As the only large-scale randomized experiment in Head Start history, the Head Start Impact Study (HSIS) included a nationally representative sample of 3- and 4-year-old children who were eligible for Head Start. The HSIS reported short-term benefits of Head Start after one-year participation, while few benefits were maintained in kindergarten, first grade, or third grade.

This study uses the HSIS data to examine the effects of Head Start on children who otherwise would have no access to center-based care. Children in the HSIS control group received a mixture of alternative care arrangements, including another Head Start program, other center-based care, relative/non-relative care, and parental care. Research has shown that the type and quality of care arrangements are closely related to children’s developmental outcomes. A preliminary analysis of the HSIS data finds that the effects of Head Start after one year of the HSIS random assignment vary substantially depending on the counterfactual care arrangement. Head Start has generally large, favorable effects on cognitive and behavioral outcomes when compared to parental care and relative/non-relative care, while few differences are found between Head Start and other center-based care.

Using the currently available HSIS data, this study investigates whether the favorable effects of Head Start on children who otherwise would have parental care or relative/non-relative care are sustained after one year of the HSIS random assignment, including at age 4 for the 3-year cohort and in kindergarten and first grade for both 3- and 4-year cohorts. The cognitive outcomes available in these waves of data were collected from direct assessments, including Peabody Picture Vocabulary Test (PPVT) Receptive Vocabulary and Woodcock-Johnson III Letter-Word Identification and Applied Problems. The behavioral outcomes were reported by parents, including Aggressive Behavior, Hyperactive Behavior, Withdrawn Behavior, and Behavior Problem. The analysis sample in preschool has a total of 3,932 children, including 2,198 in the 3-year cohort (1,360 in treatment and 838 in control groups) and 1,734 in the 4-year cohort (1,065 in treatment and 669 in control groups).

There is an issue of self-selection because children in the control group chose different care arrangements due to variation in pretreatment covariates. Similarly, children in the treatment group would also have chosen different arrangements if they had been assigned to the control group. Since the randomization in the HSIS was conducted over the eligibility to enroll in Head Start but not other specific care arrangements, a direct comparison of Head Start to other specific arrangements would be internally invalid. This study uses principal score matching to first estimate the probabilities of choosing different care arrangements for children in the control group, and then applies the predicted parameters to the treatment group to estimate the probabilities of having these arrangements for Head Start participants if they had been assigned to the control group. Children with similar principal scores are then matched and their outcomes are compared.

The findings of this study provide important implications regarding whether Head Start would be most beneficial for children who otherwise would have no access to center-based care.