It's an interesting question, and that's probably why it ends up in so many econometrics and stats courses. You have the classic causality problem, along with e

It's an interesting question, and that's probably why it ends up in so many econometrics and stats courses. You have the classic causality problem, along with ethical concerns of generating experimental data for such a study. Controlling for as many of the relevant variables as possible, there does seem to be a strong negative correlation between class size and test scores. Beyond that, who knows, but it makes sense that small classes would succeed more often (at least up to a certain point of course)