I am not in favor of giving teachers pay raises strictly based on years of experience and training/education, which I believe is the current standard (for public schools). Ideally, salary would be proportional to quality for teachers and principals. However, evaluating teachers and schools can be challenging. For example, there are several major problems/challenges with using student test scores as the marker of success.
Problems with test score based evaluations:
1. teachers teach to the test; the focus would likely be on skills and memorization; this could lead to a more homogeneous and less creative group of students than is ideal for society
2. selection bias: substantial variation in student quality across schools (by student quality here I mean independent from the effect of the school)
3. incentive to cheat: both teachers and administrators have incentive to cheat. if the metric is a change score, then there is incentive to do poorly on the pre-test.
4. at the state level there is incentive to make the test easier to show ‘improvement’ (e.g., link)
5. kids are self-centered; they have little personal incentive to try hard on these tests (when I was in school, I recall other students admitting that they were just going to randomly fill in dots, since they weren’t going to be graded on it)
I’m going to ignore the logistics of this (how to pay for it, how to implement it, etc) for now. Think of it as something more like a thought experiment.
Suppose we have a lot of fairly small schools, so that parents could choose between about 3 local schools (without having to travel long distances). Parents would have a choice of which school to send their kids to. A lottery would be used for schools that got more applicants than they could admit.
Similarly, within a school parents could choose between approximately 3 teachers. A lottery would be used if a teach got too many applicants. (note that this is a very different model than we see in public schools now, where parents are discouraged from asking for a specific teacher)
Ideally, these schools would all serve a single community, and therefore eliminates problem #2 above.
Schools would be judged primarily based on how many people wanted to go to that school. That is, schools would be judged based on demand for that school. I think it’s likely that demand would be correlated with quality (a school with a good local reputation would most desirable). It would be difficult to game this system (eliminating concerns #2 and 3). If a school was getting very few applicants, that would be a reason to consider hiring a new principal.
Teachers would be judged primarily based on demand as well. A teacher with a good reputation would get the most applicants. This would likely be correlated with quality.
In addition, the schools would likely reflect local preferences. If a community preferred skills and memorization to creativity, they’d send their kids to teachers and schools with those values. I could easily see different types of schools emerging in a single community, reflecting diverse preferences of parents.
One possible concern is that teachers and schools that were more generous with their grades (e.g., pass everybody) would be more popular. I’m not sure if that’s a problem, however. It would reflect the preferences of the community. In addition, the diploma would eventually be devalued (if these students were not having success post-high school).
Another possible problem is the small sample size. Suppose all of the local schools were excellent. Even if one had much lower demand than the others, that might not be reason to make major changes. So, having both a relative metric (like demand) and an absolute metric might be necessary. The same issue exists for teachers (i.e., a high demand teacher still might be bad, if all teachers in that grade are bad; a low demand teacher still might be good, if all teachers in that grade are good).