Our thinking on this has evolved over the years, but our work now suggests that fadeout is fairly ubiquitous across education interventions. In our recent meta-analysis of approximately 85 diverse randomized controlled trials, we found that only one intervention characteristic was a strong and consistent predictor of the follow-up impact size in the two years after programs ended: the size of the effect at the end of the intervention. Surprisingly, information about intervention features, skills targeted, and study characteristics provided little additional predictive power. In other words, knowing how big the effect of the intervention was at the end of the program was a good predictor of the follow-up effect a few years after interventions ended, but other characteristics of interventions and programs told us almost nothing about how much effects would fade or persist. While the persistence of effects varied considerably (and impacts did persist to some extent, on average), theoretically salient factors could not explain this variance. Instead, we found that effects faded fairly consistently across cognitive interventions, social-emotional interventions, early childhood interventions, and adolescent interventions.
To better understand the very long-term effects of education programs, our team has conducted a meta-analysis examining randomized controlled trials that have measured adult effects. We identified 29 programs that evaluated end-of-treatment and adult outcomes and found that, despite medium-term fadeout on child skills, these programs did produce positive long-term impacts on adult functioning, though effects are fairly small on average (~0.05 SD).