While there is growing evidence that early childhood prevention programs can have positive outcomes for children and families, research has tended to focus on short-term outcomes, with fewer studies of long-term benefits. In addition, evaluations of such programs rarely go beyond the question, “does the program work?” despite numerous calls over the past decade for more research that can help understand how these programs have their effects. Using longitudinal, experimental data from the Early Head Start (EHS) Research and Evaluation Project (EHSREP) linked to child welfare agency records for 2794 children, we examined the effectiveness of EHS birth-to-three services in preventing child maltreatment during children’s first 15 years of life. Following this, we assessed whether changes in specific child and family outcomes at ages 2 and 3 acted as mediators for later maltreatment prevention. Results showed that EHS has a long-term impact on the likelihood of child welfare system involvement that is driven by earlier impacts on parenting behaviors, family well-being, and child developmental status. By children’s second birthday, families randomly assigned to participate in EHS had lower family conflict and parenting distress, and more positive parent-child interactions; these impacts, in turn, led to later reductions in the likelihood of children being involved with the child welfare system through age fifteen years. Furthermore, at age three, children in EHS were more attentive and engaged in play with their parents and had higher scores on cognitive development assessments compared to controls; these outcomes were similarly associated with long-term reductions in the likelihood of child maltreatment. These findings suggest that early two-generational programs, like EHS, that are able to successfully decrease family conflict and stress and support positive, emotionally responsive parenting and child development, may reduce the likelihood of abuse and neglect later in life.