Data audit (P1)
All SHA-256 hashes match data/README.md and bundle.json.
| file | SHA-256 | data rows | cols | range | NA | dup dates |
|---|---|---|---|---|---|---|
| current_vintage.csv | 1af2590d…0562 | 318 | 2 (observation_date, A191RL1Q225SBEA) | 1947-04-01…2026-07-01 | 0 | 0 |
| first_release.csv | ab843c9f…95bb | 13 | 3 (quarter, first_release_value, first_release_vintage_date) | 2023-07-01…2026-07-01 | 0 | 0 |
| vintages/…2024-01-31.csv | 7ce218fb…f243 | 307 | 2 | 1947-04-01…2023-10-01 | 0 | 0 |
| vintages/…2024-04-30.csv | 62c1dfce…97cb | 308 | 2 | …2024-01-01 | 0 | 0 |
| vintages/…2024-07-31.csv | 6f942591…79da | 309 | 2 | …2024-04-01 | 0 | 0 |
| vintages/…2024-10-31.csv | c4c51ede…0003 | 310 | 2 | …2024-07-01 | 0 | 0 |
| vintages/…2025-01-31.csv | 10a5e02e…e2 | 311 | 2 | …2024-10-01 | 0 | 0 |
| vintages/…2025-04-30.csv | 1f19c975…85d1a | 312 | 2 | …2025-01-01 | 0 | 0 |
| vintages/…2025-07-31.csv | f1e7c556…a9 | 313 | 2 | …2025-04-01 | 0 | 0 |
| vintages/…2025-12-31.csv | c111c070…a8 | 314 | 2 | …2025-07-01 | 0 | 0 |
Units: percent change from preceding quarter, SAAR (percentage points of annualised growth). Quarterly spacing regular (90–92 days), no gaps.
Anomalies: pandemic quarters 2020Q2 (≈ −28) and 2020Q3 (≈ +35) are extreme outliers in every vintage; handled in the model by excluding regression observations involving them (decided a priori, before any forecast error was computed). Values of recent quarters differ across vintages (normal revisions, e.g. 2025Q1: −0.3 → −0.5 → −0.6). The eighth origin is 2025-12-31 rather than 2025-10-31 (shutdown delay; README), so origins are not evenly spaced — no action needed; targets are still last quarter +1/+2.
Information availability: every target quarter (2024Q1…2026Q1) has a first-release value; all first_release_vintage_date values for targets are after the corresponding origin (checked in analysis.py). current_vintage.csv violates the real-time rule for estimation, so it is used only for the revision statistic. first_release.csv is used only as actuals.