Over the last decades, ambulatory psychosocial interventions aimed at treating child externalized behavior problems have been developed and tested. In recent years, however, it has become clear that the reported effects of psychotherapies have been considerably overestimated due to publication bias. In addition, previous meta-analyses of psychosocial interventions aimed at treating child disruptive behaviors have not accounted for potential interaction effects among moderators. The purpose of this meta-analysis was to examine the effects of interventions while accounting for publication bias and dependencies among moderators. We searched electronic databases to identify randomized controlled trials targeting child externalized behavior problems treated in ambulatory settings. A total of 111 randomized controlled trials with 11,623 participating children were analyzed. Interventions targeting children with symptom severity below the clinical range demonstrated a non-significant effect (adjusted g = 0.12), while interventions targeting children with symptom severity within the clinical range demonstrated a small to moderate effect size (adjusted g = 0.38) that was sustained at follow-up (adjusted g = 0.27). We found the highest pooled effect size for individually administered treatments involving both the parent and the child (g = 1.11; no indication of publication bias). No other differences were found between delivery formats, including assisted self-help interventions. Publication bias was relatively smaller in high-quality studies. These findings suggest that the most effective treatment regime involves both the parent and the child, however, only when they are treated in an individual setting. Our results highlight the importance of accounting for interaction effects, publication bias and trial quality in future meta-analyses.