The Evidence on Teaching

Evaluating Public Programs with Close Substitutes: The Case of Head Start

Kline, P., & Walters, C. R. · 2016

grade Bquasi-experimentindependentreplicatednumbers spot-checked
Sample
4,667 (HSIS reanalysis)
Population
HSIS sample, decomposed by what the control group actually did.
Design
Decomposes the LATE into sub-LATEs by counterfactual care mode. CAVEAT: the partial first stage for the centre-substitution margin is weak (F=1.8).
Key findings
The most important reframing in the Head Start literature: Head Start beats staying home (~+0.37 SD) but is statistically indistinguishable from the other preschools it displaces. As alternative childcare has become near-universal, the measurable effect of *adding* Head Start has shrunk accordingly — much of the apparent 'fadeout' at scale is really a counterfactual problem.
Counterfactual
THE POINT OF THE PAPER: without an offer, 55% home care, 31.5% other centre, 13.6% Head Start anyway; 34% of compliers came from competing, mostly publicly funded preschools.
Genetic confound
Experimental base; the confound at issue is the counterfactual, not genetics.
Replication notes
Independently corroborated by Weiland et al. (2020): where controls already attend centre-based care, measured effects go to zero on academic outcomes.
DOI / URL
10.1093/qje/qjw027

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Head Start vs HOME care (sub-LATE)SD~+0.37, stable across specificationsstandardizedend-of-treatmentbusiness-as-usualend-of-treatmentdomain-skill
Head Start vs OTHER CENTRE-BASED preschool (sub-LATE)SDnever significantly different from ZERO; slightly negative in places (weak first stage, F=1.8)standardizedend-of-treatmentactive-alternativeend-of-treatmentdomain-skill
Marginal value of public fundsMVPF1.10 ignoring substitute-program savings; higher once they are countedunknownadultunclearadulthoodattainment

Cited by

Evaluating Public Programs with Close Substitutes: The Case of Head Start · The Evidence on Teaching