We design a labour market experiment to compare demand-side and supply-side policies to tackle youth unemployment, a key issue in low-income countries. The experiment tracks 1700 workers and 1500 firms over four years to contrast the effects of offering workers vocational training (VT) to offering firms wage subsidies to train workers on-the-job (FT). Both treatments lead to skill accumulation but whilst VT workers learn sector-specific skills, FT workers learn more firm-specific skills. This is associated with higher employment rates for each type of worker but the effect is 50% larger for VT (21% vs 14%) and their total earnings increase by more (34% vs 20%). Structurally estimating a job ladder model reveals the mechanisms: VT workers receive higher rates of unemployment-to-job offers and higher rates of job-to-job offers. This greater labour market mobility stems from the certifiability and transferability of their skills, and causes the wage pro…les of VT workers to diverge away from FT workers. Evidence from the firm-side of the experiment complements these findings: we find that some of the higher returns to VT are driven by workers matching to higher productivity firms. Our evidence shows both …rms and workers are constrained in this setting and that subsidies to either side of the labour market would increase workers’ employment and earnings. However, VT workers are better o¤ than FT workers as the greater certifiability and transferability of their skills allows them to climb the job ladder more quickly.