Standard Empirical Policy Analysis Does Not Support Universal Vouchers/ESAs

Abstract

At least 17 states have adopted school vouchers combined with “education savings accounts” (ESAs) that are nearly universal in student and school eligibility—just since 2020. These policies allow families to use government funds to pay for tuition at essentially any in-person private school, including religious schools, and to use the funds to pay for qualified educational expenses, such as tutors and computers. The funding, while still below that of neighborhood public schools (NPSs),1 is substantial, ranging from about $7000 to $10,000 per pupil annually.

The terms that the program's supporters use have created confusion about the program design. The term “savings account” is common and it is partially accurate in the sense that the funds are paid to providers from accounts created in students' and family names, but the first word also implies that families contribute from their own savings. Some of the state policies are also called “scholarships,” implying that they are based on academic merit. But these are neither savings accounts nor scholarships. Because these policies couple school vouchers with ESAs, it is more apt to think of them as super-vouchers. I will just call them “vouchers/ESAs” for purposes here.

The fact that vouchers/ESAs combine two types of policies complicates the analysis. As I explain, the voucher element allows schools to discriminate against certain students, entwines government with religion, involves a large fiscal cost, and has shown fairly poor, or at best inconclusive, academic results. Home and virtual schooling, funded by the ESA element, do not suffer much from the first three problems but seem to generate even worse academic outcomes. Despite their differences, this Point-Counterpoint combines the debates because vouchers/ESAs are currently being adopted as a package.

As in any policy debate, we must also consider the policy alternatives. Since the early 1990s, the status quo of American education, originally built on NPSs, has been transformed into a multifaceted system that already has many forms of school choice. Almost one third of American K–12 students no longer attended their NPSs before the recent voucher/ESA push (Harris et al. 2017; USDOE 2018). In addition to tuition-paying private schools, the modern American school system already includes almost 8000 charter schools,2 homeschooling3 supported by subsidized virtual charter schools,4 and open enrollment that allows students to attend public schools outside their attendance zones and school districts. Whether vouchers are wise relative to NPSs is still a legitimate debate, but today's debate is really about whether we should continue to expand public school choice or expand religious and other private school choice.

In assessing these policy alternatives, I adopt the well-known voucher evaluation framework by Levin (2002), which includes two criteria common to almost any empirical policy analysis: efficiency and equity. Levin's framework also includes two less common criteria that are relevant with vouchers. One is freedom of choice, or liberty, which is seen as good in and of itself. Levin's fourth criterion, social cohesion, means that schooling can be a kind of social glue that creates common language and values like tolerance, respect for individuals, and belief in democracy that are necessary for society to function.

Below, I show that the case for universal vouchers/ESAs is weak on these criteria. Even the freedom of choice criterion, undoubtedly the one that most strongly favors vouchers/ESAs, does not present as strong an argument as it seems.

These new vouchers/ESAs represent arguably the most fundamental change in U.S. K–12 education policy in 70 years. Specifically, they run against our three most significant and longstanding schooling traditions: secular teaching, public accountability, and nondiscrimination (Harris 2024).

First, the United States was founded in part on the ideas that people should be free to practice their religions, and that the government should not establish a state religion. While vouchers/ESAs now pass Constitutional muster (Wooding 2025), they still entwine church and state in practice because the government must decide which religious schools can participate in the voucher/ESA program. While this church-state entanglement is legally allowable, it is still a major break from tradition.

Second, every state constitution refers to “public” education as a state responsibility. While there remains debate about what state framers meant by this term, it has come to mean at least some level of public governance. For the last century, this has meant elected local boards governing NPSs. The extent of public governance has also increased and centralized over time, starting with state standards in the 1980s, then test-based accountability in the 1990s, and finally reaching its apex with the 2001 federal No Child Left Behind law. Vouchers/ESAs reject this public accountability tradition and, instead, rely on parents and private schools making decisions through markets.

Third, the nondiscrimination principle means that NPSs must, with limited exceptions, accept all students who live within a school's neighborhood, regardless of race, religion, sexual orientation, disability, political party, and academic ability. Vouchers/ESAs, in contrast, are used at private schools that can legally select or discriminate on any of the above factors, except for race (Petrilli 2017).

For these reasons, the new vouchers/ESAs are arguably the most significant departure from U.S. education tradition since Brown v. Board of Education sparked school desegregation in 1954. The Brown decision was important in part because it extended one of the prior principles—antidiscrimination—to African Americans. More broadly, before the recent voucher/ESA push, education policy had been evolving to reinforce the principles of secular teaching, public accountability, and nondiscrimination. Vouchers/ESAs reverse course and run against those same principles.

Vouchers/ESAs also diverge from global education policy. Many countries provide public funding to private schools. But even subsidized religious schools in these nations are subject to government standards and tests. In effect, the “voucher” policies in other countries are more like the U.S. charter school policies.

This new private school choice agenda is therefore a major departure from both U.S. and global traditions. The fact that we have so little experience with universal vouchers/ESAs raises the burden of proof on those who argue for them.

Advocates for any policy usually start with evidence of sustained and demonstrable failure of current policy on a given topic. When policies fail, we should try something new—and when they fail badly, we might experiment with something radically new and even unproven. In what follows, I describe the outcomes of the entire education system in the decades before vouchers/ESAs took hold.

Many education outcomes, especially high school graduation and college-going, are at or near all-time highs and high relative to other higher-income countries (Harris et al. 2025). Parent satisfaction with public schools has also been high and steady for decades and improving on some metrics. In the last Gallup (2025) poll before COVID, 82% of parents were completely or somewhat satisfied with the schools their children attended.5 Partly because of these academic and other successes, the U.S. workforce remains one of the most productive in the world (de Vries 2022; Harris et al. 2025).

Our test score results are more mixed. U.S. international standing has long been slightly above average among high-income countries, though we have been improving in relative terms since 1990 on the PISA (Harris et al. 2025).6 Also, our scores on the National Assessment of Educational Progress (NAEP) increased steadily for a quarter century through around 2014 but have declined somewhat since then, even before COVID (Harris et al. 2025). One of the likely causes of declining test scores is reduced test-based accountability a decade after NCLB (Dee and Jacob 2011; Goldhaber and DeArmond 2023).

Less well recognized is that today's high school students typically graduate with knowledge of much more advanced content than in prior decades. The percentage of high school students who have taken a “rigorous curriculum” more than doubled from 30% in 1990 to 62% in 2019 (USDOE 2019). This is apparently due to the ratcheting up of high school graduation requirements and competition to get into better colleges. The content of these advanced courses is not reflected in the NAEP or international assessments.

It is difficult to interpret this as evidence that a major and unproven change is necessary, especially these new vouchers/ESAs, which limit the very standards and accountability that have been partially responsible for our recent successes.

Even if recent trends do not seem to call for radical change, it is still theoretically possible that vouchers/ESAs would make noticeable academic improvements over the status quo.

In deciding what evidence to focus on, it is important to note that all forms of school choice seem to work better with disadvantaged students in urban areas and the vouchers/ESAs we are presently debating are universal (Chen and Harris 2023; Cohodes and Roy 2024). Many prior studies of vouchers also focus on small-scale voucher programs targeted to low-income students in urban areas, but only a small fraction of the recipients of these new universal programs fall into these groups.

For this reason, I focus attention on the more recent and relevant evidence from four U.S. statewide programs that are most similar to the voucher/ESAs being debated (Harris 2018). In three of the four states—Indiana, Louisiana, and Ohio—researchers have found convincing evidence of negative effects on student achievement (Erickson et al. 2021; Figlio and Karbownik 2016; Waddington and Berends 2018). Florida is the fourth state but cannot be studied because of there are no comparable test scores. In Louisiana, we also have evidence that the program showed no effect on college-going (Erickson et al. 2021); the Louisiana results are especially credible because this is the only state where students were randomly assigned to voucher access.

One possible reason the test score results might be so poor is that private school curricula are not usually aligned to public school tests. Naturally, students are not going to test well on material they have no seen yet. However, in one of the states with negative test score effects, Indiana, all private school students have long been required to take the public school test, so the test alignment issue, even if a possible partial explanation, cannot explain all of the negative results.

Other voucher advocates argue that test scores are simply a less important outcome (Kelly 2013). But if we accept that argument, it also undermines the case for vouchers in a different way. The decline in national NAEP scores, based on tests taken mostly by public school students, is the main weakness of the status quo (see above), but if test scores are not important in assessing schools, then it raises the question, why should we overhaul a system that is generating improvement on the other outcomes?

While no effects were found on college-going in Louisiana, the results for that outcome do look better in Florida and Ohio (Chingos et al. 2019; 2025), at least at first glance. The researchers were hampered by the fact that students in these two states were not randomly assigned to receive vouchers, forcing them to rely on matching techniques that do not work well when we cannot match on prior values of the outcome variable (Cook et al. 2008; Hallberg et al. 2018). Even if two students look the same on observable factors, the fact that they made different schooling choices means they are different in unobservable ways, which are likely related to their expected outcomes.

Voucher advocates also argue that the negative test score results in the above three states are due to higher-performing schools declining to participate in voucher programs because the regulations attached to the programs make participation unattractive (Sude et al. 2017). Unfortunately, this is difficult to establish because we cannot measure the performance of nonparticipating schools (or even participating ones in many states). Even if we could establish the quality of private schools, the most worrisome regulations, according to private school leaders, are restrictions on their admission criteria, which private schools report would reduce their probability of participation by 70% (DeAngelis et al. 2019). Whether to maintain their prestige through exclusivity or create homogenous groups aligned with their missions, this implies that high-quality private schools are unlikely to admit many new students even if the school participates.

These four statewide programs, while they are the closest comparison we have, also differ from today's universal voucher in one important way: They still target disadvantaged students who are most likely to benefit from school choice (Chen and Harris 2023; Cohodes and Roy 2024). So, even these poor and inconclusive academic results in the four statewide programs likely over-state what universal vouchers/ESAs will accomplish.

This lack of targeting also means the programs have some perverse fiscal effects. Because of the lack of means-testing, the vast majority of voucher/ESA funding is currently going to high-income families, mostly in the suburbs, who already have their children in private schools (Klinenberg et al. 2024). This is not just fiscally regressive but also inefficient since higher-income students benefit less educationally.

As much as the above discussion presents a weak academic case for universal vouchers, the situation is even worse for programs supporting home/virtual schooling. While less extensively and rigorously studied, what we do know suggests the home and virtual schooling that ESAs fund substantially reduce student achievement and long-term outcomes (Yoo et al. 2023). These studies, however, rely on matching methods that, as I said above, are unconvincing because of selection bias.

Vouchers/ESAs simply do not have the track record we would expect for a policy that is poised to upend the basic principles of American and international education. At the very least, it seems reasonable to ask that some share of the many states experimenting with these programs show positive and convincing results with similar universal eligibility, before scaling it up.

Since the debate is really between private/religious school choice and public school choice, we also need to look at how the latter choice policies work. Little research exists on the effects of open enrollment programs, but we know much more about charter schools. Like vouchers, charter schools are privately operated and students can attend almost regardless of where they live. Like NPSs, however, charter schools are required to follow state standards and administer state tests, to have outcome data publicly reported, and to have oversight from a government agent, called an authorizer.

The evidence is increasingly clear that charter schools raise average academic achievement for participating students (Epple et al. 2015; Cohodes and Roy 2024; Harris 2025). For example, in the most recent national study, the average participant effect was about +0.04 SD (CREDO 2023). We have less evidence on college-going, though some charter schools evidently have produced quite large improvements (Harris and Larsen 2025; Cohodes and Roy 2024), including in suburban locations (Cohodes and Pineda 2024).

As with vouchers, charter schools generate positive competitive effects on traditional public schools (Epple et al. 2015; Cohodes and Parham 2021). In fact, when looking across the whole country and across all the potential mechanisms, the net or market-level charter effects are positive for both achievement and high school graduation (Chen and Harris 2023). In other words, charter schools have demonstrated multiple types of success at scale. This is important because it means that charter schools, the main alternative to vouchers/ESAs, are more promising academically.

It can be difficult to predict the effects of policies in the long run. In economic terms, partial equilibrium responses can differ substantially from the general equilibrium. In this case, the substantial reduction in the price of private schooling, coupled with the advantage private schools have in their ability to select students and teach religion, is likely to lead to significant shifts in enrollment away from public schools over time. These long-run effects of large-scale change are difficult to predict, introducing an additional element of risk from universal voucher/ESA adoption.

Second, it is not just that charter schools are a more proven alternative but that, despite their successes, will likely be hurt by vouchers/ESAs. Charter schools (and NPSs) will be at a disadvantage because they cannot easily select students or teach religion. Vouchers/ESAs will also provide incentives for charter school operators to convert to fully private schools, sidestepping the public accountability they currently face. While this might be good for school leaders, the above analysis shows that the loss of public oversight will likely be detrimental for efficiency and equity.

A third long-run effect pertains to the fiscal costs. Replacing the previous private funding of private schools with public funding is already straining budgets in many states. To pay for this, states will either need to raise taxes or, more likely in the Republican-leaning, voucher-adopting states, cut per-pupil funding to education or cut other state programs, such as health care.

The above analysis shows that vouchers/ESAs are probably less efficient and less equitable than the relevant alternatives. The best evidence we have suggests that participating students will end up with somewhat lower achievement and no or unclear gains in college-going. The lack of targeting, combined with continued private school admission requirements, means that the spending will be regressive and resources will be devoted to those who benefit least. Vouchers/ESAs will also undermine charter schools, one of the few education policies of any kind that have generated success at a national scale, and which represent a form of public school choice that is the primary alternative to vouchers. Moreover, the very unusual nature of the reform and uncertainties about the long-run and indirect effects mean this is risky.

The third criterion, social cohesion, is closely related to discrimination. It is not hard to see how school selection of students would separate people into groups based on their beliefs—indeed, this is precisely what voucher advocates give as a reason for vouchers. Also, public accountability is about more than just test scores. It is about ensuring that schools instill common language, values like tolerance, and belief in democracy. Private schools, in contrast, will face market pressure to focus on private and parochial aims to the neglect of public ones. While the evidence on this is limited, there are reasons for concern that vouchers will make us less socially cohesive.

This leaves freedom of choice as the criterion where advocates are on the strongest ground. Vouchers/ESAs do indeed provide more freedom, though, even here, the case is not as strong as it seems. We do not need vouchers to expand families' choices, freedom, and liberty. This could be done within the public sector. Also, vouchers/ESAs allow schools to select students, which places some limits on parental choice—it is the “schools' choice” as much as parental “school choice” (Jennings 2010). These policies also are designed in ways that provide limited information to families when choosing schools. In these respects, the choice enabled by vouchers/ESAs may be more theoretical than practical. Vouchers/ESAs will increase the quantity of choice—more options will be available—but whether this is wise policy depends just as much on the quality and type of those options, whether parents have good information to choose among schools, and the way in which the entire system serves students and communities,

One potential reason for valuing parental freedom is that parents are better judges of student success and consider a wider range of outcomes, some difficult to measure. That is no doubt true in some respects, but education policy research almost universally relies on these standard academic measures because they are not only correlated with, but cause, increased employment, earnings, health, and a range of other private and social benefits (Wolfe and Haveman 2002; Jacob and Lefgren 2003; Hanushek and Woessman 2008; Machin et al. 2011; Lochner and Moretti 2004). There would be a price to be paid for any effort that undermines standard academic outcomes.

I have titled this “Standard Empirical Policy Analysis…” because the above approach reflects standard principles of policy analysis: I identified the usual program evaluation criteria, discussed evidence regarding the effects and consequences for associated outcomes, and compared the policy with the main alternative. On this basis, and especially for efficiency and equity, it seems difficult to justify scaling universal vouchers/ESAs, which upend long-held educational traditions and have generated poor, inconclusive, uncertain, and fiscally regressive results, especially as the alternative of public school choice has the same strength—freedom of choice—and has demonstrated measurable success at scale on both efficiency and equity grounds.