We investigate the determinants of teacher absences both within and across schools. We find that teachers generally respond to increased workload by decreasing their rate of absence. Teachers are less likely to be absent when they are teaching larger classes, have new grade assignments or have fewer years of experience. Moreover, we show that when teachers change schools, their absence rate quickly gravitates towards the mean absence rate of their new school, suggesting that school-level factors are an important determinant of absence rates. Finally, we show that the inverse relationship between workload and absence may lead researchers to underestimate the ceteris paribus effect of certain teacher inputs. We illustrate this point in the context of estimating the effect of teacher experience on test scores and show that controlling for absence rates increases the estimated returns to experience by approximately 10%.