High school students from non-elite backgrounds are less likely to have peers with elite educated parents than their elite counterparts in Norway. We show this difference in social capital is a key driver of the high intergenerational persistence in elite education. We identify a positive elite peer effect on enrolment in elite programmes and disentangle underlying mechanisms. Exploiting a lottery in the assessment system, a causal mediation analysis shows the overall positive peer effect reflects a positive effect on application behaviour (conditional on GPA), which dominates a negative effect on student GPA. We consider implications for income mobility finding that encouraging further mixing between elite and non-elite students in high school could improve mobility across the whole distribution.