Improving student learning outcomes through performance-based compensation for teachers
How can pay-for-performance contracts incentivize teachers to improve student learning outcomes while promoting quality teaching practices? This is the challenge addressed by the "Supporting Teacher Achievement in Rwandan Schools" (STARS) program, designed by the Rwandan Ministry of Education, alongside its partners.
In Rwanda, teacher compensation is tied to annual performance targets based on a system known as Imihigo. An initial impact evaluation conducted by Innovation for Poverty Action and the Rwanda Education Board (Leaver et al., 2021) has demonstrated that contracts linking teacher performance to more objective criteria may significantly improve learning outcomes. These contracts are structured around a common framework called the "5Ps": Presence of teachers, Preparation of lessons, Pedagogy, Participation in CAMIS (the national database for reporting student assessment grades) and Pupil learning gains. The STARS program tested and evaluated different versions of this contract including different features, in order to identify the most effective model for scale-up. The project also aims to increase CAMIS reporting, thus improving the quality of data used to assess learning outcomes and manage performance-based contracts.
Key impact results
Results from the third year of the pilot project show that the pay-for-performance contracts improved teachers' practices in several areas, particularly in terms of their lesson preparation and pedagogical approach. CAMIS reporting has gradually increased, although these results vary depending on the contract arrangements in place. Under certain models, improvements in teaching practices are associated with better learning outcomes for students. The most promising results on both teaching practices and student learning outcomes were achieved using a contract model based on the "5Ps" version that includes a performance bonus for headteachers, linked to the results of teachers in their school (version B5).
+13
percentage points average increase in available lesson plans during unannounced visits to schools using STARS contracts.
$0,15
per student is the cost of implementing the scheme
+0,26
standard-deviations improvement in teaching practices with version B5.
+0,11
standard-deviations improvement in learning outcomes with version B5.
These results are still preliminary and need to be validated through further research. For details of the contract variations tested and evaluated, see the table in the document in attachment.
Impact study methodology
This research applies an adaptive experimental approach: since 2023, various contract versions have been trialed each year, and their impact on teachers' practices and students' learning outcomes assessed. This information was then used to gradually phase out the least effective models while continuing with the most promising measures, either alone or in combination. The aim is to determine which contract model is both effective and suitable for wider scale-up.
A serie of tests (randomized A/B tests) was conducted to compare the different contract versions – from B1 to B7 – each year for three years, across 245 schools, against the "5P" reference model in 50 schools and a control group of 94 schools that did not use the improved pay-for-performance contracts. In the third year, the focus was narrowed to only the most variations: in particular the ones including the base model (5Ps), the evaluation of teachers by headteachers, re-testing of students via external audits to improve the accuracy of students assessments, and incentives for headteachers to improve their schools' performance.
The impact evaluation addressed three research questions
- 01Can pay-for-performance contracts improve teaching practices?
The results have shown that pay-for-performance contracts improved teachers' practices in several areas, particularly in terms of lesson preparation and pedagogical approach. At the end of the project's third year, the different contract models led to an increase in available lesson plans, with an estimated improvement of around 11 to 15 percentage points depending on the specific contract arrangements. Independent observations also found that teachers were better prepared and more organized. However, no significant increase in the presence of teachers was observed, as this rate was already high from the start.
Contract model B5 achieved the most significant estimated impact on teaching practices (+0.26 standard deviations). This version combines the basic model with a performance contract for headteachers, which partially ties their evaluation scores to those of teachers at their school. This finding suggests that a two-pronged approach offering incentives for both teachers and management may help to improve teaching practices.
STARS contracts also appear to encourage reporting in the national CAMIS system, with an estimated positive impact of around 2 to 3 percentage points depending on which contract model was used in the third year. This increased participation has helped to provide more comprehensive data on student learning outcomes.
- 02Which contract models are most effective in improving students' learning outcomes?
Improving students' learning outcomes is a central component of all the pay-for-performance contracts evaluated, with learning gains accounting for 50% of teachers' overall score. Student progress is measured relative to their initial level so that teachers are assessed on how much their pupils improve, regardless of their starting point.
The results have demonstrated that stronger incentives for headteachers can also have a significant impact. According to independent evaluations, version B5 achieved the highest gains out of all the models tested, with an increase of around +0.11 standard deviations.
- 03Which contract model is most suitable for broader scale-up?
The key finding from the evaluation is that contract model B5 achieved the most promising results both in terms of teachers' practices and students' learning outcomes. Headteachers could therefore play a crucial role in making basic performance contracts more effective.
B5 adds a single component to the basic model: a financial incentive for headteachers tied to the performance of teachers at their school. This version requires less additional resources than more complex models involving several mechanisms, such as teacher evaluation by headteachers and student re-testing.
Choosing the most appropriate model for scale-up depends not only on its impact, but also on its cost and feasibility. Seeking to strike this balance, Rwandan authorities have selected the B5 model for national roll-out.
Perspectives and scale-up
Initially piloted in primary schools, the B5 contract model will now be rolled-out to 3,492 primary and secondary schools across 30 districts, representing 87,945 teachers and 3.23 million students. Research will continue alongside this national roll-out to substantiate the findings from four years of trials and to assess whether the impact of these contracts can be sustained on a larger scale.
Find out more
Research
Kubie, O. ; Leaver, C. ; Mfura, L. ; Zeitlin, A.
Case study, 2025.
Leaver, C. ; Ozier, O. ; Serneels, P. ; Zeitlin, A.
American Economic Review, 2021.
Muralidharan, Karthik ; Sundararaman, Venkatesh.
Journal of Political Economy, 2011.