An important gap in the literature on two-generation early intervention programs, which provide case-management services to low-income parents and center-based child care to their very young children, is understanding which intervention components (mediators) have the largest impact on parental education and economic self-sufficiency outcomes. This article helps fill the gap by rigorously examining the relative effects of providing case-management referrals and subsidized center-based child care using randomized controlled trial data from the Early Head Start Research Evaluation. The analysis uses recent instrumental variable statistical methods to identify mediator effects by linking experimental impacts on the mediators and parent outcomes across 17 sites. The key finding is that parents’ receipt of case management appears to be the most important explanation for cross-site variation in impacts on parental human capital and self-sufficiency outcomes. (author abstract)