Schools in the shadow of toxic sites: Pollution proximity in low- and middle-income countries — replication package

Replication package for: Crawfurd, L. (2026). "Schools in the shadow of toxic sites: Pollution proximity in low- and middle-income countries." Combines geocoded data on 2,840 contaminated sites across 17 countries (Pure Earth Toxic Sites Identification Program combined with national government contaminated-site registers in India CPCB, Brazil São Paulo CETESB, Mexico RETC, and Peru PAM) with the locations of about 2.6 million schools (national EMIS data for 11 countries; Overture Maps for 6 additional countries). Contents: The full package (schools-pollution-replication-v3.zip) contains code (a single sequential Python pipeline, plus Stata and R cross-language replication), raw input data, processed intermediate datasets, and all paper figures and tables. README.md documents the structure, replication steps, and data availability. v3 changes: Restructured the code into one sequentially numbered pipeline (01–28) run end-to-end by run.py — the former code/build/ table-building subfolder was flattened into code/, the per-country EMIS proximity scripts were merged into one (04_emis_proximity.py), and the three PM2.5 scripts into one (08_pm25_analysis.py, single raster load); superseded scripts were removed (no Archive/ folder). Portability: paths resolve from the package root, with an optional SCHOOLS_OVERLEAF_DIR environment variable for Overleaf mirroring (no-op when unset) and a single REPLACE_WITH_PACKAGE_ROOT token in the Stata/R scripts; run.py enforces Python ≥ 3.10. Tables 2 and 3 now report enrolled children in millions within 1 km and 5 km (presentation only). All outputs verified byte-identical to v2 — no analytical results change. v2 changes (relative to v1): Added national government contaminated-site registers for four countries; added Argentina and Mexico EMIS data; expanded the appendix with TSIP-versus-register robustness checks (tab_wealth_gradient_robustness) and a population-baseline comparison using WorldPop (tab_population_baseline).