re: class size research: grade improvements driven entirely by grade manipulation

So is having teachers with large classes being evaluated mostly by test scores potentially optimal policy?

(assuming that policy for expensive private schools is excluded here)