Tyler mentioned, following a depressed Brad DeLong, a new paper on education vouchers in Chile that does not find large achievement gains. I have some criticisms of the paper (see below) but I was surprised that neither mentioned the most important recent paper on vouchers, Vouchers for Private Schooling in Colombia by Angrist, Bettinger, Bloom, King and Kremer in the Dec. 2002 AER.
Using data from a randomized experiment, Angrist et al. estimate that attending private school increased the probability of finishing eighth grade by 13-15 percentage points or 25 percent. Test scores increased by .29 standard deviations which is equivalent to about an extra year’s worth of schooling which has been estimated to increase yearly wages by 10 percent. Other markers such as teen cohabitation also improved.
Is this just a case of dueling papers? No, first, unlike Hsieh and Urquiola (HU), the Angrist et al. results are consistent with results found elsewhere. See in particular those found for Catholic schooling in the United States . Second, Hsieh and Urquiola (HU) are good researchers, judging by their paper, but Angrist et al. have a much more convincing research design – results from a randomized trial beat econometric identification any day. Cheer up Brad!
I shouldn’t give the impression that the results are directly comparable, however, as HU are trying to get at the general equilibrium effect of a voucher experiment and Angrist et al. are after the partial equilibrium effect of private schooling. Given the large gains found in the partial equilibrium literature, however, the GE results from HU are not plausible in my view.
Now regarding the HU paper some information is in order. First, there were no vouchers in Chile. Instead, there was public funding of some private schools on a per-student basis. Parents could not apply their voucher to the tuition at a private school of their choice.
Second, HU do not test whether students who transferred to private schools did better than other students – they tested whether aggregate scores (public and private) increased over time as more students attended private schools. Their evidence seems consistent with a nationwide decline in public school quality over time. More generally, I would have liked to have seen some information in their paper on the power of their tests. Given the size of the private sector what sort of gains could would we have expected to see in the aggregate scores and is their technique powerful enough to pick up such gains?
Third, HU claim that “cream skimming” was extensive but I find this difficult to believe because there is no price difference between public and private (voucher-accepting) schools since each was paid the same per-student amount. There are some non-pecuniary barriers but no limits on entry that HU mention.
Fourth, why did private enrollment increase if parents did not perceive a quality improvement? HU mention “freshly painted walls” which I thought was a bit flip – we ought to take revealed preference more seriously.
I do think that the HU study of Chile provides useful information about designing a good voucher program and my priors would have been that the program instituted in Chile, even though not a true voucher program, would have produced a larger effect – thus I learned something from the paper.