Data audit
| file | SHA-256 | lines (wc -l) | matches README hash |
|---|---|---|---|
| data/effects.json | 708c1ff0c4379c228ac6a98f577c23d0d61564f3c958fb35596266aca1f5f10b | 5 top-level items | n/a |
| data/README.md | 7cf7cb385f9fb4fc0df52cfc6b0644bcbfb969a38afe2b018b864b7871187dc4 | 16 | n/a |
| sources/arxiv-2609.27732v1.txt | ee2d444d...a8b | 134 | yes |
| sources/arxiv-2606.14621v1.txt | 27c463c0...238 | 58 | yes |
| sources/arxiv-2505.04555v3.txt | b1731a9b...95c1 | 66 | yes |
| sources/arxiv-2510.03658v1.txt | f8782937...127 | 91 | yes |
Schema: effects.json = list of 5 objects {id, source, effect, reported_definition, target_definition}; no duplicates, no missing fields. Sources are plain text, tables one row per line with " | ".
Anomalies found
- DCR (2609.27732): CS-DR 5 km SE printed 0.210 in Table 6 but 0.218 in Table A19 (same point estimate -0.431). Not used (no SE propagation).
- DCR: Table A19 note says 640 never-treated regions, tables say 608. Not used.
- DCR: abstract says fatal outcomes "fall by 35%"; Sec 6.1 says "approximately 30%"; 100(exp(-0.354)-1) = -29.8%. Abstract inconsistent with Table 7. Not needed for E4 (absolute target).
- DCR: Sec 6.1 says narcotics "rise by 47%"; 100(exp(0.390)-1)=47.7%. Minor (truncation). Not used.
- DCR: no baseline/pre-period/control-mean level of the fatal (or any) hospitalisation rate is printed -> E4 not harmonisable.
- MW (2505.04555): Table 2 note says "about 4% minimum wage increase" while row "% MW changes" prints 0.047. Use the table row (task instruction).
- MW: SE of "Emp. elasticity w.r.t affected wage" (0.091) equals SE of "Affected employment"; possible copy; not used.
- RESEA (2606.14621): Table 2 control mean 75.0% is the raw aggregate control mean, not block (randomisation-week) adjusted, whereas ATE 1.47 pp is block-adjusted; the paper itself shows raw comparisons are biased. Used as instructed (target = percent of control-group mean); flagged.
- Benin (2510.03658): text refers to "table 2" for -0.0855 with mean 0.455 (table not in extract); abstract calls 18.79 a "percentage change" (it is a percent). -0.0962/0.261 = -0.3686 vs printed % Change -0.3685 (rounding). Not affecting E5.
No information-availability violations in the inputs. No rows excluded.