Some people dwell on emptiness even in glasses that are three-quarters full. Thats how it is with assessments of the research on school choice. Although the research supporting school-voucher programs has greater breadth and depth than research on almost any other education policy, opponents of choice continue to chant their mantra that the results are mixed and inconclusive. As a congressional vote approaches on offering vouchers to students in Washington, D.C., it is worth reviewing how strongly existing research supports the expansion of school choice.
* * *
Consider that the research on the benefits of school choice for those who use a voucher to attend a private school includes evidence from five random-assignment experiments: programs in Charlotte, Dayton, Milwaukee, New York, and Washington, D.C. Random assignment, the approach commonly used in medical research, is the strongest social-science research design because it creates virtually identical treatment and control groups. Any difference in outcomes for the two groups over time can be attributed to the “treatment” as opposed to some pre-existing factor.
Researchers have conducted seven analyses of these five random-assignment experiments. None of them finds that students are harmed by receiving a voucher to attend a private school and all find positive estimated effects, although the results vary in whether the benefits exist for both math and reading, whether those benefits are restricted only to African-American students, and in the statistical confidence that one can have in these positive estimated effects.
Even the analysis of the New York experiment by Princetons Alan Krueger, which was trumpeted recently in a New York Times piece (by Michael Winerip) as if it were a fundamental challenge to the research on vouchers, reports positive estimated effects for African-American students who use a voucher to attend a private school. Dr. Krueger only differs from previous researchers over whether this positive effect can be found at a high level of statistical significance. And his results only fall short of statistical significance after he makes a series of poor research choices, such as failing to control for prior test scores and measuring student race in an inappropriate way.
But whatever one may think of the dispute over the results of this New York experiment, it is simply not true, as Mr. Winerip alleges in the Times, that “it is scary how many prominent thinkers in this nation of 290 million were ready to make new policy from a single study . . . .” This fear-mongering preys upon popular ignorance. The evidence supporting benefits for voucher recipients is drawn from the aforementioned five random-assignment experiments, not just one, although you would never know that from reading the Times story.
Those who dismiss the research on school choice also ignore a growing body of evidence regarding the effect choice has on public schools. The party line, of course, is that vouchers and other forms of school choice harm public schools. The academic research, on the other hand, actually suggests that school choice raises test scores in public schools by providing those schools with an incentive to improve. Studies in Milwaukee, San Antonio, Florida, Michigan, Arizona, Maine, and Vermont have found that public schools rise to the challenge of school choice, improving their performance in order to compete effectively.
Waiting for all researchers to agree is like waiting for Godot -- it isnt going to happen. As long as academics are rewarded for engaging in controversies, and as long as there are competing interests, scholars are likely to disagree about how best to interpret data from these experiments. But if one steps back from each particular feud and sees the consistently positive results and the absence of any negative results from voucher research, an honest assessment would note the remarkable strength in the evidence supporting vouchers. Theres at least enough encouraging evidence to support trying another pilot program, like the one being proposed in Washington, D.C.