# Data audit (P1)
| file | SHA-256 (matches README & bundle.json) | data rows | cols | range |
|---|---|---|---|---|
| UNRATE.csv | 621ff6fadf50910cb7a21ee20ba610c247cb043b29b4e05c4c1a72fadca2a06c | 945 | 2 (observation_date, UNRATE) | 1948-01-01 .. 2026-09-01 monthly |
| GDPC1.csv | 1c6c6e9f59e383e20c849dd299175c7a307526c09abdc18f18da9754fc5b3343 | 319 | 2 (observation_date, GDPC1) | 1947-01-01 .. 2026-07-01 quarterly |

Units: UNRATE percent (SA); GDPC1 billions of chained 2017 USD, SAAR.

Findings:
- UNRATE: no duplicate dates, no gaps in the monthly calendar, dates monotonic. Range 2.5–14.8. **One empty value at 2025-10-01** (missing observation; October 2025, outside the analysis window). All 723 months 1959-10..2019-12 present and non-missing.
- GDPC1: no missing values, no duplicate dates, no calendar gaps, all dates are first day of quarter months. Largest absolute level changes are 2020Q2/Q3 (COVID), outside window.
- Information-availability: files include data after 2019Q4 (and before 1959Q4). These are not used: the analysis restricts to 1959Q4 levels (for differencing) through 2019Q4. No violation remains after filtering. Data are current vintage, which is permitted.

Handling: missing 2025-10 UNRATE value recorded in data_issues; it is excluded by the window filter and does not affect results. Quarterly averages are only computed from quarters with 3 non-missing months (checked in analysis.py).