This paper analyzes how variation in participant take-up rates affected the impacts of the New Hope project, a random-assignment, anti-poverty program. New Hope offered experimental members four benefits--child care subsidies, wage subsidies, health insurance, and, if needed, a temporary community service job--that were available to families working full time. Take-up of the benefits was far from universal and experimental participants who used one of the benefits rarely used all of them. Clustering and propensity score methods are used to analyze take-up sub-groups and to estimate program impacts within each. The majority of take-up patterns adopted by experimental members were associated with at least one positive program impact. However, the primary beneficiaries were those parents who used the community service jobs. They increased their employment effort, felt less parenting stress, and had children with higher teacher-related academic accomplishment scores. The implication of this method for the evaluation of other multi-benefit programs is discussed. (author abstract)