In this paper we estimate relative age effects in academic performance using a unique database of students at Bocconi University. The identification exploits school entry cut-off ages that generate up to 11 months difference between the youngest and the oldest students within each cohort. Our data allow to control for potential selection issues as well as for differences in cognitive ability, as measured by an attitudinal entry test. Contrary to most of the existing evidence for primary school children, we document that in university the youngest students perform better compared to their oldest peers, particularly in the most technical subjects. To rationalize this result we produce additional evidence on relative age effects in cognitive ability and in social behavior using a combination of data from Bocconi admission tests and from a survey on the social behavior of Italian first-year university students. We find that the youngest students in the cohort perform slightly better in cognitive tests and also appear to have less active social lives: they are less likely to do sports, go to discos and have love relationships. These results suggest that negative relative age effects in university performance might be generated by two mechanisms: (i) a profile of cognitive development that might be decreasing already around age 20; (ii) psychological relative age effects that lead the youngest in a cohort to develop social skills (self-esteem, leadership) at a slower pace. Younger students, thus, have less active social lives and devote more time to studying, as confirmed by additional evidence from the PISA study.