The purpose of this paper is to analyze whether there are peer effects in criminal behavior among young students. Conventional wisdom says that education is a fast track out of delinquency for young individuals at an early stage of a possible long criminal career. However, if a young delinquent induces his peers to also become involved in criminal activity, this wisdom may be questioned and indicate a need for further supporting initiatives to young delinquents. A causal peer effect at school may also work in the opposite direction, whereby adolescents on the straight and narrow path induce delinquents to enter that path and stay out of criminal activity. We analyze data from four cohorts of Danish students (N= 27,525) when they complete lower secondary school (9th/10th grade) and commence in an upper secondary vocational education program with new classmate peers. We use the change of school between lower and upper secondary school and the exogenous variation in the composition of classmates across adjacent cohorts within the same colleges to identify causal peer effects. The results show significant nonlinear causal peer effects, as students who enter a program alongside many classmates with a high crime propensity are more likely to be charged with a criminal offense within 12 months from commencing vocational education. At the same time, we also find that students with a history of repeat offending, who enter a program with new classmates who have a low degree of prior charges, benefit from this low-crime peer group.