A High Resolution Look at Long Run Development: Evidence from 1.3 Million Historical Aerial Photographs
Presented by:
Joel Ferguson
Nelson Institute for Environmental Studies
University of Wisconsin-Madison
Friday, February 13, 2026
12:00 pm-1:15 pm
Taylor-Hibbard Seminar Room (Rm103)
To understand how to promote economic growth, alleviate poverty, and reduce inequality, we need measures of wealth that are both disaggregated over space and available over long time periods. However, reliable sub-national data on wealth only go back to 1990 for many developing countries, limiting the set of important policy questions can be answered using existing data. To begin to fill this gap, we apply frontier machine learning methods to a newly digitized archive of 1.3 million historical aerial photographs to produce a novel 5 km resolution dataset of population density and wealth per capita for 18 African countries covering the period from 1940-1990. Our machine learning pipeline combines transfer learning, in which the model learns the relationship between satellite imagery and outcomes in the present to have a good starting point for learning the relationship in the past, with subjective pairwise rankings of historical images generated by human annotators. Our models exhibit excellent performance in both tasks, achieving test r2s of 0.86 for population and 0.85 for wealth. We validate performance on historical imagery using a collection of historical censuses and similar administrative records, such as tax revenues, that we have digitized and georeferenced.