Experimental Evidence on Distributional Effects of Head Start
Bitler, M. P., Hoynes, H. W., & Domina, T. · 2014
grade Bquasi-experimentindependentunreplicated
Sample
4,667 (HSIS)
Population
HSIS sample analysed by quantile.
Design
Quantile treatment effects across the outcome distribution rather than mean impacts.
Key findings
Head Start's effect is concentrated at the BOTTOM of the distribution (>1 SD) and near-trivial at the top. This reframes fadeout: the mean effect disappears partly because schooling remediates the same floor for the control group. For achievement, fadeout is largely control-group CATCH-UP — though for IQ specifically the mechanism is treatment-group DECLINE (Protzko).
Counterfactual
25% of HSIS controls attended a non-Head Start centre.
Genetic confound
Experimental.
Replication notes
Supports a floor-remediation reading of Head Start: it lifts the bottom of the distribution, and floors get remediated for everyone once school starts.
DOI / URL
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| PPVT quantile treatment effects | SD | more than a FULL SD at the bottom of the distribution, falling to ~0.25 SD at the top | standardized | end-of-treatment | business-as-usual | end-of-treatment | domain-skill |
| PPVT, Spanish speakers | SD | +0.47 | standardized | end-of-treatment | business-as-usual | end-of-treatment | domain-skill |
| Cognitive gains, Spanish speakers, through grade 1 | SD | PERSIST — the one durable cognitive subgroup finding in the study | standardized | grade-3 | business-as-usual | over-2yr | domain-skill |
| Non-cognitive outcomes across quantiles | SD | little to no effect at any quantile | mixed | end-of-treatment | business-as-usual | end-of-treatment | non-cognitive |
Cited by
- Fadeout — why early gains disappear, and what actually persistsstrong supportconf: highgc: low
- Preschool at scale — Head Start, state pre-K, and what universal provision deliversmixedconf: highgc: low