Promoting Early Grade Reading & Numeracy in Tanzania: KiuFunza III

Our incentive scheme pays teachers for the skills their students are able to master. The design has three important features:
1) The bonus is calculated at the grade-subject level. Teachers in a grade earn money based on the performance of all students in the grade. This means it is a group-incentive, equally shared, in case a focal subject in a grade is taught by more than one teacher.
2) It has multiple thresholds. Thus, it pays teachers based on how many skills their students are able to master. The more skills (thresholds) the student masters, the more money the teachers get.
3) Harder skills (with lower overall pass-rates) earn the teachers more money.


Besides the teacher bonus, there is a head teacher bonus which equals 20 percent of the bonus total paid to subject teachers in his/her school; and an infrastructure bonus --- one per region --- paid to the school that, among a group of weak performers on the Standard Four National Assessment, performed best on the Standard III test.

The bonus budget for teachers is TZS 170 million ($\sim$ USD 74,000); for head teachers TZS 34 million ($\sim$ USD 14,800); and TZS 9 million ($\sim$ USD 4,000) for the infrastructure bonus in each of the six regions in the sample.

The KiuFunza III budget includes a teacher bonus pool that is based on an expected teacher bonus equal to about 4 percent of annual teacher salary. This bonus size is in line with previous KiuFunza implementations. In our experience it is sufficiently high to attract teacher attention, without it being so high that it could not be afforded out of future pay increases.

The curriculum for Standards I-II-III has a series of competencies that students are required to master in each grade for Kiswahili and Math (Arithmetic). KiuFunza rewards teachers for proficiency levels: the attainment of each of the curriculum skill levels by a student is rewarded. The size of the per skill pass bonus depends on difficulty of the skill and is determined by the total number of students passing in the sample. A student who masters more skills earns his/her teacher a higher amount.

At the end of the school year, we will test a sample of students in grades 1 and 2, and all students in grade 3. The total budget is split across grades in proportion to the number of students enrolled in each grade. The budget is then divided equally among subjects and skills within each subject. For example, suppose the budget allocated to one grade for demonstrating proficiency in addition (a math skill) is US\$1,000. If there are 500 students in the grade, and 250 pass the addition portion of the math test, then a teacher would receive US\$4 for every student in her class that was proficient in addition.

Since the per pass bonus paid ex-post is equal to the skill budget divided by the number of students passing the skill, the budget for easier-to-obtain skills is spread across more students --- resulting in a lower per-pass bonus. Conversely, harder-to-obtain skills have a higher per pass bonus. Thus, teachers have the potential to earn larger bonuses if their students are proficient in a larger number of skills, especially harder-to-obtain skills.

2019-01-01

2020-12-31

Learning outcomes for children
Behavioral changes for teachers

The KiuFunza III intervention will be implemented in an experimental sample of 100 public primary schools from six regions across Tanzania. These six regions are Kigoma, Simiyu, Tanga, Singida, Songwe and Pwani.

The KiuFunza III incentive program will be evaluated as part of a randomized experiment coordinated by the RISE Tanzania research team. This research design aims to evaluate two programs: (1) the new School Quality Assurance (SQA) program implemented by the Ministry of Education, Science and Technology; and (2) the KiuFunza teacher incentive program.

The treatment was randomized at the school level, stratified by district.

Randomization done in office by a computer

School

Yes

397 schools

Roughly, 12,000

\begin{table}[H]
\caption{Treatment assignment across schools\label{tab:cross}}
\centering
\begin{tabular}{lccc}
\toprule
& \multicolumn{2}{c}{KiuFunza Incentives} & Total \\

SQA & No & Yes & \\
\cmidrule(lr){2-3}\cmidrule(lr){4-4}
No & 149 & 50 & 199 \\
Yes & 74 & 25 & 99 \\
Yes+Reminders & 74 & 25 & 99 \\
\cmidrule(lr){2-3}\cmidrule(lr){4-4}
Total & 297 & 100 & 397 \\
\bottomrule
\end{tabular}
\end{table}

We set power at 0.8 and the size at 0.95. We estimate the intra-cluster correlation to be 0.15 using data from Control schools in another RCT in the same region. Finally, we assume that student and school-level covariates (e.g., previous test scores) explain around 53% of the variance in the outcome. Under these assumptions, the MDE for the treatment effect of teacher incentives is 0.19 in Grades 1 and 2, and 0.16 for Grade 3. The MDE for the treatment effect of teacher incentives+SQA (or teacher incentives+SQA+text messages) is 0.23 in Grades 1 and 2, and 0.19 for Grade 3. To test the difference between teacher incentives and the other two treatments, the MDE of the difference between treatments is 0.27 in Grade 1 and 2, and 0.22 in Grade 3. To test the difference between the two treatments that have add-on features, the MDE of the difference between treatments is 0.32 in Grade 1 and 2, and 0.26 in Grade 3.