Teacher Merit Pay Might Raise Student Achievement After All

A few weeks ago, I wrote a post on whether teacher merit pay is effective at increasing student achievement.  I felt that the evidence was pretty weak for the whole idea.

Lo and behold, a new paper came out recently on just this subject.  The authors are Roland Fryer, Steven Levitt (of Freakonomics fame), John List, and Sally Sadoff.  Here is a link, although it's gated. 

They conduct a randomized controlled trial in which teachers are randomly assigned to a bonus program, which gives money for strong gains in student achievement.  The program is fairly similar to the Nashville study that I discussed before. 

The really interesting thing is the way the bonuses are given out.  Some of the teachers are paid $4,000 at the beginning of the school year, before any teaching has taken place.  Another group are given the bonus at the end of the year, and the amount depends on how well their students perform.  For the teachers paid at the beginning of the year, if their students don't perform well, then they are forced to give potentially all of the money back.

The notion that people might be more motivated to not lose something than gain something has a long literature dating back to some work by Tversky and Kahneman (another person with some interesting books). 

The teachers are randomly assigned to the groups.  It turns out that the effect of the program on student achievement is weak and mostly statistically insignificant for teachers given money at the end of the year, as it was in the Nashville study.  However, for the group of teachers that faced the prospect of giving back part of the $4,000, the effect was pretty big!  Students with teachers that faced the loss had test scores that were .201 to .398 standard deviations higher on average compared to a control group of students.  A .398 standard deviation improvement is similar to going from a 500 on the SAT to a 544, and that is for every student in the class.

I would like some more information about the randomization, like whether principals had the opportunity to assign the program teachers to better or worse classes.  However, I think this paper is really interesting, and it's hard for me to come up with an alternate explanation for why the upfront bonus that can be taken away is so much more effective. 

I was skeptic of teacher merit pay, but when new evidence comes along we are supposed to change our minds.  Maybe teacher merit pay can work if it's done the right way.