# Contract — F6-dev-01

Question. Does the cross-country Preston curve (life expectancy rising with ln income) hold in the most recent usable WDI year, and how strong is it? Descriptive/associational only.

Estimand. Slope β of the linear projection LE_i = α + β ln(GDPpc_i) + ε_i over countries/economies (equal weights, one obs per country) in the analysis year. LE = life expectancy at birth (years); GDPpc = GDP pc PPP constant 2021 intl $. β is a best-linear-predictor coefficient (meaningful without linearity/exogeneity assumptions as a projection); inference with heteroskedasticity-robust SE (HC3 chosen), countries treated as independent draws.

Outputs. out/results.json with keys analysis_year, n_excluded_missing, baseline_metric, model_metric, pearson_r, r_squared, effect_estimate, intercept, effect_se, effect_ci_low, effect_ci_high, conclusion, task_id, feasibility, feasibility_reason, data_issues, n_input_rows, n_analysed, random_seed. Plus out/figure1.png (scatter + OLS line), out/table1.csv (income_level, n_countries, mean_log_gdp_pc, mean_life_expectancy), out/report.md (6 sections), out/analysis.py, out/claims.json, out/exclusions.csv, out/state/*.

Information rules. Exclude entities with region == Aggregates. Analysis year = latest year 2015–2023 with both indicators for ≥150 countries. Drop countries missing either indicator in that year (count n_excluded_missing), no imputation. Natural log. Only the listed files.

Invalidity / impossibility. No year with ≥150 complete countries; inability to identify aggregates; non-positive GDP (log undefined); duplicated country-years. Any of these → premise_invalid / partial.