Data audit (P1)
All SHA-256 values of the four extracts match those listed in data/README.md.
| file | SHA-256 | lines (wc -l) / items | content |
|---|---|---|---|
| effects.json | 708c1ff0c4379c228ac6a98f577c23d0d61564f3c958fb35596266aca1f5f10b | 5 items | keys id, source, effect, reported_definition, target_definition; no missing fields, ids E1–E5 unique |
| README.md | 7cf7cb385f9fb4fc0df52cfc6b0644bcbfb969a38afe2b018b864b7871187dc4 | 16 | provenance table |
| sources/arxiv-2609.27732v1.txt | ee2d444d…6a8b (matches) | 134 | DCR paper: abstract, Tables 6, 7, Sec 6.1, Table A19. Log-point ATTs; no level/mean of any hospitalisation rate printed |
| sources/arxiv-2606.14621v1.txt | 27c463c0…a238 (matches) | 58 | RESEA: Table 2 (means, diffs, ATEs), Table 3. Reemployment in % / pp |
| sources/arxiv-2505.04555v3.txt | b1731a9b…95c1 (matches) | 66 | MW paper: Table 2 (elasticities; '% MW changes' 0.047) with note ('about 4%') |
| sources/arxiv-2510.03658v1.txt | f8782937…1127 (matches) | 91 | Benin: Sec 4.1, Table 3 (LPM, Mean Dep. Var. 0.261, % Change row) |
Units: E1/E4 log points; E2 percentage points (control mean 75.0%); E3 elasticity and proportional MW change; E5 probability on 0–1 scale.
Duplicates: none (Reemployment appears in Table 2 and Table 3 of RESEA — distinct tables; Table 2 row used as specified).
Anomalies (detail in results.json data_issues): DCR abstract fatal "35%" vs Table 7 −0.354 (≈−29.8%); DCR SE 0.210 vs 0.218 and 608 vs 640 control regions; MW 0.047 vs "about 4%"; MW duplicated SE 0.091; Benin % Change −0.3685 vs computed −0.36858, "table 2" reference; RESEA abstract 1.5 pp vs table 1.47 pp.
Information-availability: no input violates the rules. The decision-relevant gap is E4: the extract prints no baseline/counterfactual level of the fatal rate per 100,000, so the absolute change cannot be computed (handled as harmonisable=false, value null). The script also searches the extract for any such printed level (0 found).
Exclusions: none (out/exclusions.csv; 5 raw items − 0 = 5 analysed).