Planes, trains, and automobiles: what drives human-made light?

This paper links the newest generation of nighttime satellite images, which offer a resolution 45 times higher than the previous generation, to nationwide administrative panel-data on population and income from the United States and Brazil for the years 2012-2020. Using this fine-grained data, I confirm that nighttime light responds strongly to changes in income even after controlling for population effects. When population is included directly in the model, light is less responsive to changes in GDP in Brazil than in the USA. In Brazil, though not in the USA, except for the highest-producing municípios, the effect of changes in population appear to track more closely with nighttime lights than changes in economic output. A between-county estimator provides identification of the effects of time-invariant characteristics and infrastructure features on night-time light. My estimates suggest that railways are associated with lower levels of nighttime light while border crossings contribute positively and significantly to nighttime light.