Data audit (P1)
| file | SHA-256 | lines (wc -l) / items | matches README/bundle hash |
|---|---|---|---|
| data/effects.json | 708c1ff0c4379c228ac6a98f577c23d0d61564f3c958fb35596266aca1f5f10b | 5 top-level items (ids E1–E5; keys id, source, effect, reported_definition, target_definition) | yes (bundle) |
| data/README.md | 7cf7cb385f9fb4fc0df52cfc6b0644bcbfb969a38afe2b018b864b7871187dc4 | 16 lines | yes (bundle) |
| data/sources/arxiv-2609.27732v1.txt | ee2d444d…6a8b | 134 lines | yes |
| data/sources/arxiv-2606.14621v1.txt | 27c463c0…a238 | 58 lines | yes |
| data/sources/arxiv-2505.04555v3.txt | b1731a9b…95c1 | 66 lines | yes |
| data/sources/arxiv-2510.03658v1.txt | f8782937…1127 | 91 lines | yes |
Schema: text extracts with provenance header; tables one row per line, cells separated by " | ", SE row follows coefficient row. No duplicates (5 unique ids). No missing fields in effects.json.
Units: DCR ATTs in log points of hospitalisation rate per 100,000; RESEA reemployment in % (means) and pp (effects); MW Table 2 elasticities (unitless) and '% MW changes' as a proportion (0.047); Benin Table 3 LPM coefficient on 0–1 scale, mean dep. var. 0.261.
Anomalies (handled in results.json data_issues): DCR abstract "fatal fall by 35%" vs Sec. 6.1 "approximately 30%" vs Table 7 (exp(-0.354)-1 = -29.8%); SE 0.210 (Table 6) vs 0.218 (Table A19) for same ATT; 608 vs 640 control regions; MW note "about 4%" vs table 0.047; Benin text cites "table 2" (not in extract) and calls a relative change a "percentage change"; RESEA abstract rounds 1.47 pp to 1.5 pp.
Information availability: all inputs are admissible extracts. E4 requires the baseline fatal hospitalisation rate per 100,000, which is not printed in the extract (grep for mean/baseline/rate found none) -> declared not harmonisable; no external value used. No row-level filtering applies; exclusion ledger records 5 items in, 5 analysed.