 37 reviewed district joins have positive derived enrollment. Missing, suppressed and unreported values elsewhere stay as published.

The 19,714-row district-office inventory is joined by exact seven-digit LEA ID. It includes agencies whose study eligibility has not been checked. It is not yet a national comparison population; unmatched PEN records cannot be interpreted as zero bans. The 19,834 candidate keys also include unresolved PEN aliases and must not be counted as distinct districts.

NCES release description · Original membership ZIP. The package includes the 37,202-row total extract, CSV source-row references, source and extract hashes, and extraction code. The 1.43 GB source archive is omitted from the delivery ZIP.

What can be concluded now

The broader search improves coverage and reveals matching errors. Enrollment now supplies a time-matched denominator for reviewed districts. Neither change establishes whether media attention preceded first restrictions across districts. The previous 22-row Alpine month/stage comparison is unchanged. New articles have not yet undergone title-level PEN linkage. No causal model, forecast or website update was produced in this pass.

Next: review the remaining candidates and inaccessible pages; reconcile title, stage and restriction dates against primary district records; screen agency types and operational status for the eligible population; review coding independently before estimating a timing model. All work in this pass used the preserved export and public downloads, with no additional BigQuery query or paid service.

Reproduce: run python run_all.py after unzipping. Frozen review decisions are inputs; the program does not pretend to regenerate human article judgments. See expansion_summary.json for 85 additional passing checks.