Data audit (P1)
All SHA-256 values below (first 16 hex shown; full hashes computed and identical to data/README.md and bundle.json). Rows = data rows after header.
| file | sha256 | rows | cols | first obs | last obs | missing | dup dates | month gaps |
|---|---|---|---|---|---|---|---|---|
| data/RSAFS_current.csv | 627d8bb386ac80f4… | 417 | 2 | 1992-01-01 | 2026-09-01 | 0 | 0 | 0 |
| data/UMCSENT.csv | 3076277f9f21f4bd… | 887 | 2 | 1952-11-01 | 2026-09-01 | 210 | 0 | 0 |
| data/first_release.csv | 1d1ae8b3cacf7799… | 28 | 5 | 2024-06-01 | 2026-09-01 | 0 | 0 | n/a |
| data/vintages/PAYEMS_vintage_2025-10-31.csv | d4b4c909ed321fbf… | 1040 | 2 | 1939-01-01 | 2025-08-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2025-11-30.csv | a05ef7fc50cfc339… | 1041 | 2 | 1939-01-01 | 2025-09-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2025-12-31.csv | 829dae087af61916… | 1043 | 2 | 1939-01-01 | 2025-11-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2026-01-31.csv | 8fd01c5d2740e869… | 1044 | 2 | 1939-01-01 | 2025-12-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2026-02-28.csv | 4a913e67ac5ffbb4… | 1045 | 2 | 1939-01-01 | 2026-01-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2026-03-31.csv | 919de86b07be9046… | 1046 | 2 | 1939-01-01 | 2026-02-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2026-04-30.csv | 06a8d663d9aaa61c… | 1047 | 2 | 1939-01-01 | 2026-03-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2026-05-31.csv | e2a152e874139acd… | 1048 | 2 | 1939-01-01 | 2026-04-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2026-06-30.csv | 19fbc6cd3547ffad… | 1049 | 2 | 1939-01-01 | 2026-05-01 | 0 | 0 | 0 |
| data/vintages/PAYEMS_vintage_2026-07-31.csv | 3a3a6603fcc31c6e… | 1050 | 2 | 1939-01-01 | 2026-06-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2025-10-31.csv | 3ecb0832a391e726… | 404 | 2 | 1992-01-01 | 2025-08-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2025-11-30.csv | 4a683d85374690f1… | 405 | 2 | 1992-01-01 | 2025-09-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2025-12-31.csv | 436df4b5509b1dfe… | 406 | 2 | 1992-01-01 | 2025-10-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2026-01-31.csv | b92ab6a5c1dbdc56… | 407 | 2 | 1992-01-01 | 2025-11-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2026-02-28.csv | 2af0a3aaeeba81ef… | 408 | 2 | 1992-01-01 | 2025-12-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2026-03-31.csv | 0316fdf238d55d9f… | 409 | 2 | 1992-01-01 | 2026-01-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2026-04-30.csv | 02dcc697a1db3475… | 411 | 2 | 1992-01-01 | 2026-03-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2026-05-31.csv | 8c3fe35302135896… | 412 | 2 | 1992-01-01 | 2026-04-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2026-06-30.csv | 2f103157f4a9b7a3… | 413 | 2 | 1992-01-01 | 2026-05-01 | 0 | 0 | 0 |
| data/vintages/RSAFS_vintage_2026-07-31.csv | 52c04a8fce2effa7… | 414 | 2 | 1992-01-01 | 2026-06-01 | 0 | 0 | 0 |
Units: RSAFS millions USD SA; PAYEMS thousands of persons SA; UMCSENT index 1966Q1=100; first_release levels in millions USD, yoy in percent (log ×100).
Findings / anomalies
- Irregular publication timing (RSAFS): last(V) = 2025-08 for V=2025-10-31 (two months lag; 2025-09 first appears 2025-11-30), 2026-01 for V=2026-03-31 (two-month lag), and V=2026-04-30 adds two months (2026-02, 2026-03). All other vintages end the month before V. Handled per task rule: targets are last(V)+h, not calendar-based.
- PAYEMS timing: vintage 2025-10-31 ends 2025-08, 2025-11-30 ends 2025-09, 2025-12-31 ends 2025-11 (no missing month; 2025-10 present from 2025-12-31). PAYEMS vintage 2026-02-28 shows a benchmark revision (2025-12 level 159,526 → 158,497). PAYEMS is not used in my model (see method), so no handling needed beyond documentation.
- UMCSENT: 210 empty values, all before 1978-01 (survey quarterly/not conducted). Not used in my model.
- Level break between last shipped vintage and current RSAFS: current file levels for 2025–2026 are ~4–5% below the vintage levels (e.g. 2026-06: 768,553 in vintage 2026-07-31 vs 732,898 now) — a large annual/benchmark revision after 2026-07-31. first_release.csv shows the break between target months 2026-07 (vintage 2026-08-31) and 2026-08 (vintage 2026-09-30). YoY rates are computed within a single vintage, so levels are consistent within each YoY; the break affects (i) the h=3 actuals of the last two origins (targets 2026-08, 2026-09, first released post-revision) and (ii) estimand (d). Documented; no adjustment.
- Revisions between consecutive shipped vintages are small (≤0.42% in level), affecting typically 4–5 recent months.
- No duplicates, no missing values in RSAFS/PAYEMS/first_release; no month gaps.
Availability-rule compliance
- RSAFS_current.csv is used only for estimand (d).
- UMCSENT current vintage is not used in forecasts (allowed only months ≤ month(V); not needed).
- first_release.csv used only as evaluation actuals.