Loss aversion: The new catch-phrase in merit pay debates

See all posts
A dynamite new paper from Harvard economist Roland Fryer and colleagues, including Freakonomics star Steven Levitt, is a one-two punch on the issue of performance pay. Pushing past evidence that such pay doesn't seem to motivate American teachers to do a better job, the researchers add a new twist to the previous experiments that have led to such blah results. This time they tested the application of loss aversion--a well-documented force in lab experiments, but never before applied to the thorny issue of teacher pay.

Here's the setup. The experiment offered 150 K-8 teachers in an extremely poor district outside Chicago a chance to earn four different types of bonuses. Teachers in two 'gain' groups were eligible to receive up to $8,000 at the end of the school year, depending on either individual or team successes. The maximum reward represented a 16 percent raise on average. Teachers in two 'loss' groups (also individual and team) each received $4,000 upfront, with the promise that they would earn up to another $4,000 for above-average student growth. If they had below-average student performance, they would have to pay the district back from the original $4,000 they'd gotten.

Consistent with the findings of a couple of other studies, the teachers in the two 'gain' groups didn't produce any higher level of performance. But the teachers in the 'loss' groups produced not just significant gains but actually monumental gains--equating to a full standard deviation improvement in teacher quality.

Several interesting sidebars:

  1. The experiment was attractive to the teachers:  94 percent of the eligible teachers in the district elected to participate.
  2. The results appeared real, not the product of cheating. The positive effects were sustained on assessments that were not tied to the experiment's financial rewards.
  3. The team structure appeared to motivate an effective kind of collaboration. The positive effects were better sustained on the independent assessments for the teachers in the 'loss' team group than for the teachers in the 'loss' individual group.

There's a whole lot to unpack here, but apart from these really, really fascinating results, the paper doesn't provide many answers to our many questions. The gains were clearly so big that other work is critical. We're keeping our ears to the ground.

Kate Walsh and Laura Johnson