The surprising consensus on school choice

THERE has been a flurry of activity in school choice research in the last few years. As a result, where we used to have only theories and limited evidence we now have a relatively solid understanding of the likely effects of school choice. I say "relatively" because all research is necessarily imperfect and additional study can always improve the confidence with which we draw conclusions. But the research on school choice includes several random-assignment studies, the "gold-standard" of research design, where subjects are randomly assigned to treatment and control groups as in a medical study. I can think of only one other education policy issue (the effect of class-size reduction) that has been the subject of even one significant random-assignment experiment, let alone several "gold-standard" studies.

What is striking about the recent research on school choice is not just its quality but that it consistently reveals positive benefits from school choice. Of course, groups with vested interests in the results, such as the teachers' unions and their allies, always prefer to describe the results as mixed or inconclusive at best. (The tobacco industry similarly used to describe the research on smoking and cancer as mixed or inconclusive at best.) Meanwhile, journalists prefer covering controversy and fear crossing those who oppose school choice, so they describe results as mixed or inconclusive as well. And researchers have incentives to highlight disagreements with each other as a matter of academic pride and professional competition, which also helps obscure the general success of school choice programs.

The research on school choice essentially addresses three questions: (1) Does school choice benefit those who receive a school voucher? (2) Does it benefit students who do not actively choose a school (or as it is sometimes negatively framed, "those left behind")? and (3) How does school choice affect integration and the democratic ideals that we may wish schools to promote? The research on the first question, does choice benefit "choosers," is the strongest and most consistently positive. Whether choice benefits "nonchoosers" is more difficult to study, and therefore the evidence is less conclusive on that question, although recent research suggests benefits for choosers and nonchoosers alike. The last question, whether choice poses a threat to our democratic system, has been the least studied, perhaps because it is so central to popular faith in the public school. (Like any central myth, people prefer not to examine it too closely.) But some recent research suggests that choice may enhance the ability of s chools to promote democratic ideals. Let us consider the evidence on each of these three questions in turn.

Does school choice benefit choosers?

One indication of the academic effects of school choice on choosers is whether they report being more satisfied with their school experience than do nonchoosers. Here the evidence in support of school choice is unambiguously and overwhelmingly positive. John Witte, one of the evaluators of the school choice program in Milwaukee, reported that "satisfaction of Choice parents with private schools was just as dramatic as dissatisfaction was with prior public schools." In Cleveland, evaluator Kim Metcalf found that "across the range of school elements, parents of scholarship students tend to be much more satisfied with their child's school than other parents. Scholarship recipient parents are more satisfied with the child's teachers, more satisfied with the academic standards at the child's school, more satisfied with order and discipline, more satisfied with social activities at the school." Also in Cleveland, Paul Peterson, William Howell, and I found that after two years of the program, choice parents were si gnificantly more satisfied with almost all aspects of their children's education than was a random sample of parents from Cleveland public schools. Nearly 50 percent of choice parents reported being very satisfied with the academic program, safety, discipline, and teaching of moral values in their private school. Only around 30 percent of Cleveland public school parents report being very satisfied with these aspects of their children's schools. Very similar results were obtained from the privately funded school choice programs in Charlotte, Washington, D.C., Dayton, New York City, and San Antonio.

If this were almost any other policy realm or consumer issue, we might consider the strong positive effect of school choice on parental satisfaction sufficient evidence that the program benefits its participants. If, for example, people report that they are happier with the maintenance of public parks, we would usually consider this sufficient proof that efforts to improve the parks have succeeded. We would not feel obliged to count the number of items of trash and repair problems to verify reports of satisfaction.

But the standards for assessing programs in education are different. Many in the education and policy communities only give serious consideration to standardized test scores, and give little credence to parental satisfaction. They suspect that parents are not informed consumers of education or experience psychological pressures to justify their choices, making their assessments of program success unreliable. To put it bluntly, they suspect that parents are stupid. And thus, despite the overwhelmingly positive effects of school choice on parental satisfaction, these findings have not significantly influenced the policy debate.

Testing school choice

Instead, the debate has focused mainly on the effect of school choice on standardized test scores. These score results, including those from several "gold-standard" random-assignment experiments, have also been consistently positive, which gives them enormous credibility. In the last few years, there have been seven analyses of random-assignment school choice experiments, from five different programs, conducted by several different researchers. Every one of those analyses finds statistically significant benefits from school choice for those who are provided with opportunities to choose a private school.

For example, in New York City, Washington, D.C., Dayton, Ohio, and Charlotte, North Carolina, privately funded programs offered scholarships for private school. Because there were many more applicants than scholarships available, scholarships were awarded by lottery, allowing for the "gold-standard" random-assignment research design. Comparing the standardized test scores of those students who won scholarships to the scores of those who lost the lottery allows researchers to identify with confidence the effect of receiving a voucher. Since we can expect the two groups to be generally alike, differences in academic achievement between them are probably the result of the voucher, not differences in backgrounds or motivations.

The results from New York City, Washington, and Dayton were contained in a report issued by the Harvard Program on Education Policy and Governance (PEPG). The PEPG report found that after two years in private school, African-American recipients of the scholarships performed significantly better on standardized tests than did the African-American members of the control group that applied for a scholarship but were not given one by lottery. The benefit of receiving one of these privately funded vouchers in New York was about 4 percentile points, in Dayton about 7 percentile points, and in Washington about 9 percentile points. Interestingly, the PEPG report found that receiving the scholarship had no significant academic effect, good or ill, for students of any other ethnic group.

A second group, Mathematica Policy Research of Washington, D.C., the research company that was involved in collecting the data in New York for PEPG, conducted their own analyses of the New York data and issued their own report. Contrary to a misleading article in the New York Times, the findings from the Mathematica and PEPG studies were essentially the same. In fact, Mathematica calculated the average benefit for African-American students receiving a scholarship to be one-tenth of a percentile point higher than that reported by PEPG.

The only difference between the Mathematica and PEPG analyses of the New York results--a difference fully exploited by the New York Times in the throes of the presidential campaign--was the spin that each report placed on the findings. The Mathematica report preferred to emphasize the results broken down by grade, while the PEPG report focused on the average for African-American students across all grades. Breaking down the results by grade produced very small samples for each grade, so that the positive effect of receiving a voucher was only statistically significant in sixth grade, but not significant in two of the other three grades studied. Mathematica expressed worries that the benefit of the scholarship might not be widespread and therefore advised against drawing any conclusions. PEPG focused instead on the statistically significant benefit for African-American students across all grades and, in light of the similar results from other cities, felt comfortable drawing a stronger conclusion than Mathema tica did.

The difference between the two approaches can be illustrated by thinking about presidential campaign polls. Let's say that a national poll found Gore was ahead of Bush by seven points. Let's then say someone wanted to break down the results by state, even though the sample in each state is fairly small, and found that only in California was the Gore lead statistically significant. Would it then be more reasonable to conclude that Gore and Bush were tied or that Gore was ahead? The PEPG interpretation of the results is analogous to focusing on the national poll results, and the Mathematica interpretation is analogous to focusing on breaking down those results by state, even though the number of subjects is very few.

All of this discussion of the different interpretations, however, obscures a basic truth: Both the PEPG and Mathematica analyses of results from a high-quality random-assignment school choice experiment in New York find statistically significant benefits from school choice. And the PEPG report finds statistically significant benefits from the other two programs it covered, in Dayton and Washington, D.C.

Three other analyses of random-assignment choice experiments confirm the existence of academic benefits. My own analysis of the privately funded scholarship program in Charlotte found that students given by lottery a voucher to attend private school outperformed their counterparts who failed to win a voucher by 6 percentile points after one year's time. I was unable to determine whether benefits occurred exclusively for African-American students because more than three-quarters of the students in the Charlotte study were African American, leaving too few students from other groups about which to draw conclusions.

Two analyses of random-assignment data from the publicly funded school choice program in Milwaukee also found significant gains for students who received vouchers to attend private schools. One study, by Paul Peterson, Jiangtao Du, and myself, found that students who won lotteries to receive a voucher scored 6 percentile points higher on their reading scores and 11 percentile points higher on their math scores than students who did not receive a voucher. Cecelia Rouse, a Princeton University economist and former member of the Clinton administration, independently analyzed the data from Milwaukee and arrived at similar results, at least in math scores. After trying several analytical strategies Rouse concluded: "Students selected for the Milwaukee Parental Choice Program ... likely scored 1.5-2.3 percentile points per year in math more than students in the comparison groups."

The big picture

In addition to these seven random-assignment studies, there have been three nonrandom-assignment studies of publicly funded school choice programs. The quality of these nonrandom-assignment studies (including one I conducted) is so much lower than the quality of the random-assignment studies that less weight should be given to their results. But it is striking that these lower-quality studies are also generally positive in their findings.

For example, the Cleveland choice program offers evidence on the academic effects of choice, but of lower quality because there are no random-assignment data nor sufficient data on the background characteristics of choice- and public-school families. Despite these data limitations, analyses of test scores have been performed by Kim Metcalf of the Indiana University School of Education and by myself, Paul Peterson, and William Howell. Both groups found at least some significant academic benefits of the choice program in Cleveland.

Metcalf observed that "the results [after two years] indicate that scholarship students in existing private schools had significantly higher test scores than public school students in language (45.0 versus 40.0) and science (40.0 versus 36.0). However, there were no statistically significant differences between these groups on any of the other scores." Metcalf's analyses were based on a comparison between one grade cohort of choice students and a nonrandom sample of public school students. He had a very limited set of controls for background differences, which could seriously bias results. But Metcalf still saw fit to conclude in favor of school choice:

The scholarship program effectively serves the population of families and children for which it was intended and developed. The program was designed to serve low-income students while maintaining the racial composition of the Cleveland Public Schools.... The majority of children who participate in the program are unlikely to have enrolled in a private school without a scholarship.

The analyses performed by myself, Peterson, and Howell also had serious data limitations. We only had test scores from two private schools, although those schools did contain nearly 15 percent of all choice students and nearly 25 percent of all choice students who had transferred from public schools. We were only able to compare scores from students over time relative to how they scored when they first entered these two schools. But since inner-city students tend to have declining scores relative to national norms over time, any gains in test scores over time should be a strong indicator of academic progress for the choice students. We found that after two years students at the two schools we examined had gains of 7.5 national percentile points (NPR) in reading and 15.6 NPR in math. These gains were achieved even though the students at these two schools were among the most disadvantaged students in Cleveland. Thus, despite shortcomings in the available data, there were indications of significant academic ben efits for choice students in Cleveland.

Rather than getting lost in the details of these various studies, it is worth stepping back and reviewing the results as a whole. There have been seven random-assignment and three nonrandom-assignment studies of school choice programs in the last few years. The authors of all 10 studies find at least some benefits from the programs and recommend their continuation, if not expansion. No study finds a significant harm to student achievement from the school choice programs. The probability that all 10 studies would be wrong is astronomically low. It is also worth noting that the private schools participating in these various school choice programs tend to have per pupil operating costs that are nearly half the per pupil expenditure in the public schools. Even if we were to find no significant academic benefit from school choice, we might still endorse the policy because parents like it, and because it costs half as much money to produce the same level of academic achievement. To increase student achievement sig nificantly, while spending less money per pupil and making parents more satisfied, as the evidence from these 10 studies consistently shows, provides strong support for school choice.

Does school choice benefit nonchoosers?

If choice helps its beneficiaries, does it do so at the expense of others? The suspicion is that choice programs "cream" the best students from the public schools, draining talent and resources from the public system. On the other hand, it is possible that "creaming" has largely already occurred in the public system. Higher-achieving students and more affluent and involved families may have already chosen a public or private school that suits them, leaving "the rest behind." In fact, the U.S. Department of Education estimates that 59 percent of students currently attend "chosen" schools. But many of the remaining 41 percent lack the financial resources to move to a desired public school district or pay private school tuition. Can vouchers exacerbate the situation in a way that harms nonchoosing families?

As we have already seen, evaluations of the Milwaukee and Cleveland programs have concluded that the programs successfully targeted very low-income families, offering them opportunities they would otherwise lack. The average income of families participating in the Milwaukee program was $10,860. In Cleveland, the mean family income was $18,750; in New York, $10,540; in Washington, D.C., $17,774; and in Dayton, $17,681. In Milwaukee, 76 percent of choice students were from single, female-beaded households. In Cleveland, the figure was 70 percent. In Washington it was 77 percent, and in Dayton it was 76 percent. The standardized tests scores of choice students before they began private school averaged below the 31st percentile in Milwaukee, below the 27th percentile in New York, below the 33rd percentile in D.C., and below the 26th percentile in Dayton. In other words, choice students were generally performing in the bottom third academically. If this is cream, then none of us need go on a diet.

The most damaging thing one could say about all of these choice programs with respect to "creaming" is that they probably attract the more capable of the disadvantaged poor. But if this is "creaming," then Food Stamps, Temporary Assistance to Needy Families, and virtually all other antipoverty programs engage in it. These programs generally fail to serve the most dysfunctional of the poor, because these persons have difficulty taking full advantage of the programs designed to help them. This is normally seen not as an indictment against antipoverty efforts but rather as an unfortunate reality that all programs must face. Like these other antipoverty programs, school choice programs can be designed to target disadvantaged populations, even if they do not always reach the most disadvantaged of the disadvantaged.

The choice challenge

But showing that school choice does not "cream" the best students does not address whether public schools respond effectively to the challenge of school choice. Does the quality of education improve for those who remain in traditional public schools? Studying this issue is difficult. None of the current school choice pilot programs is large enough or has existed long enough to allow researchers to detect with certainty the effects of choice programs on public schools. It is true that the Milwaukee public school district, home to the largest and longest-running choice program, has dramatically increased the number of public school choice programs to retain students who might be drawn to the private school choice program. The Milwaukee public schools have also promised parents that their children will read at grade level by the third grade or they can receive individual tutoring. The promise is advertised on billboards and the sides of buses to make people want to "choose" the public schools. This attentivenes s to the needs of students in Milwaukee suggests that the school district has constructively responded to the challenge of school choice. But these reports from Milwaukee are little more than anecdotes and are not the kind of evidence that social scientists require.

The evidence from a new evaluation I conducted of the A-Plus choice and accountability program in Florida provides stronger systematic evidence that the prospect of vouchers inspires significant academic improvement in public schools. Under the A-Plus program, students in schools that had received two failing grades from the state would be offered vouchers to attend a private or a different public school. I compared the changes in test scores of Florida schools that had received a failing grade to those of other schools in the state and found that the test-score gains of schools facing the imminent prospect of vouchers were more than twice as large as the gains realized by the other schools. When Florida schools had to compete to retain their students under a choice system, they made substantial progress.

Some studies address the effects of school choice on public school improvement by examining whether areas with more choice tend to have higher student test scores than areas with less choice. Harvard economist Caroline Minter Hoxhy examines the effect of choice on the quality of public and private schools by using a very innovative research strategy. Hoxby takes advantage of the fact that some families currently exercise choice by moving to different school districts within a metropolitan area or by paying the tuition to send their children to private school. Some metropolitan areas have more choices available than others because some have more school districts and more private schools. For example, Boston has several school districts in the metropolitan area (Boston, Brookline, Cambridge, Waltham, etc.), while Miami has only one school district for the entire county.

Hoxby examines whether the availability of more choices is related to higher academic achievement. As one would expect, given most economic theory, the metropolitan areas in her study with more choices available have higher academic performance at lower cost than do metropolitan areas with fewer choices available. A one standard deviation increase in the available public school district choices results in a 3 percentile point improvement in test scores and a 4 percent increase in wages for students upon entering the work force, all for 17 percent less per capita expenditure. A one standard deviation increase in choices offered by the private sector results in an 8 percentile point improvement in test scores and a 12 percent increase in wages for students upon entering the work force, without any significant change in per capita expenditure. Hoxby concludes: "If private schools in any area receive sufficient resources to subsidize each student by $1,000, the achievement of public school students rises." Choic e appears to help the nonchoosers as well as the choosers.

A similar study that I conducted for the Manhattan Institute, called the Education Freedom Index, produced similar results. We measured the extent of educational choices currently available to families in each state, including charter school choices, subsidized private school choices, homeschooling choices, and public school choices. Controlling for per pupil spending, median household income, class size, and racial composition, states that offered more choices to families in the education of their children had significantly higher student test scores. When parents have more choices, schools pay greater attention to the needs of students because families may withdraw their children and the accompanying resources.

Clearly, we could have more direct and conclusive research than these three studies on whether school choice improves the quality of education for nonchoosers as well as choosers. For that kind of evidence, we would need to have a number of large-scale voucher programs that were in existence for several years. We could compare the overall educational achievement in districts with large-scale school choice programs to comparable districts that did not have voucher programs to see if vouchers helped spur schools to improve. The evidence at this point certainly suggests that such large-scale choice programs are worth trying. We have seen that the mechanism by which some worry that choice will undermine the quality of public schools, the "creaming" off of the best students, has not occurred in the several choice programs that have been studied. And it is important to note that existing school choices, particularly the ability of wealthier families to move to different school districts or attendance zones, produce a considerable amount of "creaming" before voucher programs are introduced. We have also seen that the studies of school choice without vouchers show that when it is easier for mo re families to exercise such choices, school quality is higher.

Promoting civic ideals

Even if some were convinced that school choice could improve academic achievement for choosers and nonchoosers alike, they might still be wary of vouchers' possible effect on education's civic purposes. Our system of government-operated schools was developed to ensure the transmission of desired civic values to future generations, as much as to impart economically useful skills. Oddly, however, while promoting democratic principles is a central mission of public schools, there is virtually no evidence to support the claim that government control of schools is necessary to achieve this goal. For many persons, even academics, the importance of government control of schools for promoting civic ideals is simply an article of faith.

Some recent studies, however, cast doubt on whether government management of schools is necessary or even desirable for promoting civic ideals. In one study, several colleagues and I analyzed responses from the Latino National Political Survey (LNPS), a national sample of adult Latinos. Subjects were asked whether they went to a public, private, or foreign school for each grade. They were also asked about their willingness to let members of their least-liked group engage in political activities such as running for office or holding demonstrations. Respondents in these kinds of studies most often identify the Ku Klux Klan as their least-liked group; Latinos most often picked gay activists. The more willing people are to allow members of their least-liked group to participate in these activities, the more tolerant they are said to be. Controlling for a variety of background characteristics, we found that adult Latinos who had been educated mostly in private school were more likely to be tolerant than those who had been educated mostly in public or foreign schools. The effect was moderate, but significant. Latinos who received their education entirely in private school were willing to tolerate the political activities of their least-liked group 50 percent of the time, compared to 39 percent for Latinos who never attended private school, holding all other factors constant.

In another study, headed by Patrick Wolf of the Brookings Institution, a sample of college students was asked similar questions about their willingness to allow members of their least-liked group to engage in certain political activities. Again, controlling statistically for differences in the students' backgrounds, the more time students spent in private school before college, the more tolerant they were.

Harvard University researcher David Campbell examined a large national data set of secondary school students that contained a limited set of tolerance items focusing on whether students would tolerate antireligious activities. Campbell found that Catholic school and secular private school students are more likely to be tolerant than are public school students. Secular, Catholic, and other religious private school students also outperformed their public school counterparts on other civic measures, including their experience with volunteering and their willingness to engage in public speaking or write letters on public issues.

Rather than being the bastions of intolerance they are sometimes imagined to be, private schools and religious schools appear to be more successful than public schools at instilling tolerance in their students. And remarkably, this private school advantage appears to last into the adult lives of their students.

Racial integration

But does this tolerance in private schools extend to racial integration? School choice has a bad reputation on this issue because vouchers were endorsed in the 1960s by some southern segregationists who wanted to evade court orders to integrate schools. Vouchers do have this shameful history, but government-controlled public schools have a shameful history of their own, having been segregated by law in much of the country for almost a century. The desirability of school choice with regard to racial integration should be judged by the policy's merits, not its pedigree.

In the last few years, a number of studies have examined the effect of school choice on racial integration. In one study, I examined the racial composition of a random sample of public and private school classrooms collected by the National Education Longitudinal Study (NELS). I found that private school students were significantly less likely to be in racially homogenous classrooms. Fifty-five percent of public school students were in classrooms that were almost entirely white or almost entirely minority in their racial composition, while 41 percent of private school students were similarly segregated. When all families choose their schools, as they do in the private sector, more of their children end up in racially mixed educational settings than when most families were assigned to schools, as they are in the public sector. Choice appears conducive to integration, while government assignment to public school appears to encourage segregation.

In another study, several colleagues and I observed a random sample of public and private school lunchrooms in Austin and San Antonio, Texas, and recorded where students sat by race. We found that private school students were significantly more likely to be in racially mixed groups at lunch than were public school students. After adjusting for seating restrictions, school size, and student grade level, we found that 79 percent of private school students were in racially mixed groups, compared to 43 percent of public school students. Sitting in a racially mixed group was defined as having any one of five adjacent students being of a different racial or ethnic group. We found that religious private schools were better integrated than were secular schools, suggesting that the low tuition typically found at religious schools helped contribute to racial integration. If vouchers or tax-credits further reduced the financial barriers to private school attendance, integration in private schools might be even better.

We also found that public schools with more students from outside their attendance zones--that is, with more magnet or transfer students--had higher rates of integration. It appears that choice systems, where schooling is detached from housing, are better able to transcend racial segregation in housing patterns. Traditional public schools, on the other hand, appear to replicate and perhaps reinforce racial segregation in housing.

Recent work by Stanford economist Thomas Nechyba arrives at similar conclusions about segregation by income. Based on policy simulations, Nechyba finds that "by removing education-related incentives for high-income households to separate themselves from poor neighborhoods, vouchers introduce a desegregating force into society. [And] by reducing housing prices in high quality public school districts and raising them in low quality districts, vouchers help more low-income families afford to live in areas with better public schools." In other words, by attaching schooling to housing, the public school system has created distortions in the racial mix and pricing of housing. Housing prices are artificially high in areas with desirable public schools and artificially low in areas with undesirable public schools, contributing to sorting of housing patterns by income (and race). By detaching schooling from housing, school choice makes it easier for wealthier families to stay in economically mixed neighborhoods. And by reducing the premium placed on housing in areas with good schools, vouchers make it easier for poorer families to move into those areas. It is no wonder that vouchers are most strongly supported by poor inner-city residents and most vigorously opposed by well-to-do suburbanites.

But these findings are based on examinations of existing private schools or policy simulations. What would the effects of an actual choice program be on integration? The Cleveland and Milwaukee school choice programs offer some answers. Following a strategy similar to that used to examine the data from NELS, I looked at whether choice students in Cleveland were more likely to attend schools that were racially representative of the broader community and less likely to attend racially homogeneous schools than were public school students. I found that nearly one-fifth of recipients of a voucher in Cleveland attend private schools that have a racial composition that resembles the average racial composition of the Cleveland area (defined as having a proportion of minority students in the school that is within 10 percent of the average proportion of minorities in metropolitan Cleveland). Only 5 percent of public school students in the Cleveland metropolitan area are in comparably integrated schools. More than thre e-fifths of public school students in metropolitan Cleveland attend schools that are almost entirely white or almost entirely minority in their racial composition. Half of the students in the Cleveland Scholarship Program are in comparably segregated schools. The amount of integration is not great in either system, but it is markedly better in the choice program.

When Howard Fuller and George Mitchell of Marquette University examined racial integration data from Milwaukee, their findings were similar to those from Cleveland. In 1998-99, they observed that 58 percent of Milwaukee public elementary students attended schools with more than 90 percent or fewer than 10 percent minority students. Only 38 percent of elementary school students at a large sample of Milwaukee Catholic schools were in similarly segregated schools. In 1998-99, Catholic schools accounted for more than half of the growth of choice students in the Milwaukee voucher program.

The public systems in Cleveland and Milwaukee, despite years of busing and other forced desegregation efforts, produce highly segregated schools. Desegregation has failed in those districts because white parents lack faith in the public schools' ability to manage integration successfully, and consequently have fled to the suburbs. The school choice programs in those cities allow families to transcend racial segregation in housing by selecting a racially mixed school in which they have confidence. And families are more likely to attend racially mixed schools by means of vouchers than through their ability to purchase housing in areas with desired schools.

Contrary to popular myth, private schools are neither bastions for intolerance nor segregation. In fact, private control of schools appears to promote the civic purposes of education7 more effectively than government control of schools.

A new consensus?

Reviewing the recent evidence on the effects of school choice leaves us with a few basic conclusions. First, all seven random-assignment studies and all three nonrandom-assignment studies found important benefits for the families that participate in choice programs. Second, choice does not appear to "cream" the best students. In all studies of existing choice programs, the evidence shows that participants have very low family incomes, predominantly come from single-mother households, and have a prior record of low academic performance. Third, the existing choice programs are not large enough nor have they operated long enough to address definitively the effects, positive or negative, on the public school system. However, the results from the A-Plus program in Florida suggest that the prospect of vouchers may induce public schools to improve. And Caroline Minter Hoxby's work finds that metropolitan areas with more choices available have significantly better outcomes at lower cost, and my work on the Education Freedom Index finds that states that offer more choices to parents enjoy higher student achievement. From these studies, we can conclude that choice is likely to improve public schools. Finally, private schools are more likely to be integrated and to promote civic virtues like tolerance than public schools.

Of course, considerably more research needs to be done on these questions before the conventional wisdom among the media and among academic and policy elites might be changed. This is especially the case since even gold-standard studies that reveal consistently positive benefits from school choice are likely to be interpreted in the least favorable light possible by choice opponents. But if the evidence we have so far is any indication, future studies will only confirm what we are already beginning to suspect: school choice works. Eventually, the weight of this evidence will make it hard to ignore or disparage.

An earlier version of this article was published as a Civic Report by the Manhattan Institute. An expanded version will appear in the book Charters, Vouchers, and Public Education, forthcoming from the Brookings Institution.

JAY P. GREENE is a senior fellow at the Manhattan Institute for Policy Research. He was a professor at the University of Texas and the University of Houston.

COPYRIGHT 2001 The National Affairs, Inc.
COPYRIGHT 2002 Gale Group