This study investigates the effect of drought on economic activity globally using remote sensing data. In particular, predicted variation in greenness is correlated with changes in the density of artificial light observed at night on a grid of 0.25 degree latitude-longitude pixels. I define drought as greenness estimated by lagged variation in monthly rainfall and temperature. This definition of drought performs well in identifying self-reported drought events since 2000 compared with measures of drought that do not take greenness into account, and the subsequent analysis indicates that predicted variation in greenness is positively associated with year-on-year changes in luminosity: If a unit of observation experiences a predicted variation in greenness that lies 1 standard deviation below the global mean, on average 1.5 - 2.5 light pixels out of 900 are extinguished that year. Finally, an attempt is made to estimate the global cost of drought.