Evaluating School Choice

A recent study of Chile’s school choice program, which purports to show no significant improvement in student performance under the program, has touched off another round of Alex and Tyler’s mini-debate on educational vouchers (see here, here, here, and here). Like Alex, I continue to think that vouchers are a good idea. One study of one foreign country’s school choice program – in the context of other studies showing better results elsewhere – is not sufficient to make me a pessimist.

But what if it turned out that school choice programs (to clarify, vouchers are just one mechanism for implementing school choice) were not really effective in improving student test scores and other measures of performance? Would that necessarily mean school choice is a bad idea? I think not, for three reasons.

First, even if school choice delivered no improvement in performance, it might deliver the same level of performance at lower cost. I only skimmed the report on Chile, but I found no reference at all to the cost of education.

Second, there are reasons to favor school choice other than cost and performance, one of which is the value of choice in itself. Standardized performance measures cannot capture the value to students and their families of being able to get the kind of education they want. As I argued in one of the first posts on this blog, school choice has the potential to defuse a number of policy conflicts that result from the one-size-fits-all character of public education: bilingual education vs. immersion, prayer vs. no prayer, single-sex vs. coeducation, and so on. Even if all private schools performed equally on measures of basic skills, they could be producing greater value as perceived by the actual consumers.

Third, we should be skeptical about performance as measured by average or median scores on standardized tests, repetition rates, and drop-out rates (as in the Chilean study). Measures like these are unlikely to capture changes in the performance of the very best students. Median scores are completely unaffected by improvements in the performance of students at the top end of the scale. Average scores are affected, but possibly not much, because standardized tests typically cover basic skills that the very best students are likely to understand well anyway. (Taking calculus in high school was valuable to me, but it probably had little effect on my standardized test scores, because calculus is not a topic covered on those tests.) Repeat and drop-out rates have little to do with the best students, who are inclined to finish high school on time whether there’s choice or not. Yet the top students could be among the greatest beneficiaries of school choice, because they can attend schools that specialize in providing upper-level courses in the areas in which students are most motivated.

As a corollary to this point, we should also be skeptical of the claim that the “cream-skimming” apparently practiced by private schools in Chile (according to the study) is necessarily a bad thing. There are virtues to separating students according to their ability, and some of those virtues will not be picked up in the standard measures. (The cream-skimming in Chile seems to be related to socioeconomic status, but the authors themselves used socioeconomic status as a proxy for ability, and they do not try to separate the effects.) Public school systems often have Honors tracks and magnet schools, and while these programs have their critics, I have never heard them referred to as “cream-skimming.” The fact that the ability-level tracking might occur across schools, rather than within them, under a school choice program is hardly a compelling argument against school choice.