Clean rerun: deleted results.json/figure1.png/exclusions.csv, reran analysis.py, cmp with previous results.json |
PASS (byte-identical) |
| Independent numpy recomputation (raw-file aggregation, OLS, iid SE, Bartlett NW L=4 no df-correction, R², mean g) |
PASS all to <1e-9 relative |
| n_analysed = 240, first 1960Q1, last 2019Q4 |
PASS |
| se_iid > 0, se_hac > 0 |
PASS |
| CI contains beta_hat; symmetric around beta_hat |
PASS |
| okun_negative_5pct consistent with beta<0 and CI excluding 0 |
PASS |
| Exclusion ledger: n_before−n_after=n_removed per row; chained; 319−79=240 |
PASS |
| Window/info rule: only 1959Q4–2019Q4 levels used (script filters before computing) |
PASS |
| Magnitude: mean g 0.754 %/q ≈ 3.0 % annualised (plausible US 1960–2019); β ≈ −0.26 in line with conventional Okun estimates (~−0.3 to −0.5 for Δu on q/q growth, differs by spec); HAC/iid SE ratio 1.80 consistent with residual AC(1)=0.39 |
PASS (plausible) |
| Bandwidth sensitivity: se_hac for L∈{0,1,2,4,8,12} in 0.0288–0.0371; conclusion unchanged |
PASS |