On this page
Many parents believe a private school will raise marks. But average results mix the school's effect with the kind of families who choose it. In 2015, economists Karthik Muralidharan and Venkatesh Sundararaman published a lottery-based test of that belief in The Quarterly Journal of Economics.
What did the Andhra Pradesh voucher experiment test?
Does moving a child from a government school to a low-fee private school improve learning, and does it hurt anyone left behind? The project ran in five districts of undivided Andhra Pradesh across 180 villages that each had at least one recognised private school, starting in the 2008-09 school year, according to the published paper.
The voucher paid fees, books, uniforms and shoes for the rest of primary school.
The design used two lotteries:
Villages were drawn by lottery: 90 became voucher villages and 90 carried on as usual.
Inside voucher villages, a second lottery chose which applicant families got a voucher.
Out of 10,935 eligible households across all 180 villages, 6,433 applied. In the 90 voucher villages, 3,097 households applied, 1,980 won a voucher by lottery and 1,210 used it to enrol in a private school, the authors report. Because a lottery decided who got a voucher, later differences reflect the school, not the family.
Did private schools raise test scores?
After two and four years, lottery winners scored no better than lottery losers in Telugu and maths, the study found. In ordinary comparisons, private school pupils were 0.65 standard deviations ahead in these subjects; the lottery suggests most of that gap reflects who attends, not what the school adds.
Private schools also spent 40% less class time on Telugu and 32% less on maths, and used the time for English, science and social studies, and Hindi, which government primary schools did not teach, according to the published paper.
After four years, the effect of winning a voucher was 0.12 standard deviations in English, 0.08 in science and social studies, and 0.55 in Hindi. Averaged across all subjects, voucher winners scored 0.13 standard deviations higher, and children who actually attended a private school scored 0.26 higher. The authors note that Hindi drives this average; leaving Hindi out, the average effect is not significantly different from zero.
Were private schools cheaper, and did anyone lose out?
Private schools in the sample spent about Rs 1,849 per child per year against Rs 8,390 in government schools, and their teachers earned about Rs 2,607 a month against Rs 14,286, the paper's tables show. On unannounced visits, 9% of private and 24% of government school teachers were absent.
The village lottery let the authors check for spillovers: they found no evidence that pupils who stayed in government schools, or children already in private schools, did worse when voucher winners moved.
What does the study not show?
The authors list caveats themselves. Private schools did not do better in maths and Telugu despite longer days and lower absence; one explanation they offer is that teaching in both follows the textbook, not the child's level. If so, they write, private schools' productivity edge may mainly reflect lower teacher salaries, not better teaching. Evidence that switching to English medium held some children back is only suggestive. The study covers primary school in rural Andhra Pradesh, low-fee schools and a voucher worth about 40% of government spending per child; it says little about secondary school, cities or expensive schools.
A value-added study by Abhijeet Singh (2015) using Andhra Pradesh panel data found private school effects varied by subject, age and place, with no evidence of a positive effect in urban areas. For the Right to Education Act's 25% private school quota, a lottery study in Chhattisgarh by Romero and Singh (2022) found that over three-quarters of applicants who missed a quota seat went to a private school anyway as fee-paying students.
The fair reading is narrow: here, low-fee private schools matched government schools in core subjects at far lower cost, and added Hindi. It does not settle whether public or private schooling is better in general.
What can Indian students and families take from it?
These are our inferences, not findings of the study:
Ask about the child, not the average. A school's results list reflects who enrols; ask how much children like yours improve. Correlation is not causation in school results either.
Count the hours per subject. Here, the subject mix shaped what children gained: Hindi rose where it was taught, so check how many periods a week your weak subject gets.
Weigh a change of medium carefully. Switching language and content at once is a double load.
Fix the level gap first. If you are behind, catching up at your level matters more than the school's label; see the Teaching at the Right Level trials and the Mindspark study in Delhi.
If you are not sure where your own gaps are, the Fermi Concept Library has a lesson for every concept in physics, chemistry, biology and mathematics, so you can start from what you do not yet know.



