Managing and motivating teachers is essential for providing basic reading and mathematics skills to children in early grades. This column reports on the impact of the KiuFunza teacher incentive program, which offers a small performance-linked bonus to early grade teachers to increase their teaching effort. Experimental evidence shows that several incentive designs resulted in a substantial increase in learning among students of teachers in Tanzania with these incentives. A relatively cheap and easy to communicate version of the program was most effective.
There is a growing consensus that while traditional education investments in school inputs and teacher salaries are necessary to keep schools running, they are not going to improve learning levels in low-income settings. There is also ample research evidence that teachers can make a substantial difference in what students learn and will earn, while their effectiveness in the classroom co-determines the impact of other education spending.
But there are concerns about two major determinants of effective teaching: teacher skills and teacher motivation. Of these two levers, the first is hard to move: predicting who will become good teachers and recruiting them are not easy, while recent work uncovers ‘a sharp gap between the characteristics of teacher professional development programs that evidence suggests are effective and the global realities of most teacher professional development program’.
Thus, an important question is whether the motivation and learning effectiveness of currently employed teacher—the single most expensive budget line in education systems—can be improved, without additional training.
Our work focuses on teacher performance pay, an instrument of education governance that links teacher motivation and learning: it offers a financial reward based on teacher performance, typically measured by their students’ learning.
Teacher performance pay can complement existing input investments (such as school grants, teacher training and books) through a number of channels: improved goal-setting, improved monitoring of goal distance, and providing rewards for goal achievement. Teacher performance pay systems have been shown to improve student learning, particularly in low-income settings, but important design questions remain.
One trade-off is between simple incentives based on skill levels versus more complex designs that reward value-added (that is, learning growth). We investigated this trade-off in a teacher performance pay program named KiuFunza (shorthand for Kiu ya Kujifunza or Thirst to Learn) in Tanzania. KiuFunza has been developed and is implemented by Twaweza East-Africa, a civil-society organization, in collaboration with government and international research partners.
The main design features of KiuFunza teacher performance pay are that:
- The program offers cash rewards in public primary schools (which enroll about 97% of all primary students in Tanzania), conditional on independently measured learning outcomes only.
- The program targets Grade I–III teachers only, on the basis that these teachers are responsible for foundational skills that are essential for students’ school careers.
- The subjects with incentives are Kiswahili and mathematics (English was dropped from the Grade I–II curriculum in 2015).
- Head teachers receive a bonus equal to 20% of all rewards earned by teachers. Each program was implemented as a randomized evaluation in a nationally representative sample.
Our research examines the effectiveness of two KiuFunza incentive programs implemented in 2015-16. The main difference between the two designs tested is the mapping between test results and teacher earnings:
- Percentile Pay: a tournament system, in which students are placed in starting ability groups across all participating schools. Teachers’ earnings increase with the ranking of ‘their students’ within these groups.
- Skill Pay: a novel proficiency design, where a teacher earns a bonus for each curriculum skill test item (for example, reading words and sentences, and doing addition) that a student passes.
In addition, these designs had a fixed bonus budget, with a per teacher cash reward of about 3.5% of the average annual teacher salary in 2016.
The impact evaluations show that both programs improved learning by the equivalent of about one third of a year of schooling. We further find that both programs increased the number of students sitting the exam, which we interpret as a sign that students have become more valuable in the eyes of the teachers. We find no effects, positive or negative, on learning in grades or subjects that are not given incentives.
Skill Pay had a slightly larger learning impact than Percentile Pay, particularly for Kiswahili. Skill Pay reduced the share of students repeating a grade by 24% (a decision by the school committee that is unrelated to bonus pay), but Percentile Pay did not.
We have some evidence that teachers were more optimistic about their earnings under Skill Pay, and more goal-focused. Importantly, Skill Pay is easier to implement, manage and communicate, and requires only one test (rather than two as in Percentile Pay).
We thus find that Skill Pay has higher cost-effectiveness than Percentile Pay and good cost-effectiveness in comparison with other interventions that aim to improve learning.
Overall, our findings show that offering teachers a simple and modestly sized performance reward linked to learning can improve their performance at current levels of professional development.
Based on the 2015–16 impact results, the government of Tanzania—through the Ministry of Education, Science and Technology, the President’s Office, the Regional Administration and Local Government—asked Twaweza to formulate and test a performance pay program that can work at larger scale. This new program reduced testing costs, has government personnel working with Twaweza teams, and started in 2019. The results of evaluations of this program are expected in 2021.
