To analyze whether there are causal peer effects in criminal behavior among young students in upper secondary education. We follow four cohorts of Danish students (N=27,525) when they complete 9th or 10th grade and start upper secondary vocational education with new peers. The exogenous variation in the proportion of new peers with criminal records between cohorts within the same school and program is used to identify causal peer effects. Our analysis finds important and significant nonlinear peer effects. Students who enter a program with new peers who have a high crime propensity are 1.8 percentage points more likely to be charged with a criminal offence within 12 months of commencing vocational education. However, students with a history of repeated offenses, who enter a program with new peers with a low crime propensity, reduce their probability of future offending from 31.5 percentage points to 18 percentage points. The results support the notion that students in upper secondary vocational education might build criminal capital by interacting with schoolmates with prior charges. Yet, our study also shows that a new low-crime peer group in upper secondary education can have a positive influence to students with a history of criminal offenses.