By studying population-wide panels of 10th graders in Norway, I investigate the association between the social class composition of cohorts in schools and choosing an academic as opposed to a vocational track at upper-secondary level. Two steps are used to study this relationship. First, I use multilevel models estimating the group level model simultaneously with the regression of the individual-level model. Second, by using 11 panels of 10th graders, I employ school fixed effects to get closer to an estimate of school compositional effects happening within the school. Both models find that the proportion of upper-class peers at lower-secondary schools is associated with the increased likelihood of choosing academic tracks at the upper-secondary level, and especially for students who are not from upper-class backgrounds. This suggests that the 'classed' nature of educational decision-making is also embedded in contexts beyond familial ones.