On this page
Can a prize for top students make a whole classroom learn more? In 2001 a non-governmental organisation in western Kenya offered scholarships to the best-scoring girls in some primary schools but not others. Michael Kremer, Edward Miguel and Rebecca Thornton tracked what happened in "Incentives to Learn", published in the Review of Economics and Statistics in 2009. The answer went well beyond the winners.
What did the Kenya girls' scholarship study test?
The Girls' Scholarship Program ran in two rural districts, Busia and Teso. Of 127 primary schools, 64 were chosen at random to take part and the rest served as the comparison group, according to the published paper. The baseline sample was 11,728 pupils registered in grades 5 and 6.
Grade 6 girls in programme schools who ranked in the top 15% of programme-school girls in their district on Ministry of Education exams won the award, which ran for two school years. Each year a winner's fees of US$6.40 were paid to her school, her family received US$12.80 for school supplies, and she was honoured at a school assembly. The authors note Kenyan GDP per head was about US$400.
How much did test scores rise?
Across both districts, girls in programme schools scored 0.19 standard deviations higher than comparison girls in the year they competed, as reported in the 2009 paper. In Busia the gain was 0.27 standard deviations; in Teso it was 0.09 and not statistically significant.
The authors estimate this equals about 0.2 of a grade's worth of primary school learning.
The same paper reports two more findings. Teacher attendance in programme schools rose by 4.8 percentage points overall and by 7.0 points in Busia, roughly halving teacher absence there. And the gains lasted: a year after the competition, cohort 1 girls in Busia programme schools still scored 0.24 standard deviations higher in the paper's main sample.
Did the scholarship help girls who could not win?
Girls in the bottom half of the baseline scores had almost no chance of winning: the paper puts their odds at 3% and 5%, against 55% for the top quarter. Yet their scores rose too, and the authors could not reject that the gain was the same at every level of ability.
In Busia, boys, who were not eligible at all, gained 0.15 standard deviations in the paper's main sample (significant only at the 90% level), about 60% of the 0.25 gain for Busia girls in the same sample. Among all baseline pupils it was 0.10 and not significant. The authors' leading explanations are better teacher attendance and students lifting each other's effort; their data cannot separate the two.
What does the study not show?
It did not work everywhere. In Teso, 5 of 58 schools pulled out, parental support was lower (58% of teachers reported it as positive against 90% in Busia), and the possible effect ranges from -0.17 to 0.23 standard deviations, so no firm conclusion is possible there.
Winners were more advantaged. Winners' parents had 10.7 years of schooling on average against 9.6 for others, so the money itself went more often to better-off families even if the learning gains spread.
It is one setting. These were rural Kenyan primary pupils in 2001-2002, not teenagers facing a national entrance exam.
On the worry that rewards kill the love of learning, the authors found no sign of it, and programme pupils were no more likely to seek extra tutoring.
A later follow-up by Friedman, Kremer, Miguel and Thornton surveyed 1,387 young women in Busia four to five years on. The figures here come from its 2011 NBER working paper, which had not been peer reviewed; a revised version appeared in Economica in 2016, and we have not checked its numbers. The working paper reports test scores still 0.208 standard deviations higher and an 8.7 percentage point rise in attending some secondary school, against a 30% rate in the comparison group.
What can Indian students and families take from it?
What follows is our inference, not a finding of the study: effort may spread, and a shared goal may lift a whole group.
Set a group target, not just a personal one. Three Class 11 friends aiming to cut silly errors in physics mock tests can keep each other going.
Make the goal public and specific. The Kenyan award came with recognition at an assembly. Tell a parent or teacher your target and test date.
Do not wait for a prize. The authors' leading explanations were pupil effort and teacher attendance, not the money. Plan focused study time without one.
If the hard part is knowing where to start each evening, Fermi's AI tutor works out what you already know and builds a plan for the session. For more classroom evidence, read what the Kenya trial on grouping pupils by ability found, what the Teaching at the Right Level trials found, and how one trial helped parents explain why maths and science matter.



