A school buys an anti-bullying program. The brochure cites a study showing strong results. What it usually doesn’t say: who ran that study.
This matters more than most people realize. Across anti-bullying research, a pattern shows up again and again. When a program’s own developers evaluate it, the effects are large. When independent researchers repeat the same test, the effects shrink, sometimes to nothing. Researchers call this the developer effect.
Olweus is the clearest case
The Olweus Bullying Prevention Program is the oldest and best-known anti-bullying program in the world. Dan Olweus developed it in Norway in the early 1980s, after three teenage suicides linked to bullying.
His own evaluations in Norway showed strong effects. But when researchers in Seattle ran an independent trial in 2007, they found no overall effect on victimization. The study wasn’t even included in a major program-rating database, because it didn’t meet the inclusion criteria.
A 2018 study in Pennsylvania found a significant effect again. Olweus was a co-author.
Two researchers, Eisner and Humphreys, gave this pattern a name in 2012: studies run by a program’s own developers produce systematically larger effect sizes than independent replications. Not fraud. More likely a mix of confirmation bias and better implementation by true believers.
It’s not just Olweus
The same shape appears with Taakspel, the Dutch adaptation of the Good Behavior Game. Dutch studies show positive results. A large English trial, run independently and funded by the NIHR, found no effect on the primary outcome in 2022.
One case could be coincidence. Two starts to look like a pattern.
What this means for a school board
It doesn’t mean these programs don’t work. KiVa, for instance, has been tested in an independent Dutch trial with a solid effect: lower odds of victimization after two years, replicated again in a large UK trial in 2024 and 2025.
It means the source of a study matters as much as its result. A number from the people who sell the program and a number from researchers with no stake in the outcome aren’t the same kind of evidence, even if they look identical on a page.
One detail worth knowing: the Olweus research group didn’t share its data for a 2022 independent meta-analysis that pooled results across programs. The KiVa group did. That choice tells you something too.
Faktis is currently completing a method-audit of six anti-bullying programs used in European schools. This article draws on that ongoing research. Full sourcing and methodology will follow in the complete report.
Sources
Eisner, M., & Humphreys, D. (2012). “Measuring the impact of programs against violence and bullying.” Journal of Experimental Criminology.
Bauer, N. S., et al. (2007). Journal of Adolescent Health, Seattle Olweus trial.
Olweus, D., Limber, S. P., & Breivik, K. (2018). Pennsylvania Olweus evaluation.
Humphrey, N., et al. (2022). NIHR-funded trial of the Good Behavior Game, England.
Huitsing, G., et al. (2020). Prevention Science, independent Dutch KiVa trial.