STARS

Project Information

Background

Like many countries, Rwanda’s education system faces significant hurdles, including under-resourced schools, uneven teacher quality, and limited support for educators. Despite strides made by the Ministry of Education, the Learning Assessment in Rwandan Schools (LARS) showed that fewer than 40% of Primary 3 students meet expected English proficiency, and less than 60% do so in Math, with even greater challenges in economically disadvantaged districts.

Effective teacher management is essential to improving student outcomes, yet Rwanda grapples with high teacher attrition rates—nearly 20% of primary school teachers leave annually, and only 37% of teacher training graduates enter the workforce. To build a thriving education system, Rwanda must adopt sustainable strategies that not only invest in teacher salaries but also empower educators through development, motivation, and retention initiatives, ensuring long-term impact.

The Government of Rwanda is tackling education challenges through bold reforms and investments in teacher support. In 2022, teacher salaries were raised by 40% to 80% of net income to improve retention and motivation, laying the groundwork for enhanced educational quality. Complementing this, the government is building a robust data infrastructure to link teacher performance with student outcomes. Through the Comprehensive Assessment, launched in 2019-20, teachers now report student performance via the Comprehensive Assessment Management Information System (CAMIS). This integrated system enables annual tracking of teacher effectiveness and provides a data-driven foundation for policy decisions that strengthen Rwanda’s education system beyond the STARS initiative.

From 2015 to 2018, a pay-for-performance randomized controlled trial (Leaver et al), in Rwanda explored innovative ways to incentivize teacher performance and improve recruitment, motivation, and retention. Upper-primary teachers in the top 20% of their district were awarded bonuses of RWF 100,000—approximately 15% of their annual salary—based on a balanced set of indicators, including teacher presence, preparation, teaching quality, and student learning progress. This holistic approach ensured fair incentives across schools, discouraging “teaching to the test” and promoting genuine learning gains.

Program Design

Imihigo, a traditional Rwandan performance management system, involves teachers and civil servants setting annual targets and receiving bonuses based on their achievements. These bonuses, ranging from 0 to 5% of a teacher’s annual salary, are intended to reward performance throughout the school year. However, in practice, the subjective nature of performance evaluation criteria has often resulted in bonuses being distributed as fixed payouts. This has reduced the system’s effectiveness in motivating teachers to excel based on measurable and objective performance outcomes.

STARS builds on the trial insights from Leaver et al, and enhances the imihigo framework by introducing performance-based bonuses aligned with clear, measurable goals. The STARS program, on its path to being scaled to all primary schools in Rwanda, integrates rigorous A/B testing of various models to help the Government decide what is the best combination of indicators, and ways of measuring them, to include in the teacher performance contracts. 

Central to STARS is the measurement of teacher effectiveness across five key performance indicators, referred to as the “5Ps”: Presence, Preparation, Pedagogy, Participation in the Comprehensive Assessment Management Information System (CAMIS), and Pupil Learning Gains. These indicators provide a holistic view of teacher performance, from their commitment to lesson planning and classroom instruction to their contributions to student learning growth. Importantly, the pupil learning gains are measured relative to the pupil starting grade in last year’s Comprehensive Assessment, ensuring teachers are not penalized for being assigned lower performing students in their class. 

The program’s design directly integrates with the Government’s education infrastructure. Sector Education Inspectors (SEIs), who already play a role in school oversight, are instrumental in collecting data on these performance metrics, making STARS one of the first programs to embed teacher evaluation so deeply within a country’s civil service framework. By working with SEIs for measuring teacher attendance, lesson preparation and classroom practices, STARS gains an accurate, on-the-ground view of teacher practices. In some of the STARS treatment variations, headteachers also provide feedback on teacher performance, adding an additional layer of insight and accountability at the school level. Participation in CAMIS involves teachers systematically recording student performance data in a centralized database, providing a measurable indicator of student progress. These data play a critical role in evaluating pupil learning gains, establishing a clear link between teacher practices and student outcomes.

Program Evaluation

​​The STARS program employs a rigorous A/B testing approach to identify and refine the most effective models for improving teacher motivation, feasibility of implementation, and student learning outcomes. By testing multiple variations of teacher performance contracts, the program continuously assesses impact, phasing out models with limited effectiveness or implementation barriers while incorporating successful elements into a scalable framework for all primary schools in Rwanda.

Over the first two years, the program has implemented five distinct variants of teacher imihigo contracts in 10 of Rwanda’s 30 districts. Using a “hub and spoke” evaluation design, eligible schools are randomly assigned to one of these variants or a control group maintaining standard teacher contracts. The variants include:

• Model A (Base Design): Performance measurement across five key areas (5Ps) with payoffs up to 5% of salary based on grades.
• Model B.1: Adds headteacher evaluations of teacher inputs alongside SEI assessments.
• Model B.2: Incorporates interim feedback to teachers on their relative performance.
• Model B.4: Includes audits of student exams to ensure accountability.
• Model B.5: Introduces headteacher imihigo contracts based on aggregated teacher outcomes.

Currently operational in 345 schools, the program reaches over 6,600 teachers and 300,000 students. As results from ongoing iterations inform the optimal design, STARS is set to expand, ultimately covering all primary schools, teachers, and students nationwide.

In the first year, STARS contracts showed clear improvements in teacher input metrics, making a positive difference to teacher presence, preparation, pedagogy, and participation in CAMIS. Teacher presence increased only slightly, from a very high rate of 92% in control schools. However, treatment school teachers were 17-19 percentage points more likely to be prepared, and their pedagogy scores rose by 4.5-7%, compared to control school teachers. Participation in CAMIS also surged. 10 out of the 11 districts with the highest CAMIS involvement nationally were the STARS districts. In these districts, STARS schools largely drove these gains, with teachers in treatment groups increasing their participation by 5-9 percentage points, depending on the treatment arm. Improvements in teacher inputs were observed across the board, and didn’t differ by treatment variations.

Notably, Year 1 impacts on student learning outcomes were modest and statistically insignificant. This was in line with the previous RCT findings that showed larger learning gains emerging only in the second year of implementation, and encouraged the Government Task Force to renew all of the treatment variations for one more year without changes.

In Year 2, STARS contracts continued to strengthen teacher preparation, classroom practices, and CAMIS participation, in large and significant numbers. Importantly, Year 2 results also brought emerging evidence that some of STARS contracts design variations (models B1, B4, and B5), positively influenced student learning outcomes, while others did not. This led the Task Force and the Ministry to decide to drop the lower-performing models for the third year of implementation, and to re-allocate their schools to new variations of the contract, which combine features of the retained models. The random re-allocation of schools to these new arms, dropping low-performing ones and testing new combinations of what works, allows the Government to learn by doing. The results from further evaluation will inform the design of the final imihigo contract scaled up to the rest of the country the following year.