This paper combines multiple sources of information on early childhood development in a unified model for analysis of a wide range of early childhood policy interventions. We develop a model of child care in which households decide both the quantities and qualities of maternal and nonmaternal care along with maternal labor supply. The model introduces a novel parenting-effort channel, whereby child care subsidies that permit less parenting may enable better parenting. To estimate the model, we combine observational data with experimental data from the Infant Health and Development Program (IHDP) which randomly assigned free child care when the child was 1 and 2 years old. We estimate a cognitive skill production function and household preferences, giving insight into mechanisms driving the ex post heterogeneous effects of the IHDP intervention, accounting for alternative care substitutes available to the control group and spillovers of the child care offer across the household's decisions. We also estimate ex ante effects of counterfactual policies such as an offer of lower-quality care, requiring a co-pay for subsidized care, raising the maternal wage offer, or a cash transfer. Finally, we use the model to rationalize existing evidence from outside the US on the effects of universal child care programs. (author abstract)