The Evidence on Teaching

Third Grade Follow-up to the Head Start Impact Study: Final Report (OPRE 2012-45)

Puma, M., Bell, S., Cook, R., Heid, C., Broene, P., Jenkins, F., Mashburn, A., & Downer, J. · 2012

grade Arctindependentunreplicatednumbers spot-checked
Sample
4,667 randomized; grade-3 analysis samples reduced by attrition
Population
Same national HSIS sample followed to grade 3.
Design
Tests impacts across four domains and two cohorts at grade 3, with Benjamini-Hochberg multiplicity correction.
Key findings
By grade 3, across four domains and two cohorts, essentially nothing survives — and the one favourable result fails multiplicity correction while the one unfavourable result concerns grade promotion. Both arms remain ~0.5 SD below national norms, which is the more important fact: Head Start did not close the gap it was built to close.
Counterfactual
By grade 3 there is no schooling difference: 96-98% of both arms in public school with similar school quality.
Genetic confound
RCT.
Replication notes
HSIS has NO adulthood follow-up, so it cannot test the Boston-style 'attainment moves while test scores don't' pattern. Do not read the grade-3 null as a verdict on adult outcomes.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Cognitive domain, 4yo cohort, grade 3ITT SDONE surviving impact: ECLS-K Reading +0.11, SUGGESTIVE ONLY (p<=.10), fails Benjamini-Hochbergstandardizedgrade-3business-as-usualover-2yrdomain-skill
Cognitive domain, 3yo cohort, grade 3ITT SDONE surviving impact: parent-reported grade promotion -0.11 — UNFAVOURABLEresearcher-designedgrade-3business-as-usualover-2yrattainment
Math, both cohorts, grade 3ITT SDzerostandardizedgrade-3business-as-usualover-2yrdomain-skill
Health domain, grade 3ITTzero — all early favourable health impacts goneresearcher-designedgrade-3business-as-usualover-2yrhealth
Absolute standing vs national norms at grade 3SD below national meanBOTH arms ~0.5 SD below a national grade-3 samplestandardizedgrade-3noneover-2yrdomain-skill

Cited by