# Verification checks (P4)
1. Reproducibility: deleted results.json/figure1.png/exclusions.csv, re-ran analysis.py; `cmp` with previous results.json → byte-identical. PASS
2. CI contains estimate: -0.3260 < -0.2558 < -0.1855. PASS
3. SEs positive: se_iid=0.01996, se_hac=0.03583. PASS
4. Exclusion ledger: 319 GDP quarters − 78 − 1 = 240 = n_analysed; n_before−n_after=n_removed per row. PASS
5. CI built as beta ± 1.959964·se_hac (checked numerically). PASS
6. Window: first_quarter=1960Q1, last_quarter=2019Q4, n=240; 3 months in every quarter (assert). PASS
7. Independent recomputation (pandas resample + hand-coded Bartlett HAC, L=4, no df correction): beta=-0.2557677, se_hac=0.0358346, mean g=0.75350 — matches to ~1e-14. PASS
8. Magnitude sanity: beta≈-0.26 for non-annualised growth (≈-0.064 per annualised pp) is in the conventional Okun range; mean growth 0.75%/q (≈3%/yr) plausible; se_hac > se_iid consistent with positive residual autocorrelation (lag-1 ρ=0.39). PASS
9. Growth not annualised: g = 100·ln ratio of consecutive quarters (code inspected; mean 0.75 not 3.0). PASS