# Data audit — F4-conf-03

All SHA-256 values computed with `shasum -a 256` match data/README.md and bundle.json.

| file | SHA-256 | lines (wc -l) / items | content |
|---|---|---|---|
| data/effects.json | 708c1ff0…10b | 5 top-level items (JSON list) | ids E1–E5, fields id/source/effect/reported_definition/target_definition; no missing fields, no duplicate ids |
| data/README.md | 7cf7cb38…dc4 | 16 lines | provenance table |
| data/sources/arxiv-2609.27732v1.txt | ee2d444d…8b | 134 lines | Armendariz: abstract, Tables 6, 7, Sec 6.1, Table A19. Units: log points on log rate per 100,000 |
| data/sources/arxiv-2606.14621v1.txt | 27c463c0…238 | 58 lines | Li et al.: Table 2, Table 3. Means in %, effects in pp |
| data/sources/arxiv-2505.04555v3.txt | b1731a9b…c1 | 66 lines | Kanayama et al.: Table 2 (elasticities; '% MW changes' 0.047) |
| data/sources/arxiv-2510.03658v1.txt | f8782937…27 | 91 lines | Tossou: Sec 4.1, Table 3 (LPM, Mean Dep. Var. 0.261) |

Rows = printed table rows; no date-ordering issues (retrieval 2026-10-30T03:23:18Z).

## Anomalies / internal inconsistencies found
1. Armendariz abstract: fatal hospitalisations "fall by 35%", but Table 7 5 km fatal ATT -0.354 implies 100(exp(-0.354)-1) = -29.8%; Sec 6.1 says "approximately 30%". Table value used.
2. Armendariz Table A19: CS-DR SE (0.218) differs from Table 6 5 km SE (0.210) for the same point estimate -0.431. Not needed (no SE propagation).
3. Armendariz Table A19 note: "640 regions" never-treated vs "608" in table body and Tables 6/7. Not needed.
4. Armendariz Table A19 % change rows for Synthetic DiD (-30.3%) and Stacked DiD (-13.9%) differ by 0.1 from 100(exp(b)-1) on printed b (-30.4%, -13.8%); probably unrounded coefficients. Not needed.
5. Armendariz: no baseline/pre-treatment/counterfactual level of any hospitalisation rate is printed in the extract -> E4 not harmonisable. The panel's time unit ("per year") is also not stated explicitly.
6. Kanayama Table 2 note says "about 4% minimum wage increase" while row '% MW changes' prints 0.047 (4.7%). Task says use the size printed in the table -> 0.047 used.
7. Tossou: abstract/text quote 0.0855 and mean 0.455 from "table 2" (not in extract); Table 3 col (4) prints % Change -0.3685 while -0.0962/0.261 = -0.3686 (rounding of printed mean). Footnote 11 text garbled (rendering). Only col (4) Treatment used.
8. Li et al.: rendering defect "p ¡ 0.001"; control mean printed with one decimal (75.0%); ATE is block-adjusted while 75.0% is the raw aggregate control mean.

## Information-availability
No input violates the rules; all are designated admissible extracts. No external values used.