This paper addresses whether changes in expenditures from participation in a Medicaid home and community-based waiver are due to selection effects or true program effects. We examined North Carolina Medicaid claims data (FY1998 to FY2000) for disabled or blind adults likely to need facility-based care using two-part expenditure models for nursing home, inpatient, and total expenditures. Instrumental variables estimation controlled for selection into the program. Although no savings in total costs were found in this sample, significant decreases were found in the level of nursing home and hospital inpatient expenditures for waiver participants, achieving the goals of the home and community-based waiver program.