George — that ties in quite a bit with this Slate article from Ray Fisman. (Which, oddly enough, I also discussed in my print-issue review of Charles Murray’s Real Education.)
As opposed to pretty much every other measure of a school’s quality, teachers really do seem to make a difference on students’ test scores. The problem, as you point out, is that pushing bad teachers out of one school tends to mean they end up in another; Fisman writes of a New York principal who managed to rid his school of underperforming instructors, but given the union, had to pawn them off on other schools in the district. To make matters worse, it’s not really possible to only hire good teachers — the only measure with predictive power is whether a teacher has raised test scores in the past, and you can’t measure that without some school hiring the teacher. So, we end up with a lot of bad, union-protected teachers.
Fisman suggests three ways to get around this: (A) Make teachers do an apprenticeship before being hired, (B) devise some sort of training for bad teachers, or (C) find some way to fire bad teachers. I’m not convinced (B) is possible and I’m pretty sure (C) isn’t, but I like (A).