On this page
In 1996 a charity gave free official textbooks to randomly chosen primary schools in rural western Kenya. Test scores did not rise on average. The study behind that result, Glewwe, Kremer and Moulin, "Many Children Left Behind? Textbooks and Test Scores in Kenya", appeared in 2009 in American Economic Journal: Applied Economics. The published article's full tables are behind a paywall, so the figures below come from the authors' 2007 NBER working-paper version and are labelled as such.
What did the Kenya textbook study test?
Do pupils in poor schools learn more with textbooks? Earlier non-randomised studies said yes. The NBER working paper notes that the four papers examined in one 1988 review of developing-country studies reported gains of 0.34, 0.36, 0.30 and 0.06 standard deviations.
In Busia and Teso districts, before the programme, 80% of pupils sat in classrooms with less than one English textbook for every 20 children, according to the NBER working paper and J-PAL.
Sample: 100 lower-performing primary schools thought to need help, split into four groups of 25 (J-PAL summary).
Treatment: in early 1996, 25 schools received English textbooks for grades 3 to 7, maths books for grades 3, 5 and 7, and science books for grade 8.
Ratio: about one book for every two pupils in maths and 60 books per 100 pupils in English and science, since children already shared.
Comparison: after year 1, the other 75 schools, not yet assisted; all were tested on exams modelled on Kenya's government papers.
Groups were formed by taking every fourth school from an alphabetical list, which the authors treat as random assignment.
What did free textbooks change in Kenya?
On average, almost nothing. After one year, the effect across all subjects was 0.02 standard deviations and not statistically significant, the NBER working paper reports. With their most precise method, the authors could reject an average gain of 0.07 standard deviations or more.
For scale, the working paper explains that 0.1 standard deviations moves a pupil from the 50th to the 54th percentile.
The picture changed when pupils were sorted by their pre-test scores into five equal groups. In year 1 the estimated effects, from the weakest fifth to the strongest, were -0.05, -0.02, 0.03, 0.14 and 0.22 standard deviations (NBER working paper, Table 9). Only the top two fifths gained significantly (at the 10% and 5% levels), as J-PAL also summarises.
Grade 8 pupils, already a selective group, were more likely to enter secondary school the next year: 43% in textbook schools against 38% in comparison schools, according to the working paper. Books did not reduce dropout, repetition or absence in lower grades.
Why did only the strongest pupils gain?
The books were in English, the medium of instruction from grade 4 but the third language for most of these children. When researchers returned in the fourth year and asked the median pupil in each grade in 50 schools to read from the English textbook, only 16% in grade 3 and 28% in grade 4 could read it (Table 11, NBER working paper).
The authors argue that Kenya's curriculum was pitched at the strongest pupils, leaving the rest behind.
What are the study's limits, and has it held up?
The study covers one region, one kind of book and mainly two years, and measures test scores rather than long-run outcomes. The published 2009 abstract reports the same conclusions as the working paper.
Later work points the same way. A randomised trial of a government textbook programme in Sierra Leone, by Sabarwal, Evans and Marshak (2014), found that in many schools a large majority of the books were stored rather than handed to students, with no impact on student performance. A 2013 review in Science by Kremer, Brannen and Glennerster concluded that test scores are unresponsive to more-of-the-same inputs such as textbooks, while teaching matched to pupils' learning levels is highly cost-effective. That is the idea behind Teaching at the Right Level trials in India, and it connects to the Kenya tracking study by Duflo, Dupas and Kremer.
What should Indian students take from the Kenya textbook study?
What follows is our inference, not a finding of the study. Most Indian students have books; the Kenya result suggests asking whether you can use the one you have.
Try this three-step check on the next chapter you study for Class 9-12, JEE or NEET:
Read one page cold. If more than two sentences lose you, the book is ahead of you.
Find the missing step. Write down the earlier idea the page assumes, such as vectors before projectile motion or moles before stoichiometry.
Go back before going forward. Spend one session on that earlier idea, then return to the page.
If the gap is a concept rather than a page, the Fermi Concept Library has every concept in physics, chemistry, biology, mathematics and computing, each with its own lesson and its place in the map, so you can step back to the idea you skipped.



