Making math mentoring work at scale: The development and roll-out of a cross-age tutoring program in Kenya

• In Event: Recycle, reuse, [reexamine], reintroduce: leveraging diverse data to build sustainable learning loops

Proposal

There is an increasing wealth of evidence showing that teaching appropriate to the student’s learning level can improve learning outcomes in low-income countries. However, teachers often lack the time (or incentives) to give each child instruction tailored to their needs. Contract teachers have been proven to be effective (see Duflo et al., 2015, Muralidharan & Sundararaman, 2013 and Kremer et al., 2013), but providing schools with extra teachers is expensive. Cross-age tutoring, where older students tutor younger students, is an inexpensive alternative to providing personalized instruction to younger students in that it substitutes a trained instructor (the teacher) with an untrained one (the older student) at the cost of the older student’s time.

We present the program development strategy and implementation learnings for a cross-age tutoring program at Bridge International Academies in Kenya and how diverse forms of data informed decision making within the organization.

Phase I of the program was to first “Fail Fast” at two schools to determine whether it was possible to successfully have older pupils tutor younger pupils for 40 minutes each day under teacher supervision. Phase I included deep qualitative research into what coaching and support would be needed to train older pupils to become good mentors and the logistics of when and where mentoring should occur.

Through this “Fail Fast” research, we re-designed the program and embarked on Phase II, beta testing with 30 schools. We collected both quantitative and qualitative data through observations, test-scores, and interviews. We found that teachers were able to sustain the program, tutees in the programme scored notably higher on the Maths exam, and a survey of over 200 parents of pupils in the mentoring programme found that almost all parents viewed the programme as “good for their child.”

However, we were concerned whether the program would continue to work at scale and whether tutoring for maths was the best use of time, relative to the alternative (an English tutoring program). Phase III involved a large randomized control trial (over 180 schools, 15,000 tutees, and 15,000 tutors) in Kenya. We find that cross-age tutoring in math has a small positive effect (0.06 SD, p-value of 0.073) on math test scores while the English tutoring program has a null effect on English test scores. We find considerable heterogeneity by the students’ baseline learning levels, with the effect largest for students in the middle of the ability distribution, while the point estimates are almost zero for students with either very low or very high baseline learning levels.

These results translate into policy conclusions. First, cross-age tutoring is more effective for math than languages. Second, these type of interventions can help the average student but may not be effective for stellar students and students who need more skilled assistance suggesting the need to develop supplementary programs for these types of students. Third, although the program has modest effect sizes, it’s essentially free, and therefore cost-effective relative to other alternatives to personalized instruction.

Author

• Lisa F Chen, Bridge International Academies