1# Checks (P2/P4)2## Baseline (P2)3Naive (untransformed / linear) readings: E1 100·b = -43.1% (linear log-point approximation); E2 1.47 pp; E3 elasticity -0.388; E4 -0.354 log points; E5 -0.0962 (0-1 scale). The harmonised values replace these with the exact task conventions.4## Verification (P4)5| check | outcome |6|---|---|7| Clean rerun of out/analysis.py (results.json deleted then regenerated) byte-identical | PASS |8| n_harmonised + n_not_harmonisable = 5 = n_analysed; exclusions 5 − 0 = 5 | PASS |9| value null ⇔ harmonisable false ⇔ missing_quantity non-null | PASS |10| E1 −35.014% rounds to printed Table 6 "% change" −35.0% (5 km) | PASS |11| E1 exact conversion differs from naive linear −43.1% as expected (exp convexity) | PASS (documented) |12| E2 1.96% positive, small; consistent with 1.47 pp on a 75% base | PASS |13| E3 −1.82% consistent in magnitude with abstract "2% employment decline"; elasticity −0.388 ≈ (missing+excess)/g = (−0.03+0.012)/0.047 = −0.383 | PASS (approximate internal consistency) |14| E5 −0.0962/0.261 = −0.36858 vs printed −0.3685 (rounding/truncation) | PASS |15| E4: grep of extract for mean/baseline rate found no printed level of the fatal hospitalisation rate | PASS → not harmonisable |16| Every printed input matched by a unique regex line, printed by the script | PASS |17| No SEs/CIs produced → no CI-containment checks applicable | N/A |18