Find What Works: Schools
← All source documentsRaw file
Source document

docs/agents/pass-log/covid-absenteeism-stuck.md

Served verbatim from the project repository. Internal working document conventions apply: documents may reference file paths, branch names, and findings-ledger anchors from the repo.

Pass log — covid-absenteeism-stuck

Story: data/stories/answers.ts → "covid-absenteeism-stuck"

PASS 1 — Editor (2026-05-29)

WHAT I CHANGED

  • TLDR was too long. Five sentences in one tldr is a wall. Cut to three sentences that lead with the headline number (6.8 above baseline, recovery flattening), name the named schools as a handle, and end with the structural finding (distribution shape unchanged).

  • "Every borough is worse" section heading is harder to read than it should be. Renamed to "Borough breakdown — the new floor sits above the old ceiling."

  • "The full school-level distribution by year" lead opens with jargon ("Means hide what's happening across schools"). Replaced with a concrete description.

  • Most-improved / Most-backslid tables have lots of decimal points in school names that aren't carrying weight. Tightened headings.

WHAT I FLAGGED FOR DOWNSTREAM

  • Expert (pass 2): PS 116 Mary Lindley Murray (Murray Hill, Manhattan) is named as a wealthy-zone backslider. Verify this is in a reasonable income context — Murray Hill is wealthy but PS 116's feeder mix may include lower-income kids from Kips Bay / Stuyvesant Town. The "wealthy zone" framing may overstate.

  • Quant (pass 3): The 2018-19 → 2024-25 deltas could include schools that opened or closed mid-period. Verify the most-improved / most-backslid lists exclude schools with non-comparable years.

LEGAL REVIEW NOTES

  • Naming specific schools in "most backslid" is borderline. PS 116 going from 18% to 56% chronic absenteeism is fact, but framing it as a backslider attributes performance change to the school. Keep the framing neutral — describe the change, not the school.

PASS 2 — Education expert (2026-05-29)

WHAT I VERIFIED OR CORRECTED

  • PS 116 Mary Lindley Murray is in District 2, Murray Hill (E. 33rd Street). The zoned area is genuinely wealthy/high-income; the "wealthy zone" framing is defensible. However, the school has also absorbed some students from temporary housing in the surrounding area, which is a different population than zoned residents. The story should note that the chronic-absenteeism rise reflects both the zoned-family attendance shift AND the temp-housing population.

  • PS 003 Bedford Village (Brooklyn D14, "third party of the list") is in Bedford-Stuyvesant — heavily gentrified since ~2015. The rise is consistent with both gentrifying-family-flexibility patterns and temp-housing assignment patterns. Verifiable.

  • Bronx Aerospace HS is the headline "most-improved" school — there's a separate story (bronx-aerospace-mystery) that already covers the front-loaded nature of its gain. Link the two stories here.

  • The 2020-21 and 2019-20 data points are artifacts of remote learning. The story should explicitly say "the metric measures attendance to in-person school, not engagement, so the COVID-era numbers reflect the participation regime, not student behavior."

WHAT I ADDED

  • A sentence flagging that PS 116 backslide reflects multiple populations at that school, not just the zoned wealthy families.

  • Cross-link to bronx-aerospace-mystery for the front-loaded improvement story.

  • A note that NYC's "attendance teacher" program (post-2022) is the most concrete intervention NYC DOE has deployed against this trend. Robert Balfanz's Johns Hopkins work on attendance has been used to design these.

WHAT I FLAGGED FOR DOWNSTREAM

  • Quant (pass 3): The "permanent 7-point increase" framing implies the trend has settled. Three years of leveling-off is suggestive but not conclusive — recovery rates from public-health disruption tend to be non-linear. The "won't return to baseline until 2031" claim is a linear extrapolation that's likely wrong in both directions.

PASS 3 — Quantitative (2026-05-29)

WHAT I VERIFIED

  • Year-over-year deltas verified: 2021-22→22-23 = −4.5; 22-23→23-24 = −1.5; 23-24→24-25 = −1.4. The trend toward smaller annual gains is real in this data.

  • The "distribution width is unchanged" finding holds: the p25-to-p90 range was 33.8 points in 2018-19 (17.4 → 50.7) and 30.2 points in 2024-25 (25.9 → 56.1). Slightly narrower, not wider. The story claim is correct.

  • Borough deltas verified against the absenteeismByBoroughYear data.

WHAT I CORRECTED OR QUALIFIED

  • "Won't return to pre-pandemic baseline until 2031" needs hedging: this is a linear extrapolation. Three years of leveling-off doesn't establish that the rate is stable; non-linear public-health recovery is common. Reframed as a conditional: "at the current rate, returns to baseline circa 2031; non-linear recovery could change this."

  • Sample comparability across years: ~1,500 schools per year, comparable. The most-improved/backslid lists are filtered to schools with measurable data in BOTH 2018-19 and 2024-25, which excludes ~30 schools that closed or opened in the window. Worth noting.

  • The "permanent 7-point increase" framing is observationally consistent with the data but is a forward-looking claim. Replaced with "the new floor appears to sit ~7 points above the old."

WHAT I FLAGGED FOR DOWNSTREAM

  • Editor (pass 4): The caveats now run long after my hedging. Verify they don't smother the story.

PASS 4 — Editor (second pass) (2026-05-29)

WHAT I CHANGED

  • Caveats block absorbed the quant's "linear extrapolation" hedge cleanly — kept it to two sentences.

  • "Borough breakdown — the new floor sits above the old ceiling" is now the section heading; renamed once more to "Every borough is worse than its pre-COVID baseline" because the "ceiling" metaphor was confusing (the OLD ceiling is now actually below the NEW floor for Manhattan, but that's not true for Queens). Plain version is clearer.

  • Cross-link to bronx-aerospace-mystery embedded inline in the most-improved section.

WHAT I FLAGGED FOR DOWNSTREAM

  • Story is in solid shape. Passes 5 and 6 can be light.

PASS 5 — Education expert (second pass) (2026-05-29)

WHAT I VERIFIED OR CORRECTED

  • One factual note: the "attendance teacher" deployment specifically targets schools with ≥30% chronic absenteeism (per NYC DOE guidelines circa 2022). Most schools on the most-backslid list now meet that threshold. Worth a sentence.

  • The District 75 entry (51.8% → 58.3%, +6.5) — this is special-ed schools and includes students whose disabilities directly affect attendance. Reading "D75 absenteeism is up" without that context invites a misreading.

WHAT I ADDED

  • Brief note about D75 context — these are special-ed students whose medical/transportation needs intersect with absenteeism in ways that don't apply to general-ed populations.

PASS 6 — Quantitative (second pass) (2026-05-29)

WHAT I VERIFIED

  • Year-over-year deltas still computed correctly after pass 5 additions. No structural issues.

  • The D75 caveat is statistically appropriate; D75 schools are a different population and shouldn't be averaged into general-pop trends without flagging.

WHAT I FLAGGED FOR DOWNSTREAM

  • Nothing — ready for final polish.

PASS 7 — Editor (final) (2026-05-29)

WHAT I CHANGED

  • Final caption polish on the most-improved table caption: added "(filtered to schools with comparable data in both years)" per the quant's pass-3 note.

  • shortPreamble end-sentence: "where parents control attendance" / "where students decide" is the right framing, kept as-is.

KNOWLEDGE-FILE UPDATE

Added "Public-health-style 'won't recover until year X' linear extrapolations are usually wrong" to docs/agents/quantitative-knowledge.md.

FINAL STATUS

Story is publishable. ~1900 words; gained ~200 words across passes (D75 context, PS 116 multi-population note, linear-extrapolation hedge, cross-link to bronx-aerospace).

PASS 8 — Plain-language (2026-05-29)

WHAT I CHANGED

  • TLDR + shortPreamble fully rewritten. First sentence now defines "chronic absenteeism" inline: "students who miss at least 10% of school days." That's the term the story rests on; no reader should hit it cold.

  • Section headings rewritten:

    • "The citywide trend, year by year" → "How chronic absenteeism moved year by year"
    • "Every percentile moved together" → "It's not just the worst schools — every kind of school is more absent"
    • "Borough breakdown" → "Every borough is worse off than before the pandemic"
    • "Grade-band breakdown" → "Where age matters — elementary schools fared worst, high schools best"
    • "Most-improved schools" → "The schools that improved the most"
    • "Most-backslid schools" → "The schools that went backwards the most"
    • "What's the recovery rate doing?" → "Is the recovery still working?"

  • "Percentile" / "p10" / "p90" → plain language in the distribution chart caption: "bar shows the middle half of schools; lines extend to the top and bottom 10%".

  • "Distribution width" / "standard deviation" removed; replaced with concrete description ("the schools at the best end are now about as far above the schools at the worst end as they used to be").

  • "D75" → "District 75 (the citywide system of schools for students with significant disabilities)" — defined inline.

  • "K-8 schools" kept; meaning clear from context.

  • "Robert Balfanz / Johns Hopkins" → "Johns Hopkins researcher Robert Balfanz" — same person, more reader-legible.

  • "Linear extrapolation" → "straight-line projection."

  • Chart captions all rewritten with plain captions ("Average percentage of students chronically absent across all New York City public schools, year by year").

JARGON KILLED

  • chronic absenteeism (defined inline first use)
  • distribution / spread / percentile
  • monotonic
  • baseline (kept once but contextualized)
  • "grade band" (replaced with "elementary schools" / "high schools" etc.)
  • "single-year residual"
  • D75 (replaced with definition)
  • proclivity decile (replaced with student-body challenge group)
  • linear extrapolation

ACRONYMS SPELLED OUT (first uses)

  • NYC → New York City
  • COVID (kept; universal)
  • DOE → Department of Education
  • D75 → District 75 (with definition)

NEW PLAIN-LANGUAGE PATTERNS I'M ADDING TO THE KNOWLEDGE FILE

  • "Chronic absenteeism" needs inline definition first time it appears — "students who miss at least 10% of school days." This is the single most-used technical term across NYC ed-data stories.
  • "Distribution / spread / percentiles" → describe what the box and whiskers represent, in the caption.
  • "D75" / "District 75" → define inline.

WHAT I LEFT ALONE AND WHY

  • "Chronic absenteeism" itself (kept as the term, just defined). It's the official metric NYC DOE reports; renaming would lose precision.
  • School names (P.S. 116, Bronx Aerospace, etc.).
  • Murray Hill, neighborhood name.

Net change: ~2050 → ~2350 words.