I find it troubling that discussions around merit pay (or performance incentives, or whatever the iteration at hand or preferred terminology may be), never quite seems to be completely honest. Even when people are trying to be honest, it seems that part of the discussion is based on half-truths and misconceptions.
Take this recent anti-merit pay op-ed in EdWeek by Kim Marshall, for example. He points out a number of faulty assumptions that many make when discussing merit pay, and then makes some of her own.
He's more or less correct when he says "The best teachers are already working incredibly long hours, and there’s no evidence that extra pay will make them work harder or smarter". One could argue that other fields provide some evidence, but to date there's virtually no evidence (certainly no experimental evidence in the U.S.) that merit pay will make teachers work harder -- or that if they did work harder that this would subsequently yield better results. It may be the case that many teachers are working pretty much as hard as they can and/or wouldn't be better teachers if they worked harder (they may pursue the wrong strategies or simply become more stressed).
But then he says "Teachers who are rewarded for their own students’ test-score gains are less likely to share ideas with their colleagues." This is demonstrably false. Every merit pay scheme I know of is designed to prevent teachers from competing with teachers at their own school for a share of a defined pool of money. The experiment that just concluded in Nashville compared the performance of teachers' students to the typical historical performance -- not to how other kids in the school performed.
His other points are mostly valid, though not necessarily precise. It's true that researchers say it takes three years of data to accurately estimate the effectiveness of a teacher (as measured by standardized tests). It's true that incentivizing higher test scores also incentivizes more test prep and even cheating -- but by that logic it would also incentivize harder work, which he earlier dismissed. He says half of all teachers teach untested subjects, but in some states it's closer to 70%.
I have yet to find a discussion of merit pay that's both based on facts rather than conjecture and approaches the topic in an unbiased way. People on both sides of the argument are making many dangerous assumptions, often based on incorrect information. The fact is that merit pay is utterly unproven in American schools and that while we can guess how it might affect teachers and schools, we simply can't know for sure until we try.
Right now, the idea is spreading rapidly, and I worry that the continuation or termination of the trend is going to depend more on half-informed arguments rather than sober analysis of research.