This paper compares the relative effectiveness of selected active labour market policies available to young unemployed people in Poland in the years 2015-2016. We use rich administrative data and propensity score matching techniques to control for the non-random selection of unemployed individuals into alternative interventions. We find large negative employment effects of participating in public works programmes, particularly among disadvantaged individuals. The differences in effectiveness between other interventions are rather small, and most become insignificant over time. We also show that vouchers that allow unemployed individuals find on-the-job training providers themselves are more effective than on-the-job training schemes in which the unemployed individuals are directed to the training providers by the public employment services (PES).