This multilevel meta-analysis examined the effects of treatment for juveniles with harmful sexual behavior on psychosocial functioning, and the potential moderating effects of outcome, treatment, participant, and study characteristics. In total, 23 studies, comprising 31 independent samples and 1342 participants, yielded 362 effect sizes (Cohen's d). A moderate overall effect size was found of d = 0.60, indicating that groups receiving treatment achieved an estimated relative improvement in psychosocial functioning of 33%. Type of outcome did moderate the effect of treatment, indicating that effects on atypical sexual arousal and empathy (a trend) were smaller, compared to effects on other outcomes. Most prominently, studies of weak quality produced larger effect sizes. Unexpectedly, non-established treatments had more effect than did established treatments, which may be explained by the use of less rigorous study designs. Treatment groups with a higher percentage of juveniles with similar age victims or mixed type problem behavior also yielded larger effect sizes. Lastly, evaluation of treatment effects by professionals produced higher effect sizes, compared to other sources of information (e.g., adolescent self-report). Although only a marginal to no indication was found for publication bias by means of funnel plot analysis of the distribution of effect sizes, articles published in peer reviewed journals showed relatively large effect sizes. Implications for future research and clinical practice are discussed.