Michael Lovenheim, Cornell University
Incentive Strength and Teacher Productivity: Evidence from a Group-Based Teacher Incentive Pay System
Scott A. Imberman∗
University of Houston and NBER
Michael F. Lovenheim
***PRELIMINARY AND INCOMPLETE***
***DO NOT CITE OR DISSEMINATE WITHOUT AUTHORS’ PERMISSION***
We use data from a group-based teacher incentive program implemented in high schools in an a large urban school district to estimate the impact of incentive strength on student outcomes. The program provides cash bonuses to teachers who are in schools where students perform well on achievement exams as measured by a value-added model. Unlike other incentive programs, these awards are based on the performances of students within a grade, school and subject, providing much more variation in group size than awards based on the performances of all students in a school. We use the share of students in a grade-subject that a teacher has in his or her classes as a proxy for incentive strength since, as the student share increases, a teacher’s impact on the probability of award receipt rises. Hence, if teachers respond to the award incentive we should expect to see their students’ achievement improve as the teacher becomes responsible for more students. Indeed, this is what we find. Overall, difference-in-differences estimates show that a 10 percentage point increase in student share increases math, English and social studies scores by 0.01 to 0.02 standard deviations. However, we also find evidence of heterogeneity by share. Local linear regressions show that a similar increase in share for teachers who are responsible for only a handful of students improves achievement by 0.05 to 0.09 in all subjects with the effects fading out to zero as share increases. These findings suggest that small groups that provide teachers with larger shares of the responsibility for awards provide productivity gains over large groups with an optimal group size of around 3 to 5 teachers (20% to 33% share for each teacher). Further, our results suggest that the lack of effects found in US teacher incentive pay experiments are probably due to specific aspects of program design rather than failure of teachers to respond to incentives more generally.