We examine the effect of air pollution from power production on students' cognitive outcomes by leveraging year-to-year production variation, wind patterns, and plant closures. We find that every one million megawatt hours of coal-fired power production decreases student performance in schools within ten kilometers by 0.02 standard deviations. Gas-fired plants exhibit no such relationship. Extrapolating our results nationwide indicates that the decline in coal use over the last decade raises test scores by 0.008 standard deviations and reduces the black-white test score gap by 0.006 standard deviations. The nationwide effect obscures substantial spatial variation: The respective numbers for the Midwest are 0.016 and 0.023.