We exploit a rich administrative panel data-set for cohorts of Economics students at a UK university in order to identify causal effects of class absence on student performance. We utilise the panel properties of the data to control for unobserved heterogeneity across students and hence for endogeneity between absence and academic performance of students stemming from the likely influence of unobserved effort and ability on both absence and performance. Our estimations also exploit features of the data such as the random assignment of students to classes and information on the timetable of classes, which yield potential instruments in our identification strategy. Among other results, we find that there is a causal effect of absence on performance for students: missing class leads to poorer performance. There is evidence from a quantile regression specification that this is particularly true for better-performing students, consistent with our hypothesis that effects of absence on performance are likely to vary with factors such as student ability.