There's an article in the National Bureau of Economic Research (NBER) Digest about the relationship between merit pay for teachers and student achievement:
Figlio and Kenny find that teacher salary incentives are associated with higher levels of student performance. They cannot be certain whether the test score improvement is driven by teacher incentives or whether the incentives are proxy variables for unobserved school quality. In general, they find, teacher salary incentives are associated with a 1.3 to 2.1 point rise in test scores, about the same increase associated with increasing maternal education by three years. The correlation exists in schools with predominantly low- and middle-income students.
Figlio and Kenny note that one case in which merit pay systems don't work is when most of the teachers get merit pay. If everybody's a winner, there's no incentive to improve.
In Missouri, the Ladue School District has used merit pay for over 50 years. However, Ladue does suffer from the merit-pay-for-everyone problem. A report by Ladue's Compensation Systems Task Force notes that 70 percent of the district's staff received 12 or 13 out of 13 possible merit points for the 2004-2005 school year. Ladue should reconsider the way it awards points if it wants to encourage excellence in teaching.