We find that student achievement improves when a teacher becomes responsible for more students post program implementation: mean effects are between 0.01 and 0.02 standard deviations for a 10 percentage point increase in share for math, English and social studies, although mean science estimates are small and are not statistically significant.
While these effects seem small, they are bigger than are found in many previous studies. So, if the incentive dissipates for larger groups, why use group incentive rather than measuring performance at the individual level? I can think of a few reasons (and maybe you can think of more):
- Individual incentive would lead teachers to try to "cherry-pick" students. These could just be the better students if raw test average is the performance measure or those thought to be able to improve the most if change in score is the metric.
- To better sort students into classes. Students differ in abilities as well as other characteristics (e.g. unruly versus well-mannered). Some teachers are better with one type of student than another (e.g., former drill sergeant). The group of teachers all benefit from better matching students to teachers.
- Spillovers within a group. Teachers have heterogeneous abilities and no one likes to be 'corrected' by a colleague. But now there is a stronger incentive for the better teachers to share their methods with those who can improve.
- Demonstration effects across teacher groups. Other groups of teachers can observe the successful groups and learn.