Daily changes in human activity are difficult to detect using nightlight imagery because many factors that influence nightlights are changing from night to night. We propose using a difference-in-differences methodology for detecting daily changes in human behavior using NASA’s Black Marble product suite. We find that total top-of-atmosphere radiance on the Indonesian island of Bali decreases by close to 100% during the Hindu holiday of Nyepi, relative to counterfactual islands. We then use changes in total radiance during Nyepi to predict census information on religion in this region of Indonesia; nighttime imagery can explain 78% of the variation in religion at the equivalent of the county level, and 62% of the variation at the equivalent of the zipcode level. We use the same method to correctly identify pockets of large Hindu populations on the predominantly Muslim islands of Java and Lombok.
