5 Big Ideas That Don't Work In Education

[Daniel Smith]

Has anyone heard of this guy or read his book?

http://www.npr.org/sections/ed/2015/08/13/430050765/five-big-ideas-that-don-t-work-in-education

The article has a link to a chart of interventions by effect size, which is pretty interesting if accurate.

[Tom Adams]

If I understand correctly an effect size is 1 if effect size = standard deviation (sd).  effect size = size of effect/sd.  I guess sd is estimated from the control group.

But note that class size does not seem to be in the effect size table, or at least I cannot find it.  The evidence given for the low effect size of class size in the link is observational and subject to confounders, I think the most one can say from the data presented is that the effect size is lower than the confounders for some pairwise country comparisons. Also that low class size is not needed in some countries to outperform relatively well.