We analyze the effect of school start time on academic performance. Sleep patterns are determined in part by sunrise times, which vary across time zones. Because school start times do not fully reflect this difference, we instrument for the hours of sunlight before school with the time zone boundary in Florida. We find that moving start times one hour later relative to sunrise increases test scores by 0.08 and 0.06 standard deviations for adolescents in math and reading, respectively. In math, the effect is larger for older children and co-varies with entry into an important pubertal stage. School districts can improve performance while maintaining the current distribution of start times by moving classes earlier for younger children and later for older children.