Public Investment and Development Outcomes in Colombia

This project started with a wide-open mandate: find administrative data on development outcomes and public financing across Latin America. I spent weeks mapping data availability across more than ten countries, government portals, international grants, project financing, before we converged on Colombia’s Terridata as the richest source available.

From there I built a dataset from the ground up, at the territorial level, combining multiple public-financing sources with population and development-outcome data, the kind of dataset that let us test theories like the O-Ring Theory of development. I pulled the data through web scraping and APIs in Python, then cleaned and structured it in Stata.

With outcomes spanning education, health, and water and sanitation, I managed iterative research cycles across a range of identification strategies, OLS, instrumental variables, LASSO regressions, and machine-learning models, running robustness checks and design adjustments to see which relationships held up, and contributing to both the analytical and writing stages.

I also documented the work as we went, fully commented analytical scripts, data visualization outputs, and technical memoranda that laid out our methodological decisions clearly, both for the research team and for broader audiences following the project.