Replication package for "What do payments to waste pickers buy? Landfill diversion and formalization in Colombia"

Data and code for the article “What do payments to waste pickers buy? Landfill diversion and formalization in Colombia” by Daniel D. Otero Meza and Fernando Vidal-Barrero (2026). The package rebuilds a municipality-year panel for 1,104 Colombian municipalities over 2010–2024 (16,560 municipality-years, 316 of them adopting the scheme) from public administrative records, and reproduces every estimate reported in the tables and figures of the article and of its supplementary material. Contents of replication.zip data/panel_2010_2024.csv: the analysis panel, with the recovered tonnage reported for payment, the year each municipality joined the tariff-financed payment to waste-picker organizations, final disposal before and after the audit, population, the value of recovered material and baseline covariates. data/inputs/ and data/raw/: the records used to build the panel. These are recovery and final-disposal records and recovery stations from the SSPD (SUI), municipal fiscal categories, DANE population projections and consumer prices, IGAC municipal boundaries, ACOPLASTICOS prices of recyclables, the market exchange rate and US CPI, and the price links and local price checks compiled by the authors. code/: four Jupyter notebooks. 00 creates a Python environment inside the package. 01 audits the disposal records and rebuilds the panel. 02 estimates the effects of adoption with Callaway and Sant’Anna difference-in-differences, the event studies and the diversion ratio θ. 03 runs the robustness checks, including the imputation and Sun and Abraham estimators, randomization inference, covariate overlap and HonestDiD bounds. run_notebooks.py runs 01 to 03 without opening them. README.md: instructions, running times, data sources and a table that links each table and figure of the article to the result file that contains its numbers. How to run. Unzip the package into a short path outside synchronized folders. Then open code/00_environment.ipynb with any Python 3.11 to 3.14 and run it, and run notebooks 01 to 03 in order (about 15 minutes once the environment is installed). Everything the notebooks create is written inside the package folder, in .venv/ and outputs/. The package runs on Windows, macOS and Linux. Disposal audit. Non-reports are treated as missing rather than zero, and modeled records are flagged. Values outside 5–1,500 kg per capita-year, or more than ten times above or below the municipality’s own median, are set to missing. The audit log lists every municipality-year set to missing, with its reason and source label. Licenses. The code is released under the MIT License. The data compiled by the authors (panel, price links and checks, and results) are released under CC BY 4.0. Third-party records remain under the terms of their issuers and are included only to make the analysis reproducible (see LICENSE).