Identifying the predictors of problem behavior is essential both for understanding the causes of such behavior and for preventing it. Although a great deal of research has sought to identify the factors predictive of problem behavior, much of the research to date has been correlational and tells us little about causality. This study attempts to improve on the correlational research by applying meta-analytic techniques to existing experimental and quasi-experimental studies of school-based prevention. The following 3 risk factors were examined: academic performance, bonding to school, and social competency skills. The most convincing evidence of a relationship between risk and problem behavior was found for bonding to school. Positive changes in attachment and commitment to school resulting from the preventive interventions were consistently accompanied by positive changes in problem behavior. Preventive interventions that produced improvements in academic performance produced moderate improvements in problem behavior. With regard to social competence, the association depended in large part on the type of measure used to assess social competency skills. Changes in self-report measures of social competency were unrelated to changes in problem behavior, whereas a strong positive correlation was observed between changes in ratings and observations of social competency by others and improvements in problem behavior.