Comments

Miguel Urquiola: Mizala and Romaguera present a thorough and useful review of empirical research on educational quality in Latin America, particularly as it relates to measurement and analysis using test scores. They ably distill the debate in this area into a few key issues, which is in itself an important contribution. In discussing so many different perspectives and results, however, the paper does not sufficiently emphasize a key point, namely, that as a result of empirical and theoretical difficulties, general knowledge on how to improve educational outcomes is, in fact, very weak, in the sense that it is insufficient to warrant the unambiguous policy prescriptions often observed in the literature.

My comments here illustrate this point for two of the central issues the authors address: the question of which inputs (like textbooks, teacher training, and class size) raise outcomes most cost effectively and the extent to which an expansion of the private sector would improve educational outcomes.

Educational Inputs

Economists would be well placed to influence educational policy if they could credibly identify which inputs are likely to raise educational outcomes most cost effectively. As discussed by the authors, the first difficulty in approaching this issue is empirical. Put briefly, the extent to which a child enjoys a given level of inputs (such as textbooks or small classes) is unlikely to be independent of other characteristics (including parental education) that also affect his or her achievement. This complication is severe because such correlations do not always go in the expected direction. For instance, in an earlier paper I consider class size, which is perhaps the most widely studied educational input, and find that in the case of Bolivia, children from lower socioeconomic levels are taught in smaller classes. Naïve interpretations of the data, like those that emerge from standard ordinary least squares (OLS) regressions, suggest that increasing class sizes would raise achievement. ¹ Because of such correlations, the literature is full of contradictory findings.

To make matters worse, such complications are not even purely empirical in nature. Lazear argues that one might actually expect that on the surface class size would be empirically unrelated to achievement. ² This would happen, for instance, if the optimal class size is larger for better-behaved students, and schools take this into account in setting its level.

These considerations suggest that until further quasi-experimental work takes place in Latin America, there is little anyone can say with certainty on this issue. This may warrant a more skeptical reading than the authors give to existing results.

Expanding the Private Sector

To determine whether an expansion of the private sector should be a policy goal, it is key to ascertain whether private schools are indeed more effective than public institutions. In other words, all other things being equal, would a given student perform better in a private than in a public school? The same complications discussed above arise in empirically evaluating this issue: namely, students' own characteristics are not unrelated to their probability of enrolling in private school.

As above, the implication might be for further experimental or quasi-experimental research, such as Angrist and others carry out for Colombia. ³ The ideal situation would be to randomly select a group of students to transfer from the public to the private sector; if they performed better than those who remained behind, one might conclude that private schools are more effective. Unfortunately, here again the complications are not purely empirical. Hsieh and Urquiola suggest that even experimental work might not suffice, because an experiment does not guarantee that all other things are equal. ⁴ What an experiment does is raise the likelihood that the group of students transferred out of public schools is identical to those left behind. It is still true, however, that those who transferred would be more likely to benefit from, say, better peer groups and better trained teachers (to the extent that, on average, private schools are better endowed with those inputs). These factors, rather than any difference in...