This paper evaluates a socio-emotional skills training programme for 4,728 urban Tanzanian youth who were not in full-time employment, education, or training. A randomized design compared awareness (for example, self-awareness, empathy, and active listening), management (for example, self-control, personal initiative, and negotiation), and combined curricula. Socio-emotional skills were measured using self-reported and behavioral indicators. Training increased self-reported socio-emotional skills in the short run across both domains, but had limited effects on behavioral measures, and all socio-emotional skills gains faded after one year. Modest but sustained employment gains were observed among men who were job seekers at baseline. Training did not improve labor market outcomes for women. No differential effects were found across training types, and each training affected skills in the alternative domain, suggesting that socio-emotional skills domains are interrelated. These findings indicate that socio-emotional skills training may improve labor market outcomes only for specific subgroups, particularly in the absence of complementary interventions tackling barriers to employment.