I feel like the real problem is in how we measure teachers. It shouldn't be about how well the students do, it should be about how well the students do relative

I feel like the real problem is in how we measure teachers. It shouldn't be about how well the students do, it should be about how well the students do relative to expectations. Imagine a school in a well-to-do area where we expect all the students to do well and get A+'s. If an entire class of students gets A- on a standard test, that means the teacher actually negatively impacted the students. But because they all got A-, we reward the teacher for having done such a great job. I propose this: for each student, far ahead of time, guess how well they're going to do on a standardized test based on previous tests, family situation, zip code they live in, etc. Predict based on all the things other than the teacher. Now measure a teacher on how well they did compared to expectations. To do even better, don't rate teachers based on how well the students do on the test the same year they teach them, but rate them based on the following 3 years. Now you're asking what the long term impact the teacher has on students, on average. That's a useful measure to pay bonuses on. The problem is that politically, it's hard to explain to the general public (or to politicians).