On this page
In September 2003, an NGO in rural Udaipur handed some of its teachers a camera and changed how it paid them. The result, published by Esther Duflo, Rema Hanna and Stephen P. Ryan in the American Economic Review in 2012, is a widely cited randomised test of whether getting teachers to turn up changes what children learn.
What did the Rajasthan teacher camera study test?
The question: if you check whether teachers come to school and pay them for the days they do, does absence fall and do children learn more? The setting was non-formal education centres (NFEs) run by the NGO Seva Mandir in the tribal villages of Udaipur district, not government schools. According to the authors' 2010 working-paper manuscript, each centre had one teacher from the local community, usually with a Class 10 education, and about 20 children, most aged 7 to 10, learning basic Hindi and maths.
Absence was the problem: the same manuscript records a 44 percent teacher absence rate at the August 2003 baseline, before the programme began.
How did the cameras and pay work?
Seva Mandir picked 120 centres and assigned half at random to the programme; seven closed before it started for unrelated reasons, leaving 57 treatment and 56 comparison centres, as the 2010 manuscript explains. Details below are from it too.
A student photographed the teacher and classmates at the start and end of each day, with a tamper-proof date and time stamp.
A day counted only if the two photos were at least five hours apart and showed at least eight children.
Comparison teachers got a flat Rs 1,000 a month; camera teachers earned Rs 50 for each valid day beyond 20 and lost Rs 50 for each day missed below 20, with the fine capped at Rs 500, so pay ranged from Rs 500 to Rs 1,300.
An independent team measured attendance in both groups through one unannounced visit a month.
What did the study find?
Teacher absence fell by 21 percentage points relative to the comparison group, and children's test scores rose by 0.17 standard deviations, according to the published abstract.
J-PAL's evaluation summary puts the same result another way: teacher attendance was 58 percent in comparison centres and 79 percent in camera centres. The other findings below come from the 2010 manuscript, not the published abstract:
Teachers did not teach less once present: they were as likely to be using the blackboard and addressing students in both groups.
Children in camera centres got about 30 percent more days of instruction, roughly 2.7 extra days a month.
By February 2006, 26 percent of camera-centre children had moved into government schools, against 16 percent in comparison centres, a 62 percent increase.
Using an instrumental-variable estimate, the authors suggest a 10 percentage point fall in absence would raise test scores by about 0.10 standard deviations, an extrapolation they say needs caution.
The published abstract adds that the authors' structural model finds teachers respond strongly to financial incentives. J-PAL puts the cost at about Rs 5,379 per teacher per year.
Does it work in government schools?
That is the big limit. These were NGO-run centres staffed by para-teachers, who the authors argue are easier to monitor and incentivise than government teachers, partly because para-teacher contracts are renewed yearly. The evidence is also mixed when government tries it. In a related Udaipur trial by Banerjee, Duflo and Glennerster, time-clock monitoring of government nurses first raised attendance, but within about six months the local health administration began excusing absences and the gap closed. J-PAL's own summary says the camera programme was not designed to be a scalable intervention.
So the study shows attendance can be raised and extra days turn into learning. It does not show cameras suit every classroom, and says nothing directly about Class 9 to 12 students. It sits alongside other Indian trials such as Teaching at the Right Level and the Balsakhi and computer maths trials, which also found that basic instruction time and targeting matter.
What can students and parents take from it?
This part is our inference, not a finding of the study: lost teaching days are lost learning, and they add up quietly.
A worked example: if your physics teacher misses two periods a week across a 30-week term, that is 60 periods gone. A plan for when it happens:
Keep a simple log of cancelled or missed classes per subject, so the gap is visible.
Within 48 hours, read that day's NCERT section yourself and note three questions you could not answer.
Take those questions to the next class, a classmate, or a teacher in a free period.
Parents: ask the school, politely, how substitute classes are arranged.
If you are working through a missed topic alone, Fermi's AI tutor can work out what you already know and build a plan for the session, so you start with a clear next step.



