# Contract — F4-conf-03 (written 2026-10-30)

Question. Re-express five printed treatment-effect estimates (from four arXiv extracts) on a target definition given per item in data/effects.json, using only quantities printed in the cited extract and an explicit transformation; where impossible, state so and name the missing quantity.

Estimands. Harmonised point estimates (no uncertainty propagation):

• E1 DCR overall ATT 5 km (log points) -> relative % change of counterfactual rate. Valid under log-outcome interpretation: 100(exp(b)-1).
• E2 RESEA ATE on reemployment (pp) -> relative % of control mean: 100·Δ/p (control mean taken as counterfactual level).
• E3 MW employment elasticity -> % employment change for observed MW increase: 100·η·g, g printed in same table.
• E4 DCR fatal ATT 5 km (log points) -> absolute change per 100,000 per year: requires a baseline (counterfactual) level of the fatal rate.
• E5 Tossou LPM coefficient (0-1 scale) -> percentage points: ×100.

Outputs. out/results.json keys: harmonised{E1..E5: value, unit, transformation, harmonisable, missing_quantity, locator}, n_harmonised, n_not_harmonisable, task_id="F4-conf-03", feasibility, feasibility_reason, data_issues[{issue,handling}], n_input_rows{file: lines/items}, n_analysed, random_seed=null. Also out/report.md (Question; Data and exclusions; Method; Results; Uncertainty; Limitations and open issues), out/analysis.py (prints source lines used), out/claims.json.

Information rules. Extracts are the only admissible evidence; nothing from memory/other versions. harmonisable=false only if a required quantity is not printed; then value null + missing_quantity. Report printed numbers verbatim.

Invalidity/impossibility. Using a non-printed quantity (e.g., guessed baseline rate for E4) would invalidate. Task is "complete" if each item has a value or justified impossibility statement.