School-based social and emotional learning (SEL) programmes represent a practical method of improving social and emotional well-being in students. To date, however, what is less well understood is why a theoretically sound, appropriately administered, engaging universal SEL programme may be more effective for some children over others. In the present study, an established SEL programme, KooLKIDS Whole of Class, was delivered to 524, 8 to 12-year-old Australian primary school students. Classroom teachers completed pre- and post-measures regarding children's social-emotional competence and emotional and behavioural difficulties. Hierarchical linear modelling was used to examine which characteristics predicted differential improvement from the programme. Findings revealed that the pre-programme total score on emotional and behavioural difficulties was the strongest predictor of improvements in SEL, independent of age, sex or socioeconomic status. This implies that while universal SEL programmes represent an effective and relatively low-cost method of improving social competence for all children, they may be particularly effective for improving social and emotional competence in those with greater emotional and behavioural difficulties.