Low-cost private schools are popping up rapidly in many parts of the developing world, especially India where even in rural areas 28% of students attend private schools. Should governments be supporting these schools as a cheap way to boost learning for the poor? Or is privatization reducing equity and undermining public institutions? A year ago I participated in a somewhat heated online debate on this topic, see here and here. Since then, I've tried to stay on the sidelines, out of a sense that there was no new evidence and not much new to say.
As of yesterday, that changed. After six years of work, Karthik Muralidharan of U.C. San Diego and Venkatesh Sundararaman of the World Bank released the results of their much anticipated, large-scale randomized trial of a school voucher program in Andhra Pradesh, India. This promises to be one of those landmark studies that will shape the debate on an important policy question for years to come. It's also timely, emerging just as India starts implementing a program of public subsidy for private schools at an unprecedented scale as part of the new Right to Education Act. Here are seven questions we can finally answer:
1. Do low-cost private schools improve learning?
Before the experiment started, Muralidharan and Sundararaman found that students in private schools scored dramatically better on standardized tests of math and Telugu, the local language. While we all know correlation is not causation, these gaps seemed too big to dismiss. Surely, a disparity of more than half a standard deviation couldn't be entirely due to the sorting of advantaged kids into private schools, right?
As it turns out, yes it could.
The results of the randomized evaluation should be sobering to private school cheerleaders: after two years of the originally-scheduled four-year study, voucher winners saw absolutely no relative gains in math or Telugu scores. Zero, zip, nada.
1b. Something tells me you're hiding something. What else did they find?
These results were really puzzling because the private schools appeared to do better on every measure of processes that the study collected data on based on surprise school observations. The private schools had longer school days, a longer school year, lower teacher absence, higher rates of teaching activity, significantly lower pupil-teacher ratios and less multi-grade teaching .
Then a colleague suggested that maybe in addition to teaching better or worse, private schools were teaching different things. The authors examined detailed school time table data, and found that private schools were spending much less time teaching math and Telugu than public schools, and much more time on English, Hindi, Science, and Social Studies.
At the end of four years, they measured effects across all subjects and found a positive, statistically significant effect. Voucher winners scored about 0.13 standard deviations higher across all tests, and kids who actually used the voucher and went to a private school gained about 0.23 standard deviations.
The moral of the story is that while private schools are more productive than public schools, they also tend to focus on different things.
2. Don't private schools get an unfair advantage by spending more money on the kids?
Well, yes and no. Yes, private schools provide far better school facilities (more electricity, more computers, more toilets) and lower pupil-teacher ratios (17:1 rather than 26:1), but private schools actually spend only about a third of what public schools spend per pupil.
(That last fact sheds a somewhat more positive light on the finding of zero impact of private schools on math and Telugu scores: same scores, one-third the price.)
But how can all those extra amenities be cheaper? The big savings come in teacher salaries. Private schools hire younger, less-educated teachers -- who are more likely to be female and from the local community -- and pay substantially lower wages. Private school teachers in AP earn about one-sixth what their public school counterparts do. This wage disparity makes private schools a great bargain, but of course private school teachers would reasonably like to see those costs creep up a bit.
3. Aren't private schools just skimming the richest and the brightest?
To belabor the point, the randomized evaluation design used here allows Muralidharan and Sundararaman to measure the clean, causal impact of school choice on test scores, with no confounding effects from any initial differences between students (even in unmeasured skills or aptitude).
But it's still possible that vouchers had a positive effect on scores, but only the rich captured these gains.
If we look outside the experiment, at enrollment in private schools in Andhra Pradesh when no vouchers are available, there is no doubt that private schools serve less poor families. At the baseline, the authors found that kids in private schools had parents who were about 20% more likely to have completed primary school, 23% less likely to be from a scheduled caste, and owned significantly more household assets.
The vouchers were designed specifically to erase that difference. Did they succeed at overcoming the equity-efficiency trade-off?
To a surprising extent, yes they did. A common worry about voucher programs is that they will subsidize the flight of wealthier families to the private sector, but poorer kids will not participate -- either because the subsidy is too small, or their parents just aren't able or interested in taking advantage of the vouchers.
In this case, Muralidharan and Sundararaman find no evidence that better-off kids benefit more. There was no relationship between baseline tests scores, scheduled caste membership, parental literacy, or household assets and the probability that a child (a) signed up for the voucher lottery, or (b) actually went to the private school if they won the lottery. Efficiency, meet equity.
4. Won't private competition undermine public schools?
The Andhra Pradesh experiment was carefully, and very cleverly, designed to study not only the effects of private schooling on voucher winners, but also the spillover effects on nearby public schools and on children 'left behind' in those schools.
Families had to choose whether to apply for vouchers before they knew whether their village would be randomly chosen to be eligible. So Muralidharan and Sundararaman can compare kids who didn't apply for vouchers in treatment villages to kids who didn't apply in control villages. That allows them to study the effect of the village level voucher program on non-applicants, i.e., those 'left behind'.
Long story short, they find no negative effects on kids left behind.
For what it's worth, they can also estimate the externalities on another group of kids you're probably less concerned about: the rich kids in private schools who suddenly faced an influx of poor kids in their classrooms. As it turns out, they experienced no effects either. This turns out to be a very important result in the context of India’s Right to Education Act, which may be one of the largest ever attempts to integrate schools across socio-economic classes in history (and which has been resisted by elites).
5. Will private competition spur public schools to perform better?
As it turns out, no. See previous answer. That's the negative flip side of the result that children left behind suffered no ill consequences. They didn't suffer any positive consequences either. At least in Andhra Pradesh, public schools didn't respond to the new competition.
For readers with deep-seated faith in the benefits of competition, the authors acknowledge that this question was really beyond their scope. It’s possible supply-side responses to competition might only emerge in a bigger voucher program conducted over a longer period -- here the vouchers were only offered to 2 grade cohorts at one point in time.
6. This all seems a little bit too good to be true. What else should I be asking you?
As the authors stress, there are two features of this voucher program that are relevant to India's Right to Education act, but not necessarily applicable to all such programs.
First, the vouchers covered the full cost of tuition, fees, textbooks and school supplies. So there were no additional top-up fees to prevent poor households from taking equal advantage of this program. When vouchers only cover part of the cost of private schooling, they may end up making it cheaper for the rich to isolate themselves in expensive schools.
Second, private schools which participated had to agree to accept whoever the researchers sent them -- there could be no selective admissions. This matters not only for equity, but also for the role of competition in raising performance. if schools can simply select the brightest students, they can earn a good reputation as a 'selective school' without adding any value. Randomized admissions, as used in many American charter schools, help force schools to compete on value added.
Of course, there is no reason why these features couldn’t be replicated elsewhere. But it is important to note that the ‘design’ of systems matters, and can’t be reduced to a generic comparison of public vs. private schools.
7. Is this the end of the low-cost private schools debate?
Hardly. Nor should it be. For readers opposed to private schooling, here's a handy guide for how to argue with research results you don't like. What works in Andhra Pradesh might not work everywhere.
But showing something worked in a representative sample of rural locations in a state of 80 million people is nothing to scoff at. And even if you're interested in policies outside India, I think it's fair to say the burden of proof on the school vouchers debate in poor countries just shifted a little bit.