Predicting Community-Level Poverty in Bolivia: Insights from Satellite Imagery, Census Data, and Spatial Modeling
This repository contains the code for the study: "Predicting Community-Level Poverty in Bolivia: Insights from Satellite Imagery, Census Data, and Spatial Modeling".
This project develops and implements a methodology to predict community-level poverty headcount ratios in Bolivia for 2022 by integrating machine learning, remote sensing (satellite imagery), and spatial modeling techniques. Focusing on Unsatisfied Basic Needs (UBN) poverty in 953 communities from 2012 to 2022, the study documents a broad decline in poverty rates — with approximately 50% of communities projected to fall below the 41.8% threshold by 2022. Poverty reductions are found to be especially significant in communities with initially lower poverty levels, although regional and urban–rural disparities persist.
The approach highlights the value of combining machine learning with geospatial data to support more precise, targeted poverty reduction strategies in Bolivia. Moreover, this methodology offers a scalable and replicable framework for other developing countries facing limited and outdated high-resolution spatial data, providing actionable insights for policymakers aiming to address poverty at a granular level despite data constraints.