Speaker: Marshall Burke, Stanford University
Most developing countries collect little or no data on local-level economic wellbeing, which makes it difficult to evaluate development interventions and to target assistance to those most in need. Here we demonstrate how new high-resolution satellite imagery can be used to measure both smallholder agricultural productivity and asset wealth at a granular level. For the agricultural productivity measurements, using multiple seasons of data from Kenya, we show that plot-level satellite-based estimates of productivity are roughly as accurate as traditional survey based measures. For the wealth measurements, we use data from five African countries and show how a neural network can be trained to identify features that can explain up to 75% of the variation in local-level economic outcomes. Both approaches are inexpensive and scalable, and together could accelerate efforts to understand and improve economic well-being in poor regions.