# Data audit (P1)
| file | SHA-256 (computed; matches README) | data rows | cols | range | missing | dup dates |
|---|---|---|---|---|---|---|
| UNRATE.csv | 621ff6fadf50910cb7a21ee20ba610c247cb043b29b4e05c4c1a72fadca2a06c | 945 | 2 (observation_date, UNRATE float, percent) | 1948-01 to 2026-09, monthly, no gaps in dates | 1 (2025-10-01 empty field; likely 2025 federal shutdown) | 0 |
| GDPC1.csv | 1c6c6e9f59e383e20c849dd299175c7a307526c09abdc18f18da9754fc5b3343 | 319 | 2 (observation_date, GDPC1 float, bn chained 2017 $ SAAR) | 1947Q1 to 2026Q3, quarterly, no gaps | 0 | 0 |

- UNRATE range 2.5–14.8 %, GDPC1 range 2172–24611; all positive, no anomalies.
- The single missing UNRATE value (2025-10) lies outside the 1959Q4–2019Q4 range needed: no effect. Recorded in `data_issues`.
- Information rule: files are current vintage (allowed, "treat as given"). Observations after 2019Q4 and before 1959Q4 are filtered out by the script before any computation; no violation.
- Exclusion ledger (`out/exclusions.csv`) follows the quarterly panel (unit of analysis): 319 GDPC1 quarters − 78 (outside 1959Q4–2019Q4) − 0 (incomplete UNRATE quarter) − 1 (1959Q4 differencing) = 240 = n_analysed. Monthly UNRATE rows in window: 723 (=241×3), all non-missing.