At its core, merit pay is a bet on human self-interest: will teachers perform better if there is more money to be made? In 2010, four economists took that idea and stood it on its head. Would teachers perform better if they stand to lose money? The technical term is loss aversion.

Economists Roland Fryer, Steven Levitt, John List, and Sally Sadoff selected nine K-8 Chicago Heights schools, most of which were predominantly low-income schools, to conduct their experiment.[i] Teachers at these schools earn an average base pay of $50,000, and were offered the chance to make up to $8,000 in bonus pay. However, the bonus structure wouldn’t look the same for everyone.

Some teachers had a traditional pay schedule – they earned their bonuses based on their students’ test scores. But some teachers got a bonus at the beginning of the year, with this important caveat: if their students earned below-average test scores, they would have to pay some of that bonus back.

This plan worked – math test scores improved overall – but it worked especially well for the teachers who could potentially lose money. “The impacts we observe are large – roughly the same order of magnitude as increasing average teacher quality by more than one standard deviation.”[ii]

Teaching incentives

The Experiment 

The “gain” group would earn the bonus on a traditional schedule. At the end of the year, they would earn $80 per test score percentile point. If the average percentile in Teacher A’s class was 50 percent, Teacher A would earn a bonus of $4,000. Teacher B would receive a $5,600 bonus for an average percentile of 70 percent. 

The “loss” group had an unorthodox schedule. These teachers would receive $4,000 at the beginning of the year, but they would have to pay back whatever part of the bonus they didn’t earn. So if Teacher C’s students finished the year with an average percentile of 45 percent (which would earn a bonus of $3,600), Teacher C would have to return $400. But if her students had an average percentile over 50 percent, Teacher still could still earn up to $8,000.

On average, the “gain” group increased math scores by 1.884 percentage points. The “loss” group increased them by 6.84.

The economists also split these groups into teachers who were rewarded based on team-level scores, and teachers who were rewarded exclusively on their own improvement. The difference between team and individual scores was insignificant – teachers who had to achieve team-level improvement did just as well as teachers who only had to worry about their own classrooms.

There would be numerous obstacles to expand this concept to an entire school district or state, but the experiment finds that negative incentives have their own merit.


[i] Roland G. Fryer, Jr; Steven D. Levitt; John List; and Sally Sadoff. (2012). Enhancing the Efficacy of Teacher Incentives Through Loss Aversion: A Field Experiment. National Bureau of Economic Research. Available at http://rady.ucsd.edu/docs/faculty/Fryer_et_al_Teacher_Incentives_NBER_WP18237_2012.pdf
[ii] See note i.