We analyze the effectiveness of a vocational training (VT) programme targeting unemployed youth in Latvia, contributing to the scant literature on active labour market policies in transition countries. The programme we analyse is part of the Youth Guarantee scheme (2014–2020), the largest action launched by the European Union to combat youth unemployment after the 2008 financial crisis. Although the programme was targeted to youths aged between 15 and 29, priority was given to those younger than 25 years of age. We exploit this eligibility rule in a fuzzy regression discontinuity design framework to estimate the impact of VT participation on the probability of being employed and gross monthly labour income at given dates after the training. Using rich administrative data, we find that the age priority rule increased programme participation for the youngest group by about 10 percentage points. However, participation in the programme did not lead to statistically significant positive effects in labour market outcomes. We argue that this result could be due to some specific characteristics of the programme, namely the voucher system (potentially inducing lock-in effects) and the type of training (classroom instead of on-the-job training). Moreover, the programme was targeted at ex-ante low-employable individuals (e.g. without vocational qualifications), a fact that is confirmed by our analysis of the characteristics of the population of compliers with the age priority rule.