# Data audit
| file | SHA-256 | rows (excl. header) | cols | range | missing | duplicates dates |
|---|---|---|---|---|---|---|
| GDPC1.csv | 1c6c6e9f59e383e20c849dd299175c7a307526c09abdc18f18da9754fc5b3343 (matches README) | 319 | 2 (observation_date, GDPC1 bn chained 2017 $ SAAR) | 1947-01-01..2026-07-01, no gaps in quarterly grid | 0 | 0 |
| UNRATE.csv | 621ff6fadf50910cb7a21ee20ba610c247cb043b29b4e05c4c1a72fadca2a06c (matches README) | 945 | 2 (observation_date, UNRATE %) | 1948-01-01..2026-09-01, no gaps in monthly grid | 1 (2025-10-01 empty value) | 0 |

Anomalies:
- UNRATE 2025-10-01 is blank (outside sample; dropped by the sample window, no effect). 
- Values in range (GDPC1 2172–24611; UNRATE 2.5–14.8).
- Data vintage is 2026 (GDPC1 in chained 2017 $): estimates reflect current revised data, not real-time data. Not a violation of the rules (task defines these files as inputs).
- Observations after 2019Q4 violate the sample rule if used: handled by restricting to 1959Q4–2019Q4 levels before differencing/estimation.