There is some pretty nice theory supporting the idea of merit pay. There is also some pretty weak evidence that it actually works.
Merit pay could work in two ways.
The first is that it will change who chooses to teach. If we can pay all of the best teachers well and the lousy teachers poorly, then over time hopefully we can get the lousy teachers to do something else. The idea is to change the makeup of the teaching profession, hopefully by washing out the poor teachers by not paying them very much.
I am not totally convinced that merit pay will do this. The measures of how much knowledge a teacher has added to their students are far from perfect. In fact, teachers rated highly in one year may be rated in the bottom the next year using these measures (see Aaronson et al (2007), McCaffrey et al (2009), and Stacy (20??) as soon as I finish it). Anyway, in a merit pay system, a teachers salary might swing a lot from year to year. People tend not to like that, so it may be that the best teachers, who also have the best outside opportunities, decide to leave teaching altogether. This is all conjecture though.
How merit pay affects the teaching profession in the long term is an empirical question. Unfortunately, I don't think anyone has answered it convincingly. We don't know, basically.
The other way merit pay can work is to motivate teachers already in the profession. This notion has a big economics literature. It's called moral hazard. Society hires teachers to teach their children. The teachers don't (so the argument goes) have as much a stake in the outcomes of the students as their parents. Without any sort of tie of salary to performance, teachers might be lazy and not work hard. Merit pay gives teachers an incentive to work harder, and hopefully this will lead to more effort and better student outcomes.
We do have some good evidence on this idea. In a study ran from 2006-2009, teachers in Nashville, TN were paid bonuses of up to $15,000 for improving student test scores. This was a randomized experiment, which despite what some people say, is the gold standard of empirical research. You can read through the report here. It's not perfect, but it's probably as good as we can do.
Basically they found zero effect of the merit pay bonus system on student achievement. This was a short run study, so it doesn't offer us much insight into the first argument. However, this is a big blow to the second argument. The authors of the report were pretty careful, and it's hard to come up with a good reason for why the chance for a $15,000 bonus didn't motivate people if they could be.
There is a great sentence in the conclusion of the report:
In sum, the introduction of performance incentives in MNPS middle schools did not set off sigificant negative reactions of the kind that have attended the introduction of merit pay elsewhere. But neither did it yield consistent and lasting gains in test scores. It simply did not do much of anything.As far as I could tell, the authors best explanation for why their was little effect was that teachers were mostly working pretty hard to begin with. They surveyed the teachers, and 80% reported that they were teaching as effectively as they could even before the program. If that's the case, then we should only expect a marginal effect.
I am not convinced that merit pay works. Maybe it can with the right design, but I don't know. I don't think there is much data supporting the idea that we can improve student achievement with teacher merit pay. There is strong evidence against the second argument, and there is little or no evidence either way for the first argument.
Also, if anyone has any papers that contradict these thoughts, then let me know.