We analyse a rich dataset of Economics students at a UK university to identify causal effects of class absence on student performance, exploiting the random assignment of students and information on students' class timetables to avoid selection problems. We use panel properties of the data to control for unobserved student factors such as ability and effort. Quantile regression results suggest that absence has adverse effects on performance--but these effects are causal only for the upper part of the conditional performance distribution. (Contains 5 tables and 3 figures.)