The objective is to empirically test the incentives associated with a Medicaid capitated mental health carve-out contract, whether outpatient services (less expensive, inside the contract) and residential treatment center care (costly care, outside of the contract) were substituted for inpatient psychiatric hospitalization used by children and adolescents. Data sources include Medicaid fee-for-service (FFS) claims for the non-capitated comparison sites and for residential treatment center use, and "shadow billing" encounter data for the experimental capitated managed care sites that provided public mental health services for children and adolescents with Medicaid insurance statewide in Colorado from September 1994 to June 1997. Two part least squares regression models are used to decompose services. Managed care sites are compared to sites that remained under FFS financing, before and in two post-periods after the carve-out. Principal findings show that children and adolescents who received mental health services from a capitated managed care provider were significantly less likely to receive inpatient care, and significantly more likely to receive residential treatment center care. In addition, insurance contract design contains financial incentives that affect the amount and mix of clinical care provided to clients by risk-bearing provider agencies. Findings provide evidence of cost substitution from inpatient care both inside the specialty system and outside the carve-out to other child-serving systems.