1# Verification checks (P4)2| check | outcome |3|---|---|4| Clean re-run (deleted results.json, table1.csv, figure1.png, exclusions.csv; re-ran analysis.py); results.json byte-identical | PASS |5| Exclusion ledger: 319 GDPC1 quarters − 78 − 1 = 240 = n_analysed | PASS |6| n_sub1 + n_sub2 = 96 + 144 = 240 = n_analysed | PASS |7| sample_start 1960-01-01 < sample_end 2019-10-01; no quarter after 2019Q4 used (asserted in script) | PASS |8| All 3 monthly UNRATE values present for every quarter 1959Q4–2019Q4; no missing GDPC1 (asserted) | PASS |9| effect_se > 0; CI contains estimate; CI symmetric with z=1.96 | PASS |10| model_metric (0.250) < baseline_metric (0.325); 1 − (model/baseline)^2 = R² (0.408) to 1e-12 | PASS |11| Magnitude: β ≈ −0.26 pp per 1% (non-annualised) quarterly growth, consistent with the commonly cited US Okun coefficient of roughly −0.3 to −0.5 in level-gap form; order of magnitude plausible | PASS (plausible) |12| Chow df = (2, 236); F > 0 | PASS |13| Full-sample β (−0.256) lies outside [β_sub2, β_sub1] = [−0.273, −0.259]: explained by the intercept shift between subperiods (pooled fit with a common intercept tilts the slope); not an error | NOTED |14| HAC SE sensitivity to lag length (0,2,4,8,12): 0.029–0.037, conclusion unchanged | PASS |15