The debate over public provision of early childhood education (ECE) programs has been much in the news lately.  Central to the debate over whether federal and state governments ought to support ECE programs is a question about the payoffs of providing these programs. A full accounting of costs and benefits should include “downstream” as well as immediate impacts of programs. This project uses plausibly exogenous variation in Head Start access and uptake to examine a previously unstudied source of downstream program outcomes: effects on the children of adults who participated in Head Start in their own childhoods. Specifically, I test whether Head Start has long-term impacts on the parenting style of adults (Generation 1 [G1]) who attended Head Start in their youth, and whether benefits (or disadvantages) accrue to the children of former Head Start participants (Generation 2 [G2]) in their early childhood years.

I pool analyses from several nationally representative datasets, including the Panel Study of Income Dynamics (PSID) and the National Longitudinal Survey of Youth (1979) (NLSY-79, linked to the NLSY-79 Children and Young Adults dataset). Both provide data on G1 parents’ Head Start participation, as well as county of residence in parents’ youth (at birth or in early childhood years); both datasets include G1 sibling pairs; both datasets include measures of parenting outcomes for a sizable subset of parents; and both datasets include measures on early childhood outcomes for G2 children. Parenting outcomes include measures of cognitive and emotional stimulation in the home environment as well as parenting attitude scales (e.g., scales of parental self-efficacy and parental warmth in the PSID; scales of parental disciplinary attitudes in the NLSY). Child outcomes include measures like birthweight, early cognitive and behavioral assessments, grade progression, and academic self-concept.

I adapt two empirical strategies used in prior work on the long-term impact of Head Start to address concerns about non-random selection into the program. The first strategy uses sibling fixed effects to explore whether G1 sibling differences in Head Start use are also associated with sibling differences in G1 parenting differences and the early-childhood outcomes of their G2 children. The intuition is that families are unlikely to selectively enroll their G1 children based on child characteristics; it is more likely that idiosyncratic differences drive differences in enrollment, such as a family move meaning that the family was proximate to a Head Start center when one child was age-eligible but not proximate for the other.

The second strategy exploits a 1965 program that provided grant-writing assistance for Head Start funding to the poorest 300 counties—and no other counties—based on 1960 census measurements of county poverty rates. This policy can be leveraged in a regression discontinuity design (Ludwig & Miller, 2007) to explore whether children whose G1 parents lived in one of the counties that received technical assistance have systematically better or worse outcomes than children whose parents lived in counties just missing the grant-writing aid.

Results will provide important new evidence about long-term costs and benefits to Head Start attendance.