Persistence and Fadeout in the Impacts of Child and Adolescent Interventions
Bailey, D., Duncan, G. J., Odgers, C. L., & Yu, W. · 2017
grade Cmeta-analysisindependentfailednumbers spot-checked
Sample
67 early-childhood interventions, 1960-2007 (only ~1/3 followed past end of treatment)
Population
Child and adolescent intervention studies across domains.
Design
Quantitative synthesis plus the field's dominant taxonomy of persistence mechanisms: skill-building, foot-in-the-door, and sustaining environments.
Key findings
Establishes the base rate every early-childhood claim must be judged against: intervention effects decay roughly geometrically, halving every 12-18 months. Note the internal tension — Building Blocks fadeout was control-group catch-up, while Protzko shows IQ fadeout is treatment-group decline. Institutional outcomes (special-ed placement, retention) persist better than test scores.
Genetic confound
Aggregates mostly experimental studies; the taxonomy is theoretical rather than tested.
Replication notes
Direct tests of the 'sustaining environments' hypothesis are sparse and mostly NULL (e.g. Abecedarian's supplementary school-age arm added nothing). Treat the taxonomy as a framework, not an established set of mechanisms.
DOI / URL
10.1080/19345747.2016.1232459
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Cognitive/test-score effects over time | ES | ~0.23 SD at end of treatment -> ~0.10 SD at 12 months, halving again over the next 1-2 years (roughly geometric, ~50% per 12-18 months) | mixed | grade-3 | business-as-usual | over-2yr | g |
| Abecedarian institutional outcomes (foot-in-the-door) | ES | special education -0.45; grade retention -0.54 — more durable than IQ/achievement | mixed | grade-6+ | business-as-usual | over-2yr | attainment |
| Control-group catch-up (Building Blocks math) | SD | controls gained ~1 SD/yr after the intervention, closing the gap | standardized | grade-3 | none | over-2yr | domain-skill |
Cited by
- Fadeout — why early gains disappear, and what actually persistsstrong supportconf: highgc: low