The National Bureau of Economic Resarch just released a new study on peer effects in the classroom:

The marginal effect of a one percent increase in the quality of peers on student achievement is equivalent to between 8−15% of a one percent increase in one’s own earlier achievement…We find that peer effects operate in a heterogeneous manner. High ability students benefit more from having higher achieving schoolmates and from having less variation in peer quality than students of lower ability.

If I understand the paper correctly, there were three separate findings. 1) Kids achieve more when their peers are high-achieving, 2) high achievers achieve even more when they aren’t mixed in with low achievers, and 3) this peer effect is most pronounced for high-achievers.

Now comes the difficult policy decision. This study makes it clear that you can increase achievement and decrease equality (segregate kids according to achievement level), or decrease achievement and increase equality (muddle everyone together), but you can’t do both. In other words, we have to pick our poison. Do we want a small group of isolated A students, or a large population of integrated C students? The permutations of peer effects means that it’s very difficult to have both…