The Evidence on Teaching

Adult Outcomes as a Function of an Early Childhood Educational Program: An Abecedarian Project Follow-Up

Campbell, F. A., Pungello, E. P., Miller-Johnson, S., Burchinal, M., et al. · 2012

grade Brctdeveloper-ledmixednumbers spot-checked
Sample
111 (57 T / 54 C); 101 assessed at age 30 (91% retention)
Population
Low-SES African-American children, Chapel Hill NC, entering at mean age 4.4 MONTHS; 8h/day, 5d/wk, 50wk/yr to school entry.
Design
The most intensive early-childhood RCT ever run (~$18,514/participant-year in 2014 USD; ~$92k total). Treatment also included two daily meals plus health care, so it is not a pure 'education' contrast.
Key findings
The most intensive early intervention ever tested moved attainment substantially (BA+ 23% vs 6%) and welfare receipt, but was NULL on income-to-needs and NULL on crime. Its small durable IQ effect sits at, not above, the environmental ceiling seen in adoption studies. Note the absolute numbers: the headline BA result is 12 people versus 3.
Counterfactual
NOT nothing: controls received iron-fortified formula, free diapers and social services; ~75% used lower-quality alternative childcare.
Genetic confound
Spitz's critique: the treatment group already led on the Bayley MDI at SIX MONTHS by roughly the eventual gap, suggesting baseline imbalance rather than a treatment effect on IQ. Ramey's reply is substantially ad hominem.
Replication notes
Abecedarian's ~+4.4 IQ points at 21 is the only durable IQ effect from a preschool RCT — but it is NOT unique in the literature: full adoption produces +4.41 (Kendler 2015) and schooling ~1-5 points/year (Ritchie & Tucker-Drob 2018). ~3-5 points looks like the general ceiling for whole-childhood environmental manipulations.
Cost
~$92,000 per participant total — the most expensive intervention in the database.
DOI / URL
10.1037/a0026644

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Years of education completedyears13.46 vs 12.31 (significant)mixedadultactive-alternativeadulthoodattainment
Bachelor's degree or higher by 30%23.1% vs 6.1% — i.e. roughly 12 people vs 3 peoplemixedadultactive-alternativeadulthoodattainment
Criminal conviction by 30% ever convicted27.5% vs 28.6% — NULL. Directly fails to replicate Perry's crime channelmixedadultactive-alternativeadulthoodbehaviour
Income-to-needs ratio at 30ratioNULLmixedadultactive-alternativeadulthoodattainment
Public assistance receipt at 30%3.9% vs 20.4%mixedadultactive-alternativeadulthoodattainment
IQ trajectory (wider Abecedarian literature)IQ points T-C~1.08 SD at 18mo-4.5yr; shrinks DURING treatment; ~+4.7 at 12; ~+4.4 at 21standardizedadultactive-alternativeadulthoodg

Cited by