← All proof trees

1. 1
   How it was computedSteps

   1. below-90 share = 100 − published 90%+ share, per school, year, and subgroup.
   2. Where exact counts are published (district file), aggregates use Σ(students − students at 90%+) / Σ students rather than averaging rounded percentages.

   Inputs: attendance_rate_above_90 (district); attendance_persistence_rate (state)

   Evidence: philly-absenteeism-descriptives.ts — search "pct_below_90"

2. 2
   Assumptions & caveatsAssumptions

   • 'Attending less than 90% of days' is treated as chronic absence. The federal definition counts a student who misses 10% OR MORE, so a student at exactly 90% attendance is chronically absent federally but is NOT in the below-90 share: for the same students and day-counting rules, the proxy cannot exceed the federal rate, and it understates it by the share of students sitting exactly on the line (not published by either publisher). That ordering does not carry over to a separately published federal statistic with different eligibility or attribution rules.
   • Minimum-enrollment rules differ: the district counts students enrolled 10+ days at the school (SDP, June 2023 attendance report); the state counts students enrolled 90+ school days (PDE glossary). PDE also lets each local education agency decide how partial-day absences are counted; identical coding across district and charter schools has not been verified.
3. 3
   How it was checkedHow it was checked

   District input: the district-file series is recomputed independently from the source CSV and matches the served database at every common cell (working-analysis §5). State input: every served cell matches a re-read of the state workbooks under the aligned attendance year (full QA §B), and the aligned cross-publisher check finds it tracks the district's figure for the same district-run schools (r ≥ 0.99 in every year from 2020-21 on) with a small median offset whose cause is not established. Charter schools and the two pre-pandemic years are outside that check.