# Data audit
| file | SHA-256 | lines (wc -l) | matches README/bundle hash |
|---|---|---|---|
| data/README.md | 16ebdff389eeaee9508fe81c4edc43690f31f3463c469680c814134086ef9929 | 10 | yes (bundle.json) |
| data/sources/arxiv-2609.27732v1.txt | ee2d444d2f78f2af2ca2cf69b59a61fcd7ac5e10ea29e8ddcd67917bf07d6a8b | 134 | yes |
| data/sources/arxiv-2606.14621v1.txt | 27c463c0a144d9d5a3916dc641f15b6029f6c68b5e574ad9074e7df1aeb0a238 | 58 | yes |

Schema: plain-text tables with " | " separators. In DCR regression tables the SE line follows each coefficient row. DCR units are log points; the controls are 608 never-treated MedStat regions; the panel covers 1998-2022 openings. RESEA Table 2 cells hold "estimate; p; 95% CI"; wages are USD/yr, reemployment is pp and UI duration is weeks; the trial ran Feb 2022 to Sep 2023 with N=23,549.
Missing: no exact p-values anywhere in the DCR extract (stars only). No SEs are printed in RESEA Table 2.
Anomalies (also in results.json data_issues):
1. Table 6 5 km SE (0.210) differs from A19 CS-DR SE (0.218) for the same estimate.
2. 608 vs 640 control regions (A19 note).
3. The abstract says fatal hospitalisations fall "35%", but Table 7 gives -0.354 log points (about -29.8%; Sec 6.1 says ~30%).
4. The text calls the HIV estimates "close to zero", but the 5 km value is -0.340.
5. The RESEA text gives the weeks CI as unsigned magnitudes.
6. "p ¡ 0.001" rendering defect.
7. RESEA tables were re-set by the editor.
Availability: neither file breaks the rules. No external information is used.