Robin and I are supporters of low-cost private schools in developing countries ever since reading Tooley’s incredible “The Beautiful Tree”.
So I was both happy and saddened to see this awesome paper by Lant Pritchett and Yamini Aiyar, called Value Subtraction in Public Sector Production: Accounting vs. Economic Cost of Primary Schooling in India.
It turns out that the median public school spent 14,600 rupees per pupil in 2011-12, while the median private school cost around 6,000 per pupil. In other words, Public schools require twice as much resources to deliver a year of education that the private sector. When you consider how many school kids there are in India, that is quite a large amount of “wasted” resources.
But the story gets weirder, because private school kids learn more than public school kids! Lant and Yamini use test scores to create an amount learned metric and then show that given the structure of public schools, it would cost almost 30,000 rupees per pupil to get their learning up to the level enjoyed by the private school kids (which is achieved at a cost of only 6,000 per pupil).
Now any of my grad students reading this will be yelling “selection bias” at their screens at this point.
The paper acknowledges the issue:
“we don’t adjust for student selection effects and hence our estimates are not estimates of “true” learning productivity effects across the two sectors. It is obvious that if higher socio- economic status of a child’s household is associated with better learning outcomes (and it typically is) and if children in private schools are more likely to be from higher socio- economic status (and they typically are) then the differences in costs and learning outcomes reflects both higher productivity of private schools and the demographic composition of students.”
and argues from other studies that the composition effect can account for something between 20 and 60 percent of the observed gaps in public / private outcomes.