Using publicly available satellite imagery and deep learning to understand economic well-being in Africa

Accurate and comprehensive measurements of economic well-being are crucial for research and policy, but such data are often unavailable at the local level in many parts of the world. This study uses deep learning models trained on publicly available multispectral satellite imagery to predict survey-based estimates of asset wealth across 20,000 African villages. The models successfully explain a significant portion of the variation in ground-measured village wealth, even in countries where the model was not trained. Satellite-based estimates also show promise in tracking changes in wealth over time. The study demonstrates the potential of satellite imagery for research and policy applications, showcasing its scalability by generating a wealth map for Nigeria.

Christopher Yeh, Anthony Perez, Anne Driscoll, George Azzari, Zhongyi Tang, David Lobell, Stefano Ermon, Marshall Burke