The Evidence on Teaching

Lifetime Effects: The HighScope Perry Preschool Study Through Age 40

Schweinhart, L. J., Montie, J., Xiang, Z., Barnett, W. S., Belfield, C. R., & Nores, M. · 2005

grade Crctdeveloper-ledfailednumbers spot-checked
Sample
123 (58 treatment / 65 control); 72 males, 51 females
Population
Low-SES African-American children, Ypsilanti MI, 1962-67; entry Stanford-Binet 70-85.
Design
2.5h/day, 5d/wk, Oct-May, ~2 years, plus weekly 1.5h home visit; ~$17,759/child (2006 USD). Graded C not B: randomization was compromised in documented ways (siblings force-assigned to the same arm; treatment-to-control transfers reported inconsistently as 2, 3-6, or 5 across HighScope's own publications).
Key findings
The most policy-influential early-childhood study ever run, on n=123. End-of-treatment IQ gains were large and vanished by age 8. What persisted to 40 was attainment, employment, and reduced arrests — not measured ability. Effects are strongly female-driven on attainment and male-driven on crime, and the whole result rests on a tiny, single-site, 1960s sample with a no-preschool counterfactual that no longer exists.
Counterfactual
NOTHING — controls received no preschool and no home visits. A 1962-67 counterfactual, not the modern one where most controls attend some preschool.
Genetic confound
RCT (though compromised). Note the pattern: IQ/g gains vanish while attainment/behavior persist — the signature of this whole literature.
Replication notes
Anderson (2008) FDR reanalysis of the FULL Perry outcome family leaves exactly TWO effects at q=.05: early MALE IQ and FEMALE high-school graduation. Abecedarian's crime effect is NULL, so Perry's crime-driven ROI does not replicate.
Cost
~$17,759/child in 2006 USD. 67-88% of modeled public return is crime savings; ~93% attributable to males.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
IQ: share scoring 90+ at age 5% of group67% program vs 28% controlstandardizedend-of-treatmentbusiness-as-usualend-of-treatmentg
IQ fadeout (whole sample)IQ points T-C~+11 at program end (95 vs 84) -> +4 at age 6 -> ~0/ns by age 8 -> 80 vs 81 at 14standardizedgrade-6+business-as-usualover-2yrg
Achievement at 14 (California Achievement Test, 'basic achievement')% of group49% vs 15% — achievement PERSISTED even though IQ did notstandardizedgrade-6+business-as-usualover-2yrdomain-skill
High school graduation% of group65% vs 45% overall; FEMALES 84% vs 31-32%; MALES 54% vs 48% (ns)mixedadultbusiness-as-usualadulthoodattainment
Earnings at 40% earning >=$20K; median60% vs 40%; median $20,800 vs $15,300mixedadultbusiness-as-usualadulthoodattainment
Employment at 40% employed76% vs 62% overall; MALES 70% vs 50%mixedadultbusiness-as-usualadulthoodattainment
Arrests% with 5+ arrests36% vs 55%mixedadultbusiness-as-usualadulthoodbehaviour

Cited by