School vouchers are making a comeback. Under legislation championed by Governor Bush, the number of Florida students using vouchers, either because they are disabled or were attending struggling public schools, has crossed the 10,000 mark. Utah’s governor has signed into law a new voucher program for students with disabilities. Participation in Milwaukee’s voucher program is nearing the legal limit, 15% of public-school enrollment. Meanwhile, Mr. Bush has raised the ante further by asking the Florida Legislature to provide vouchers for all low-performing students.

All this, after a two-or-more-year quiescence that followed the 2002 Supreme Court’s 5-4 decision finding vouchers constitutional. If the voucher movement narrowly avoided a beheading by nine justices in black robes in 2002, it came close to suffering death by a thousand cuts administered by a newly aroused opposition in 2003 and 2004. Voucher proposals came to a standstill in state legislatures everywhere except Colorado, where the State Supreme court killed the idea, at least for now. Only a very limited voucher bill for students in Washington, D.C., was enacted by Congress.

School reformers instead concentrated on the implementation of No Child Left Behind, or NCLB. To its credit, the federal law neatly combines accountability and parental choice. Either public schools must perform well, or parents must be given the choice of another, non-failing public school within the district.

If effective, the law could give new learning opportunities to students whose families exercise choice. No less important, it could stimulate effort in low-performing schools by threatening their enrollments.

But the choices offered parents under NCLB fall far short of President Bush’s original proposal, which, as in Florida, would have given vouchers to students attending schools deemed inadequate. That idea proved too controversial to navigate the congressional labyrinth. So, too, did any suggestion of opening seats in suburban districts to inner-city residents. Minority and disadvantaged students were left with only a limited, within-district set of choices administered by local district bureaucrats with scant incentive to promote movement among schools. Predictably, less than 1% of all eligible students are taking advantage of the choices offered them as a result of NCLB.

Still, if not many students are benefiting from new opportunities, the fear of losing students might still be enough to stimulate failing public schools to improve. Is public school choice under NCLB enough to promote school improvement?

It is now possible to examine this question systematically in Florida, side by side with a similar study of the voucher component of that state’s own accountability scheme. Vouchers remain available to Floridians in all those schools that received an “F” grade under the state accountability system twice in any four-year period. Roughly 2% of Florida’s elementary schools received that grade in the summer of 2002 under the revised Florida state accountability system. The dismal grade meant all students at these schools would become eligible for vouchers, if the schools received the same grade again the next year. The voucher threat was expected to give them an incentive to improve in the school year 2002-2003. Did it?

And what happened the next school year when a much larger number of schools were initially found to be “inadequate” under NCLB? Once so designated, the school, if it did not the next year make “adequate yearly progress,” faced the threat that parents could move their child to another public school within the district. Did that public-school-choice threat motivate higher performance?

Information from every elementary school student in Florida for the years 2002-04 allows us, for the first time, to answer these questions. To do so, we compared the performance on the Florida Comprehensive Accountability Test, or FCAT, of students in schools that received an “F” under the Florida accountability plan with those in “D” schools that narrowly missed being so designated. The “F” schools faced a voucher threat: One more “F” and students could leave for the private sector. We also compared the “D” schools, who suffered only the “stigma” of a very low grade, with otherwise comparable “C” schools.

When making these comparisons, we adjusted for the initial test scores and demographic characteristics of students as well as the quality of the peer group at the school. Also, by comparing very similar schools, we greatly reduced the likelihood that some other factor was at work.

As it turns out, the stigma of receiving a “D” by itself, motivated schools to work harder the next year. They outperformed by a significant margin the “C” schools that just missed getting the worse grade. Still, the “F” schools, faced with the threat of vouchers, did even better, raising scores by an amount roughly equivalent to three to four months of student learning above the performance of students in the “C” schools. They did so despite the fact that these “F” schools had highly disadvantaged, predominantly minority populations.

These are statistically significant, nontrivial results. If they continue over time, they would go a considerable distance toward achieving the very goals set out by No Child Left Behind.

Meanwhile, we found no impact at all of the threat of public-school choice under NCLB as it was implemented in Florida the next school year. Students in schools facing that threat gained no more than students in comparable schools not facing the threat. No harm done, but no good either. As Congress thinks about how to enhance NCLB, the voucher option should be put back on the table.

Mr. Peterson, a member of the Koret Task Force on K-12 Education at the Hoover Institution and the Henry Shattuck Professor of Government at Harvard University, is the director of its Program on Education Policy and Governance, in which Mr. West is a research associate. Their paper is available at www.ksg.harvard.edu/pepg.