PhiladelphiaMethodologymedium
Philly publishes both Acct and Actual PSSA cuts. They're byte-identical.
Thesis
SDP publishes PSSA in two flavors: 'Acct' (accountability cohort, Oct-31-register attribution — the canonical SPREE-feeding figure) and 'Actual' (all tested students). NYC's analogous distinction drove known snapshot-agreement gaps on the test metrics (73% ELA / 62% math). In Philly's 2024-25 file, all 4,535 paired cells show Δ=0. The two cuts are byte-identical at the school × subgroup rollup grain. Why?
Supporting findings
- Acct vs Actual paired cells: 4,535. Δ=0 across every single one.
- Mean delta: 0. Median: 0. Stdev: 0. Range: 0-0.
- Two interpretations: (a) SDP's pipeline produces both files from the same intermediate at the school-aggregate grain; (b) the student-level Acct/Actual gap exists but cancels at school aggregation because the in/out flips offset.
- Either way the validation harness keeps both load paths so that future divergence (rule changes, edge cases) surfaces immediately.
Reporting directions
- Reach out to SDP's research office to ask which interpretation is correct.
- Compare to NYC's reported test-metric gaps (73% snapshot agreement on ELA, 62% on math; NYC's 67% absenteeism gap was later explained as a grade-band display, not a cohort cut — NYC ledger A.6) — the structural difference is itself the story.
- Test 2023-24 + 2022-23 backward — does this trivial-agreement pattern hold across years?
Methodology & replication recipe — the queries + sources behind the findings above.
More on methodology
- SPREE agrees with our DB on 616/616 cells — Philly's data pipeline is structurally clean
- Computed-value (iv) validation: OSS + PSES rollups agree to the cell with the raw source
- Philly PSSA: 2023-24 vs 2024-25 year-over-year — which schools moved most
- With 287 peer-grouped schools and k=40, the rankings barely move at k=20