Advancements in satellite imagery and machine learning can be used to infer the causal impact of electricity access on livelihoods, providing a low-cost, generalizable approach to evaluating public policy in data-spare environments.