with whom Justice Stevens and Justice Souter join, dissenting.

I join Justice Souter’s opinion, and I agree substantially with Justice Stevens. I write separately, however, to emphasize the risk that publicly financed voucher programs pose in terms of religiously based social conflict. I do so because I believe that the Establishment Clause concern for protecting the Nation’s social fabric from religious conflict poses an overriding obstacle to the implementation of this well-intentioned school voucher program. And by explaining the nature of the concern, I hope to demonstrate why, in my view, “parental choice” cannot significantly alleviate the constitutional problem. See Part IV, infra.

I

The First Amendment begins with a prohibition, that “Congress shall make no law respecting an establishment of *718religion,” and a guarantee, that the government shall not prohibit “the free exercise thereof.” These Clauses embody an understanding, reached in the 17th century after decades of religious war, that liberty and social stability demand a religious tolerance that respects the religious views of all citizens, permits those citizens to “worship God in their own way,” and allows all families to “teach their children and to form their characters” as they wish. C. Radcliffe, The Law & Its Compass 71 (1960). The Clauses reflect the Framers’ vision of an American Nation free of the religious strife that had long plagued the nations of Europe. See, e. g., Freund, Public Aid to Parochial Schools, 82 Harv. L. Rev. 1680, 1692 (1969) (religious strife was “one of the principal evils that the first amendment sought to forestall”); B. Kosmin & S. Lachman, One Nation Under God: Religion in Contemporary American Society 24 (1993) (First Amendment designed in “part to prevent the religious wars of Europe from entering the United States”). Whatever the Framers might have thought about particular 18th-century school funding practices, they undeniably intended an interpretation of the Religion Clauses that would implement this basic First Amendment objective.

In part for this reason, the Court’s 20th-century Establishment Clause cases — both those limiting the practice of religion in public schools and those limiting the public funding of private religious education — focused directly upon social conflict, potentially created when government becomes involved in religious education. In Engel v. Vitale, 370 U. S. 421 (1962), the Court held that the Establishment Clause forbids prayer in public elementary and secondary schools. It did so in part because it recognized the “anguish, hardship and bitter strife that could come when zealous religious groups struggle] with one another to obtain the Government’s stamp of approval... .” Id., at 429. And it added:

“The history of governmentally established religion, both in England and in this country, showed that when*719ever government had allied itself with one particular form of religion, the inevitable result had been that it had incurred the hatred, disrespect and even contempt of those who held contrary beliefs.” Id., at 431.

See also Lee v. Weisman, 505 U. S. 577, 588 (1992) (striking down school-sanctioned prayer at high school graduation ceremony because “potential for divisiveness” has “particular relevance” in school environment); School Dist. of Abington Township v. Schempp, 374 U. S. 203, 307 (1963) (Goldberg, J., concurring) (Bible-reading program violated Establishment Clause in part because it gave rise “to those very divisive influences and inhibitions of freedom” that come with government efforts to impose religious influence on “young impressionable [school] children”).

In Lemon v. Kurtzman, 403 U. S. 602 (1971), the Court held that the Establishment Clause forbids state funding, through salary supplements, of religious school teachers. It did so in part because of the “threat” that this funding would create religious “divisiveness” that would harm “the normal political process.” Id., at 622. The Court explained:

“[Political debate and division . . . are normal and healthy manifestations of our democratic system of government, but political division along religious lines was one of the principal evils against which [the First Amendment’s religious clauses were] . . . intended to protect.” Ibid.

And in Committee for Public Ed. & Religious Liberty v. Nyquist, 413 U. S. 756, 794 (1973), the Court struck down a state statute that, much like voucher programs, provided aid for parents whose children attended religious schools, explaining that the “assistance of the sort here involved carries grave potential for . . . continuing political strife over aid to religion.”

When it decided these 20th-century Establishment Clause cases, the Court did not deny that an earlier American soci*720ety might have found a less clear-cut church/state separation compatible with social tranquility. Indeed, historians point out that during the early years of the Republic, American schools — including the first public schools — were Protestant in character. Their students recited Protestant prayers, read the King James version of the Bible, and learned Protestant religious ideals. See, e. g., D. Tyack, Onward Christian Soldiers: Religion in the American Common School, in History and Education 217-226 (P. Nash ed. 1970). Those practices may have wrongly discriminated against members of minority religions, but given the small number of such individuals, the teaching of Protestant religions in schools did not threaten serious social conflict. See Kosmin & Lach-man, supra, at 45 (Catholics constituted less than 2% of American church-affiliated population at time of founding).

The 20th-century Court was fully aware, however, that immigration and growth had changed American society dramatically since its early years. By 1850, 1.6 million Catholics lived in America, and by 1900 that number rose to 12 million. Jeffries & Ryan, A Political History of the Establishment Clause, 100 Mich. L. Rev. 279, 299-300 (Nov. 2001). There were similar percentage increases in the Jewish population. Kosmin & Lachman, supra, at 45-46. Not surprisingly, with this increase in numbers, members of non-Protestant religions, particularly Catholics, began to resist the Protestant domination of the public schools. Scholars report that by the mid-19th century religious conflict over matters such as Bible reading “grew intense,” as Catholics resisted and Protestants fought back to preserve their domination. Jeffries & Ryan, supra, at 300. “Dreading Catholic domination,” native Protestants “terrorized Catholics.” P. Hamburger, Separation of Church and State 219 (2002). In some States “Catholic students suffered beatings or expulsions for refusing to read from the Protestant Bible, and crowds . . . rioted over whether Catholic children could be *721released from the classroom during Bible reading.” Jeff-ries & Ryan, 100 Mich. L. Rev., at 300.

The 20th-century Court was also aware that political efforts to right the wrong of discrimination against religious minorities in primary education had failed; in fact they had exacerbated religious conflict. Catholics sought equal government support for the education of their children in the form of aid for private Catholic schools. But the “Protestant position” on this matter, scholars report, “was that public schools must be ‘nonsectarian’ (which was usually understood to allow Bible reading and other Protestant observances) and public money must not support ‘sectarian’ schools (which in practical terms meant Catholic).” Id., at 301. And this sentiment played a significant role in creating a movement that sought to amend several state constitutions (often successfully), and to amend the United States Constitution (unsuccessfully) to make certain that government would not help pay for “sectarian” (i e., Catholic) schooling for children. Id., at 301-305. See also Hamburger, supra, at 287.

These historical circumstances suggest that the Court, applying the Establishment Clause through the Fourteenth Amendment to 20th-century American society, faced an interpretive dilemma that was in part practical. The Court appreciated the religious diversity of contemporary American society. See Schempp, supra, at 240 (Brennan, J., concurring). It realized that the status quo favored some religions at the expense of others. And it understood the Establishment Clause to prohibit (among other things) any such favoritism. • Yet how did the Clause achieve that objective? Did it simply require the government to give each religion an equal chance to introduce religion into the primary schools — a kind of “equal opportunity” approach to the interpretation of the Establishment Clause? Or, did that Clause avoid government favoritism of some religions by insisting upon “separation” — that the government achieve *722equal treatment by removing itself from the business of providing religious education for children? This interpretive choice arose in respect both to religious activities in public schools and government aid to private education.

In both areas the Court concluded that the Establishment Clause required “separation,” in part because an “equal opportunity” approach was not workable. With respect to religious activities in the public schools, how could the Clause require public primary and secondary school teachers, when reading prayers or the Bible, only to treat all religions alike? In many places there were too many religions, too diverse a set of religious practices, too many whose spiritual beliefs denied the virtue of formal religious training. This diversity made it difficult, if not impossible, to devise meaningful forms of “equal treatment” by providing an “equal opportunity” for all to introduce their own religious practices into the public schools.

With respect to government aid to private education, did not history show that efforts to obtain equivalent funding for the private education of children whose parents did not hold popular religious beliefs only exacerbated religious strife? As Justice Rutledge recognized:

“Public money devoted to payment of religious costs, educational or other, brings the quest for more. It brings too the struggle of sect against sect for the larger share or for any. Here one [religious sect] by numbers [of adherents] alone will benefit most, there another. This is precisely the history of societies which have had an established religion and dissident groups.” Everson v. Board of Ed. of Ewing, 330 U. S. 1, 53-54 (1947) (dissenting opinion).

The upshot is the development of constitutional doctrine that reads the Establishment Clause as avoiding religious strife, not by providing every religion with an equal opportunity (say, to secure state binding or to pray in the public *723schools), but by drawing fairly clear lines of separation between church and state — at least where the heartland of religious belief, such as primary religious education, is at issue.

II

The principle underlying these cases — avoiding religiously based social conflict — remains of great concern. As religiously diverse as America had become when the Court decided its major 20th-century Establishment Clause cases, we are exponentially more diverse today. America boasts more than 55 different religious groups and subgroups with a significant number of members. Graduate Center of the City University of New York, B. Kosmin, E. Mayer, & A. Keysar, American Religious Identification Survey 12-13 (2001). Major religions include, among others, Protestants, Catholics, Jews, Muslims, Buddhists, Hindus, and Sikhs. Ibid. And several of these major religions contain different subsidiary sects with different religious beliefs. See Lester, Oh, Gods!, The Atlantic Monthly 37 (Feb. 2002). Newer Christian immigrant groups are “expressing their Christianity in languages, customs, and independent churches that are barely recognizable, and often controversial, for European-ancestry Catholics and Protestants.” H. Ebaugh & J. Chafetz, Religion and the New Immigrants: Continuities and Adaptations in Immigrant Congregations 4 (abridged student ed. 2002).

Under these modern-day circumstances, how is the “equal opportunity” principle to work — without risking the “struggle of sect against sect” against which Justice Rutledge warned? School voucher programs finance the religious education of the young. And, if widely adopted, they may well provide billions of dollars that will do so. Why will different religions not become concerned about, and seek to influence, the criteria used to channel this money to religious schools? Why will they not want to examine the implementation of the programs that provide this money — to determine, for ex*724ample, whether implementation has biased a program toward or against particular sects, or whether recipient religious schools are adequately fulfilling a program’s criteria? If so, just how is the State to resolve the resulting controversies without provoking legitimate fears of the kinds of religious favoritism that, in so religiously diverse a Nation, threaten social dissension?

Consider the voucher program here at issue. That program insists that the religious school accept students of all religions. Does that criterion treat fairly groups whose religion forbids them to do so? The program also insists that no participating school “advocate or foster unlawful behavior or teach hatred of any person or group on the basis of race, ethnicity, national origin, or religion.” Ohio Rev. Code Ann. § 3313.976(A)(6) (West Supp. 2002). And it requires the State to “revoke the registration of any school if, after a hearing, the superintendent determines that the school is in violation” of the program’s rules. § 3313.976(B). As one amicus argues, “it is difficult to imagine a more divisive activity” than the appointment of state officials as referees to determine whether a particular religious doctrine “teaches hatred or advocates lawlessness.” Brief for National Committee for Public Education and Religious Liberty as Ami-cus Curiae 23.

How are state officials to adjudicate claims that one religion or another is advocating, for example, civil disobedience in response to unjust laws, the use of illegal drugs in a religious ceremony, or resort to force to call attention to what it views as an immoral social practice? What kind of public hearing will there be in response to claims that one religion or another is continuing to teach a view of history that casts members of other religions in the worst possible light? How will the public react to government funding for schools that take controversial religious positions on topics that are of current popular interest — say, the conflict in the Middle East or the war on terrorism? Yet any major funding program *725for primary religious education will require criteria. And the selection of those criteria, as well as their application, inevitably pose problems that are divisive. Efforts to respond to these problems not only will seriously entangle church and state, see Lemon, 403 U. S., at 622, but also will promote division among religious groups, as one group or another fears (often legitimately) that it will receive unfair treatment at the hands of the government.

I recognize that other nations, for example Great Britain and France, have in the past reconciled religious school funding and religious freedom without creating serious' strife. Yet British and French societies are religiously more homogeneous — and it bears noting that recent waves of immigration have begun to create problems of social division there as well. See, e. g., The Muslims of France, 75 Foreign Affairs 78 (1996) (describing increased religious strife in France, as exemplified by expulsion of teenage girls from school for wearing traditional Muslim scarves); Ahmed, Extreme Prejudice; Muslims in Britain, The Times of London, May 2, 1992, p. 10 (describing religious strife in connection with increased Muslim immigration in Great Britain).

In a society as religiously diverse as ours, the Court has recognized that we must rely on the Religion Clauses of the First Amendment to protect against religious strife, particularly when what is at issue is an area as central to religious belief as the shaping, through primary education, of the next generation’s minds and spirits. See, e. g., .Webster, On the Education of Youth in America (1790), in Essays on Education in the Early Republic 43, 53, 59 (F. Rudolph ed. 1965) (“[Education of youth” is “of more consequence than making laws and preaching the gospel, because it lays the foundation on which both law and gospel rest for success”); Pope Paul VI, Declaration on Christian Education (1965) (“[T]he Catholic school can be such an aid to the fulfillment of the mission of the People of God and to the fostering of dialogue between *726the Church and mankind, to the benefit of both, it retains even in our present circumstances the utmost importance”).

III

I concede that the Establishment Clause currently permits States to channel various forms of assistance to religious schools, for example, transportation costs for students, computers, and secular texts. See Everson v. Board of Ed. of Ewing, 330 U. S. 1 (1947); Mitchell v. Helms, 580 U. S. 793 (2000). States now certify the nonsectarian educational content of religious school education. See, e. g., New Life Baptist Church Academy v. East Longmeadow, 885 F. 2d 940 (CA1 1989). Yet the consequence has not been great turmoil. But see, e. g., May, Charter School’s Religious Tone; Operation of South Bay Academy Raises Church-State Questions, San Francisco Chronicle, Dec. 17, 2001, p. Al (describing increased government supervision of charter schools after complaints that students were “studying Islam in class and praying with their teachers,” and Muslim educators complaining of “‘post-Sept. 11 anti-Muslim sentiment’”).

School voucher programs differ, however, in both kind and degree from aid programs upheld in the past. They differ in kind because they direct financing to a core function of the church: the teaching of religious truths to young children. For that reason the constitutional demand for “separation” is of particular constitutional concern. See, e. g., Weisman, 505 U. S., at 592 (“heightened concerns” in context of primary education); Edwards v. Aguillard, 482 U. S. 578, 583-584 (1987) (“Court has been particularly vigilant in monitoring compliance with the Establishment Clause in elementary and secondary schools”).

Private schools that participate in Ohio’s program, for example, recognize the importance of primary religious education, for they pronounce that their goals are to “communicate the gospel,” “provide opportunities to ... experience a faith community,” “provide ... for growth in prayer,” and “pro*727vide instruction in religious truths and values.” App. 408a, 487a. History suggests, not that such private school teaching of religion is undesirable, but that government funding of this kind of religious endeavor is far more contentious than providing funding for secular textbooks, computers, vocational training, or even funding for adults who wish to obtain a college education at a religious university. See supra, at 720-722. Contrary to Justice O’Connor’s opinion, ante, at 665-666 (concurring opinion), history also shows that government involvement in religious primary education is far more divisive than state property tax exemptions for religious institutions or tax deductions for charitable contributions, both of which come far closer to exemplifying the neutrality that distinguishes, for example, fire protection on the one hand from direct monetary assistance on the other. Federal aid to religiously based hospitals, ante, at 666 (O’Connor, J., concurring), is even further removed from education, which lies at the heartland of religious belief.

Vouchers also differ in degree. The aid programs recently upheld by the Court involved limited amounts of aid to religion. But the majority’s analysis here appears to permit a considerable shift of taxpayer dollars from public secular schools to private religious schools. That fact, combined with the use to which these dollars will be put, exacerbates the conflict problem. State aid that takes the form of peripheral secular items, with prohibitions against diversion of funds to religious teaching, holds significantly less potential for social division. In this respect as well, the secular aid upheld in Mitchell differs dramatically from the present ease. Although it was conceivable that minor amounts of money eould have, contrary to the statute, found their way to the religious activities of the recipients, see 530 U. S., at 864 (O’Connor, J., concurring in judgment), that case is at worst the camel’s nose, while the litigation before us is the camel itself.

*728 >

I do not believe that the “parental choice aspect of the voucher program sufficiently offsets the concerns I have mentioned. Parental choice cannot help the taxpayer who does not want to finance the religious education of children. It will not always help the parent who may see little real choice between inadequate nonsectarian public education and adequate education at a school whose religious teachings are contrary to his own. It will not satisfy religious minorities unable to participate because they are too few in number to support the creation of their own private schools. It will not satisfy groups whose religious beliefs preclude them from participating in a government-sponsored program, and who may well feel ignored as government funds primarily support the education of children in the doctrines of the dominant religions. And it does little to ameliorate the entanglement problems or the related problems of social division that Part II, supra, describes. Consequently, the fact that the parent may choose which school can cash the government’s voucher check does not alleviate the Establishment Clause concerns associated with voucher programs.

V

The Court, in effect, turns the clock back. It adopts, under the name of “neutrality,” an interpretation of the Establishment Clause that this Court rejected more than half a century ago. In its view, the parental choice that offers each religious group a kind of equal opportunity to secure government funding overcomes the Establishment Clause concern for social concord. An earlier Court found that “equal opportunity” principle insufficient; it read the Clause as insisting upon greater separation of church and state, at least in respect to primary education. See Nyquist, 413 U. S., at 783. In a society composed of many different religious creeds, I fear that this present departure from the Court’s earlier understanding risks creating a form of reli*729giously based conflict potentially harmful to the Nation’s social fabric. Because I believe the Establishment Clause was written in part to avoid this kind of conflict, and for reasons set forth by Justice Souter and Justice Stevens, I respectfully dissent.

delivered the opinion of the Court.

The State of Ohio has established a pilot program designed to provide educational choices to families with children who *644reside in the Cleveland City School District. The question presented is whether this program offends the Establishment Clause of the United States Constitution. We hold that it does not.

There are more than 75,000 children enrolled in the Cleveland City School District. The majority of these children are from low-income and minority families. Few of these families enjoy the means to send their children to any school other than an inner-city public school. For more than a generation, however, Cleveland’s public schools have been among the worst performing public schools in the Nation. In 1995, a Federal District Court declared a “crisis of magnitude” and placed the entire Cleveland school district under state control. See Reed v. Rhodes, No. 1:73 CV 1300 (ND Ohio, Mar. 3, 1995). Shortly thereafter, the state auditor found that Cleveland’s public schools were in the midst of a “crisis that is perhaps unprecedented in the history of American education.” Cleveland City School District Performance Audit 2-1 (Mar. 1996). The district had failed to meet any of the 18 state standards for minimal acceptable performance. Only 1 in 10 ninth graders could pass a basic proficiency examination, and students at all levels performed at a dismal rate compared with students in other Ohio public schools. More than two-thirds of high school students either dropped or failed out before graduation. Of those students who managed to reach their senior year, one of every four still failed to graduate. Of those students who did graduate, few could read, write, or compute at levels comparable to their counterparts in other cities.

It is against this backdrop that Ohio enacted, among other initiatives, its Pilot Project Scholarship Program, Ohio Rev. Code Ann. §§3313.974-3313.979 (Anderson 1999 and Supp. 2000) (program). The program provides financial assistance to families in any Ohio school district that is or has been “under federal court order requiring supervision and opera*645tional management of the district by the state superintendent.” § 3313.975(A). Cleveland is the only Ohio school district to fall within that category.

The program provides two basic kinds of assistance to parents of children in a covered district. First, the program provides tuition aid for students in kindergarten through third grade, expanding each year through eighth grade, to attend a participating public or private school of their parent’s choosing. §§ 3313.975(B) and (C)(1). Second, the program provides tutorial aid for students who choose to remain enrolled in public school. § 3313.975(A).

The tuition aid portion of the program is designed to provide educational choices to parents who reside in a covered district. Any private school, whether religious or nonreligious, may participate in the program and accept program students so long as the school is located within the boundaries of a covered district and meets statewide educational standards. § 313.976(A)(3). Participating private schools must agree not to discriminate on the basis of race, religion, or ethnic background, or to “advocate or foster unlawful behavior or teach hatred of any person or group on the basis of race, ethnicity, national origin, or religion.” § 3313.976(A)(6). Any public school located in a school district adjacent to the covered district may also participate in the program. § 3313.976(C). Adjacent public schools are eligible to receive a $2,250 tuition grant for each program student accepted in addition to the full amount of per-pupil state funding attributable to each additional student. §§ 3313.976(C), 3317.0B(I)(1).1 All participating schools, *646whether public or private, are required to accept students in accordance with rules and procedures established by the state superintendent. §§3313.977(A)(l)(a)-(c).

Tuition aid is distributed to parents according to financial need. Families with incomes below 200% of the poverty line are given priority and are eligible to receive 90% of private school tuition up to $2,250. §§ 3313.978(A) and (C)(1). For these lowest income families, participating private schools may not charge a parental copayment greater than $250. §3313.976(A)(8). For all other families, the program pays 75% of tuition costs, up to $1,875, with no copayment cap. §§ 3313.976(A)(8), 3313.978(A). These families receive tuition aid only if the number of available scholarships exceeds the number of low-income children who choose to participate.2 Where tuition aid is spent depends solely upon where parents who receive tuition aid choose to enroll their child. If parents choose a private school, checks are made payable to the parents who then endorse the checks over to the chosen school. § 3313.979.

The tutorial aid portion of the program provides tutorial assistance through grants to any student in a covered district who chooses to remain in public school. Parents arrange for registered tutors to provide assistance to their children and then submit bills for those services to the State for payment. §§ 3313.976(D), 3313.979(C). Students from low-income families receive 90% of the amount charged for such assistance up to $360. All other students receive 75% of that amount. § 3313.978(B). The number of tutorial assistance grants offered to students in a covered district must equal the number of tuition aid scholarships provided to stu*647dents enrolled at participating private or adjacent public schools. § 3313.975(A).

The program has been in operation within the Cleveland City School District since the 1996-1997 school year. In the 1999-2000 school year, 56 private schools participated in the program, 46 (or 82%) of which had a religious affiliation. None of the public schools in districts adjacent to Cleveland have elected to participate. More than 3,700 students participated in the scholarship program, most of whom (96%) enrolled in religiously affiliated schools. Sixty percent of these students were from families at or below the poverty line. In the 1998-1999 school year, approximately 1,400 Cleveland public school students received tutorial aid. This number was expected to double during the 1999-2000 school year.

The program is part of a broader undertaking by the State to enhance the educational options of Cleveland’s schoolchildren in response to the 1995 takeover. That undertaking includes programs governing community and magnet schools. Community schools are funded under state law but are run by their own school boards, not by local school districts. §§ 3314.01(B), 3314.04. These schools enjoy academic independence to hire their own teachers and to determine their own curriculum. They can have no religious affiliation and are required to accept students by lottery. During the 1999-2000 school year, there were 10 startup community schools in the Cleveland City School District with more than 1,900 students enrolled. For each child enrolled in a community school, the school receives state funding of $4,518, twice the funding a participating program school may receive.

Magnet schools are public schools operated by a local school board that emphasize a particular subject area, teaching method, or service to students. For each student enrolled in a magnet school, the school district receives $7,746, including state funding of $4,167, the same amount received *648per student enrolled at a traditional public school. As of 1999, parents in Cleveland were able to choose from among 23 magnet schools, which together enrolled more than 13,000 students in kindergarten through eighth grade. These schools provide specialized teaching methods, such as Montessori, or a particularized curriculum focus, such as foreign language, computers, or the arts.

In 1996, respondents, a group of Ohio taxpayers, challenged the Ohio program in state court on state and federal grounds. The Ohio Supreme Court rejected respondents’ federal claims, but held that the enactment of the program violated certain procedural requirements of the Ohio Constitution. Simmons-Harris v. Goff, 86 Ohio St. 3d 1, 8-9, 711 N. E. 2d 203, 211 (1999). The state legislature immediately cured this defect, leaving the basic provisions discussed above intact.

In July 1999, respondents filed this action in United States District Court, seeking to enjoin the reenacted program on the ground that it violated the Establishment Clause of the United States Constitution. In August 1999, the District Court issued a preliminary injunction barring further implementation of the program, 54 F. Supp. 2d 725 (ND Ohio), which we stayed pending review by the Court of Appeals, 528 U. S. 983 (1999). In December 1999, the District Court granted summary judgment for respondents. 72 F. Supp. 2d 834. In December 2000, a divided panel of the Court of Appeals affirmed the judgment of the District Court, finding that the program had the “primary effect” of advancing religion in violation of the Establishment Clause. 234 F. 3d 945 (CA6). The Court of Appeals stayed its mandate pending disposition in this Court. App. to Pet. for Cert, in No. 00-1779, p. 151. We granted certiorari, 533 U. S. 976 (2001), and now reverse the Court of Appeals.

The Establishment Clause of the First Amendment, applied to the States through the Fourteenth Amendment, prevents a State from enacting laws that have the “purpose” *649or “effect” of advancing or inhibiting religion. Agostini v. Felton, 521 U. S. 203, 222-223 (1997) (“[W]e continue to ask whether the government acted with the purpose of advancing or inhibiting religion [and] whether the aid has the ‘effect’ of advancing or inhibiting religion” (citations omitted)). There is no dispute that the program challenged here was enacted for the valid secular purpose of providing educational assistance to poor children in a demonstrably failing public school system. Thus, the question presented is whether the Ohio program nonetheless has the forbidden “effect” of advancing or inhibiting religion.

To answer that question, our decisions have drawn a consistent distinction between government programs that provide aid directly to religious schools, Mitchell v. Helms, 530 U. S. 793, 810-814 (2000) (plurality opinion); id., at 841-844 (O’Connor, J., concurring in judgment); Agostini, supra, at 225-227; Rosenberger v. Rector and Visitors of Univ. of Va., 515 U. S. 819, 842 (1995) (collecting cases), and programs of true private choice, in which government aid reaches religious schools only as a result of the genuine and independent choices of private individuals, Mueller v. Allen, 463 U. S. 388 (1983); Witters v. Washington Dept. of Servs. for Blind, 474 U. S. 481 (1986); Zobrest v. Catalina Foothills School Dist., 509 U. S. 1 (1993). While our jurisprudence with respect to the constitutionality of direct aid programs has “changed significantly” over the past two decades, Agostini, supra, at 236, our jurisprudence with respect to true private choice programs has remained consistent and unbroken. Three times we have confronted Establishment Clause challenges to neutral government programs that provide aid directly to a broad class of individuals, who, in turn, direct the aid to religious schools or institutions of their own choosing. Three times we have rejected such challenges.

In Mueller, we rejected an Establishment Clause challenge to a Minnesota program authorizing tax deductions for various educational expenses, including private school tu*650ition costs, even though the great majority of the program’s beneficiaries (96%) were parents of children in religious schools. We began by focusing on the class of beneficiaries, finding that because the class included “all parents,” including parents with “children [who] attend nonsectarian private schools or sectarian private schools,” 463 U. S., at 397 (emphasis in original), the program was “not readily subject to challenge under the Establishment Clause,” id., at 399 (citing Widmar v. Vincent, 454 U. S. 263, 274 (1981) (“The provision of benefits to so broad a spectrum of groups is an important index of secular effect”)). Then, viewing the program as a whole, we emphasized the principle of private choice, noting that public funds were made available to religious schools “only as a result of numerous, private choices of individual parents of school-age children.” 463 U. S., at 399-400. This, we said, ensured that “no ‘imprimatur of state approval’ can be deemed to have been conferred on any particular religion, or on religion generally.” Id., at 399 (quoting Widmar, supra, at 274)). We thus found it irrelevant to the constitutional inquiry that the vast majority of beneficiaries were parents of children in religious schools, saying:

“We would be loath to adopt a rule grounding the constitutionality of a facially neutral law on annual reports reciting the extent to which various classes of private citizens claimed benefits under the law.” 463 U. S., at 401.

That the program was one of true private choice, with no evidence that the State deliberately skewed incentives toward religious schools, was sufficient for the program to survive scrutiny under the Establishment Clause.

In Witters, we used identical reasoning to reject an Establishment Clause challenge to a vocational scholarship program that provided tuition aid to a student studying at a religious institution to become a pastor. Looking at the program as a whole, we observed that “[a]ny aid . . . that ulti*651mately flows to religious institutions does so only as a result of the genuinely independent and private choices of aid recipients.” 474 U. S., at 487. We further remarked that, as in Mueller, “[the] program is made available generally without regard to the sectarian-nonsectarian, or public-nonpublic nature of the institution benefited.” 474 U. S., at 487 (internal quotation marks omitted). In light of these factors, we held that the program was not inconsistent with the Establishment Clause. Id., at 488-489.

Five Members of the Court, in separate opinions, emphasized the general rule from Mueller that the amount of government aid channeled to religious institutions "by individual aid recipients was not relevant to the constitutional inquiry. 474 U. S., at 490-491 (Powell, J., joined by Burger, C. J., and Rehnquist, J., concurring) (citing Mueller, supra, at 398-399); 474 U. S., at 493 (O’Connor, J., concurring in part and concurring in judgment); id., at 490 (White, J., concurring). Our holding thus rested not on whether few or many recipients chose to expend government aid at a religious school but, rather, on whether recipients generally were empowered to direct the aid to schools or institutions of their own choosing.

Finally, in Zobrest, we applied Mueller and Witters to reject an Establishment Clause challenge to a federal program that permitted sign-language interpreters to assist deaf children enrolled in religious schools. Reviewing our earlier decisions, we stated that “government programs that neutrally provide benefits to a broad class of citizens defined without reference to religion are not readily subject to an Establishment Clause challenge.” 509 U. S., at 8. Looking once again to the challenged program as a whole, we observed that the program “distributes benefits neutrally to any child qualifying as ‘disabled.’ ” Id., at 10. Its “primary beneficiaries,” we said, were “disabled children, not sectarian schools.” Id., at 12.

*652We further observed that “[b]y according parents freedom to select a school of their choice, the statute ensures that a government-paid interpreter will be present in a sectarian school only as a result of the private decision of individual parents.” Id., at 10. Our focus again was on neutrality and the principle of private choice, not on the number of program beneficiaries attending religious schools. Id., at 10-11. See, e. g., Agostini, 521 U. S., at 229 (“Zobrest did not turn on the fact that James Zobrest had, at the time of litigation, been the only child using a publicly funded sign-language interpreter to attend a parochial school”). Because the program ensured that parents were the ones to select a religious school as the best learning environment for their handicapped child, the circuit between government and religion was broken, and the Establishment Clause was not implicated.

Mueller, Witters, and Zobrest thus make clear that where a government aid program is neutral with respect to religion, and provides assistance directly to a broad class of citizens who, in turn, direct government aid to religious schools wholly as a result of their own genuine and independent private choice, the program is not readily subject to challenge under the Establishment Clause. A program that shares these features permits government aid to reach religious institutions only by way of the deliberate choices of numerous individual recipients. The incidental advancement of a religious mission, or the perceived endorsement of a religious message, is reasonably attributable to the individual recipient, not to the government, whose role ends with the disbursement of benefits. As a plurality of this Court recently observed:

“[I]f numerous private choices, rather than the single choice of a government, determine the distribution of aid, pursuant to neutral eligibility criteria, then a government cannot, or at least cannot easily, grant special *653favors that might lead to a religious establishment.” Mitchell, 530 U. S., at 810.

See also id., at 843 (O’Connor, J., concurring in judgment) (“[W]hen government aid supports a school’s religious mission only because of independent decisions made by numerous individuals to guide their secular aid to that school, ‘no reasonable observer is likely to draw from the facts ... an inference that the State itself is endorsing a religious practice or belief’ ” (quoting Witters, 474 U. S., at 493 (O’Connor, J., concurring in part and concurring in judgment))). It is precisely for these reasons that we have never found a program of true private choice to offend the Establishment Clause.

We believe that the program challenged here is a program of true private choice, consistent with Mueller, Witters, and Zobrest, and thus constitutional. As was true in those cases, the Ohio program is neutral in all respects toward religion. It is part of a general and multifaceted undertaking by the State of Ohio to provide educational opportunities to the children of a failed school district. It confers educational assistance directly to a broad class of individuals defined without reference to religion, i. e., any parent of a school-age child who resides in the Cleveland City School District. The program permits the participation of all schools within the district, religious or nonreligious. Adjacent public schools also may participate and have a financial incentive to do so. Program benefits are available to participating families on neutral terms, with no reference to religion. The only preference stated anywhere in the program is a preference for low-income families, who receive greater assistance and are given priority for admission at participating schools.

There are no “financial incentive[s]” that “ske[w]” the program toward religious schools. Witters, supra, at 487-488. Such incentives “[are] not present... where the aid is allocated on the basis of neutral, secular criteria that neither favor nor disfavor religion, and is made available to both reli*654gious and secular beneficiaries on a nondiscriminatory basis.” Agostini, supra, at 231. The program here in fact creates financial disincentives for religious schools, with private schools receiving only half the government assistance given to community schools and one-third the assistance given to magnet schools. Adjacent public schools, should any choose to accept program students, are also eligible to receive two to three times the state funding of a private religious school. Families too have a financial disincentive to choose a private religious school over other schools. Parents that choose to participate in the scholarship program and then to enroll their children in a private school (religious or nonreligious) must copay a portion of the school’s tuition. Families that choose a community school, magnet school, or traditional public school pay nothing. Although such features of the program are not necessary to its constitutionality, they clearly dispel the claim that the program “creates .. . financial incentive[s] for parents to choose a sectarian school.” Zobrest, 509 U. S., at 10.3

Respondents suggest that even without a financial incentive for parents to choose a religious school, the program creates a “public perception that the State is endorsing religious practices and beliefs.” Brief for Respondents Simmons-Harris et al. 37-38. But we have repeatedly rec*655ognized that no reasonable observer would think a neutral program of private choice, where state aid reaches religious schools solely as a result of the numerous independent decisions of private individuals, carries with it the imprimatur of government endorsement. Mueller, 463 U. S., at 399; Witters, supra, at 488-489; Zobrest, supra, at 10-11; e.g., Mitchell, supra, at 842-843 (O’Connor, J., concurring in judgment) (“In terms of public perception, a government program of direct aid to religious schools... differs meaningfully from the government distributing aid directly to individual students who, in turn, decide to use the aid at the same religious schools”). The argument is particularly misplaced here since “the reasonable observer in the endorsement inquiry must be deemed aware” of the “history and context” underlying a challenged program. Good News Club v. Milford Central School, 533 U. S. 98, 119 (2001) (internal quotation marks omitted). See also Capitol Square Review and Advisory Bd. v. Pinette, 515 U. S. 753, 780 (1995) (O’Connor, J., concurring in part and concurring in judgment). Any objective observer familiar with the full history and context of the Ohio program would reasonably view it as one aspect of a broader undertaking to assist poor children in failed schools, not as an endorsement of religious schooling in general.

There also is no evidence that the program fails to provide genuine opportunities for Cleveland parents to select secular educational options for their school-age children. Cleveland schoolchildren enjoy a range of educational choices: They may remain in public school as before, remain in public school with publicly funded tutoring aid, obtain a scholarship and choose a religious school, obtain a scholarship and choose a nonreligious private school, enroll in a community school, or enroll in a magnet school. That 46 of the 56 private schools now participating in the program are religious schools does not condemn it as a violation of the Establishment Clause. The Establishment Clause question is whether Ohio is coerc*656ing parents into sending their children to religious schools, and that question must be answered by evaluating all options Ohio provides Cleveland schoolchildren, only one of which is to obtain a program scholarship and then choose a religious school.

Justice Souter speculates that because more private religious schools currently participate in the program, the program itself must somehow discourage the participation of private nonreligious schools. Post, at 70S-705 (dissenting opinion).4 But Cleveland’s preponderance of religiously af*657filiated private schools certainly did not arise as a result of the program; it is a phenomenon common to many American cities. See U. S. Dept, of Ed., National Center for Education Statistics, Private School Universe Survey: 1999-2000, pp. 2-4 (NCES 2001-330, 2001) (hereinafter Private School Universe Survey) (cited in Brief for United States as Amicus Curiae 24). Indeed, by all accounts the program has captured a remarkable cross-section of private schools, religious and nonreligious. It is true that 82% of Cleveland’s participating private schools are religious schools, but it is also true that 81% of private schools in Ohio are religious • schools. See Brief for State of Florida et al. as Amici Curiae 16 (citing Private School Universe Survey). To attribute constitutional significance to this figure, moreover, would lead to the absurd result that a neutral school-choice program might be permissible in some parts of Ohio, such as Columbus, where a lower percentage of private schools are religious schools, see Ohio Educational Directory (Lodging of Respondents Gatton et al., available in Clerk of Court’s case file), and Reply Brief for Petitioners in No. 00-1751, p. 12, n. 1, but not in inner-city Cleveland, where Ohio has deemed such programs most sorely needed, but where the preponderance of religious schools happens to be greater. Cf. Brief for State of Florida et al. as Amici Curiae 17 (“[T]he percentages of sectarian to nonsectarian private schools within Florida’s 67 school districts . . . vary from zero to 100 percent”). Likewise, an identical private choice program might be constitutional in some States, such as Maine or Utah, where less *658than 45% of private schools are religious schools, but not in other States, such as Nebraska or Kansas, where over 90% of private schools are religious schools. Id., at 15-16 (citing Private School Universe Survey).

Respondents and Justice Souter claim that even if we do not focus on the number of participating schools that are religious schools, we should attach constitutional significance to the fact that 96% of scholarship recipients have enrolled in religious schools. They claim that this alone proves parents lack genuine choice, even if no parent has ever said so. We need not consider this argument in detail, since it was flatly rejected in Mueller, where we found it irrelevant that 96% of parents taking deductions for tuition expenses paid tuition at religious schools. Indeed, we have recently found it irrelevant even to the constitutionality of a direct aid program that a vast majority of program benefits went to religious schools. See Agostini, 521 U. S., at 229 (“Nor are we willing to conclude that the constitutionality of an aid program depends on the number of sectarian school students who happen to receive the otherwise neutral aid” (citing Mueller, 463 U. S., at 401)); see also Mitchell, 530 U. S., at 812, n. 6 (plurality opinion) (“[Agostini] held that the proportion of aid benefiting students at religious schools pursuant to a neutral program involving private choices was irrelevant to the constitutional inquiry”); id., at 848 (O’Connor, J., concurring in judgment) (same) (quoting Agostini, supra, at 229). The constitutionality of a neutral educational aid program simply does not turn on whether and why, in a particular area, at a particular time, most private schools are run by religious organizations, or most recipients choose to use the aid at a religious school. As we said in Mueller, “[s]uch an approach would scarcely provide the certainty that this field stands in need of, nor can we perceive principled standards by which such statistical evidence might be evaluated.” 463 U. S., at 401.

*659This point is aptly illustrated here. The 96% figure upon which respondents and Justice Souter rely discounts entirely (1) the more than 1,900 Cleveland children enrolled in alternative community schools, (2) the more than 13,000 children enrolled in alternative magnet schools, and (3) the more than 1,400 children enrolled in traditional public schools with tutorial assistance. See supra, at 647-648. Including some or all of these children in the denominator of children enrolled in nontraditional schools during the 1999-2000 school year drops the percentage enrolled in religious schools from 96% to under 20%. See also J. Greene, The Racial, Economic, and Religious Context of Parental Choice in Cleveland 11, Table 4 (Oct. 8,1999), App. 217a (reporting that only 16.5% of nontraditional schoolchildren in Cleveland choose religious schools). The 96% figure also represents but a snapshot of one particular school year. In the 1997-1998 school year, by contrast, only 78% of scholarship recipients attended religious schools. See App. to Pet. for Cert, in No. 00-1751, p. 5a. The difference was attributable to two private nonreligious, schools that had accepted 15% of all scholarship students electing instead to register as community schools, in light of larger per-pupil funding for community schools and the uncertain future of the scholarship program generated by this litigation. See App. 59a-62a, 209a, 223a-227a.5 Many of the students enrolled in these schools *660as scholarship students remained enrolled as community school students, id., at 145a-146a, thus demonstrating the arbitrariness of counting one type of school but not the other to assess primary effect, e. g., Ohio Rev. Code Ann. §3314.11 (Anderson 1999) (establishing a single “office of school options” to “provide services that facilitate the management of the community schools program and the pilot project scholarship program”). In spite of repeated questioning from the Court at oral argument, respondents offered no convincing justification for their approach, which relies entirely on such arbitrary classifications. Tr. of Oral Arg. 52-60.6

*661Respondents finally claim that we should look to Committee for Public Ed. & Religious Liberty v. Nyquist, 413 U. S. 756 (1973), to decide these cases. We disagree for two reasons. First, the program in Nyquist was quite different from the program challenged here. Nyquist involved a New York program that gave a package of benefits exclusively to private schools and the parents of private school enroll-ees. Although the program was enacted for ostensibly secular purposes, id., at 773-774, we found that its “function” was “unmistakably to provide desired financial support for nonpublic, sectarian institutions,” id., at 783 (emphasis added). Its genesis, we said, was that private religious schools faced “increasingly grave fiscal problems.” Id., at 795. The program thus provided direct money grants to religious schools. Id., at 762-764. It provided tax benefits “unrelated to the amount of money actually expended by any parent on tuition,” ensuring a windfall, to parents of children in religious schools. Id., at 790. It similarly provided tuition reimbursements designed explicitly to “offe[r] ... an incentive to parents to send their children to sectarian schools.” Id., at 786. Indeed, the program flatly prohibited the participation of any public school, or parent of any public school enrollee. Id., at 763-765. Ohio’s program shares none of these features.

Second, were there any doubt that the program challenged in Nyquist is far removed from the program challenged here, we expressly reserved judgment with respect to “a case involving some form of public assistance ie.g., scholarships) made available generally without regard to the sectarian-nonsectarian, or public-nonpublic nature of the institution benefited.” Id., at 782-783, n. 38. That, of course, is the very question now before us, and it has since been answered, first in Mueller, 463 U. S., at 398-399 (“[A] program ... that neutrally provides state assistance to a broad spectrum of citizens is not readily subject to challenge under the Establishment Clause” (citing Nyquist, supra, at 782-783, n. 38)), *662then in Witters, 474 U. S., at 487 (“Washington’s program is ‘made available generally without regard to the sectarian-nonsectarian, or public-nonpublic nature of the institution benefited’ ” (quoting Nyquist, supra, at 782-783, n. 38)), and again in Zobrest, 509 U. S., at 12-13 (“[T]he function of the [program] is hardly ‘to provide desired financial support for nonpublic, sectarian institutions’” (quoting Nyquist, supra, at 782-783, n. 38)). To the extent the scope of Nyquist has remained an open question in light of these later decisions, we now hold that Nyquist does not govern neutral educational assistance programs that, like the program here, offer aid directly to a broad class of individual recipients defined without regard to religion.7

In sum, the Ohio program is entirely neutral with respect to religion. It provides benefits directly to a wide spectrum of individuals, defined only by financial need and residence in a particular school district. It permits such individuals to exercise genuine choice among options public and private, secular and religious. The program is therefore a program of true private choice. In keeping with an unbroken line of *663decisions rejecting challenges to similar programs, we hold that the program does not offend the Establishment Clause.

The judgment of the Court of Appeals is reversed.

It is so ordered.

concurring.

The Court holds that Ohio’s Pilot Project Scholarship Program, Ohio Rev. Code Ann. §§3313.974-3313.979 (Anderson 1999 and Supp. 2000) (voucher program), survives respondents’ Establishment Clause challenge. While I join the Court’s opinion, I write separately for two reasons. First, although the Court takes an important step, I do not believe that today’s decision, when considered in light of other longstanding government programs that impact religious organizations and our prior Establishment Clause jurisprudence, marks a dramatic break from the past. Second, given the emphasis the Court places on verifying that parents of voucher students in religious schools have exercised “true private choice,” I think it is worth elaborating on the Court’s conclusion that this inquiry should consider all reasonable educational alternatives to religious schools that are available to parents. To do otherwise is to ignore how the educational system in Cleveland actually functions.

I

These cases are different from prior indirect aid cases in part because a significant portion of the funds appropriated for the voucher program reach religious schools without restrictions on the use of these funds. The share of public resources that reach religious schools is not, however, as significant as respondents suggest. See, e.g., Brief for Respondents Simmons-Harris et al. 1-2. Data from the 1999-2000 school year indicate that 82 percent of schools participating in the voucher program were religious and that 96 percent of participating students enrolled in religious *664schools, see App. in Nos. 00-3055, etc. (CA6), p. 1679 (46 of 56 private schools in the program are religiously affiliated; 3,637 of 3,765 voucher students attend religious private schools), but these data are incomplete. These statistics do not take into account all of the reasonable educational choices that may be available to students in Cleveland public schools. When one considers the option to attend community schools, the percentage of students enrolled in religious schools falls to 62.1 percent. If magnet schools are included in the mix, this percentage falls to 16.5 percent. See J. Greene, The Racial, Economic, and Religious Context of Parental Choice in Cleveland 11, Table 4 (Oct. 8, 1999), App. 217a (reporting 2,087 students in community schools and 16,184 students in magnet schools).

Even these numbers do not paint a complete picture. The Cleveland program provides voucher applicants from low-income families with up to $2,250 in tuition assistance and provides the remaining applicants with up to $1,875 in tuition assistance. §§ 3313.976(A)(8), 3313.978(A) and (C)(1). In contrast, the State provides community schools $4,518 per pupil and magnet schools, on average, $7,097 per pupil. Affidavit of Caroline M. Hoxby ¶¶4^ 4c, App. 56a. Even if one assumes that all voucher students came from low-income families and that each voucher student used up the entire $2,250 voucher, at most $8.2 million of public funds flowed to religious schools under the voucher program in 1999-2000. Although just over one-half as many students attended community schools as religious private schools on the state fisc, the State spent over $1 million more — $9.4 million — on students in community schools than on students in religious private schools because per-pupil aid to community schools is more than double the per-pupil aid to private schools under the voucher program. Moreover, the amount spent on religious private schools is minor compared to the $114.8 million the State spent on students in the Cleveland magnet schools.

*665Although $8.2 million is no small sum, it pales in comparison to the amount of funds that federal,- state, and local governments already provide religious institutions. Religious organizations may qualify for exemptions from the federal corporate income tax, see 26 U. S. C. § 501(c)(3); the corporate income tax in many States, see, e. g., Cal. Rev. & Tax. Code Ann. §23701d (West 1992); and property taxes in all 50 States, see Turner, Property Tax Exemptions for Nonprofits, 12 Probate & Property 25 (Sept./Oct. 1998); and clergy qualify for a federal tax break on income used for housing expenses, 26 U. S. C. § 1402(a)(8). In addition, the Federal Government provides individuals, corporations, trusts, and estates a tax deduction for charitable contributions to qualified religious groups. See §§ 170, 642(c). Finally, the Federal Government and certain state governments provide tax credits for educational expenses, many of which are spent on education at religious schools. See, e.g., §25A (Hope tax credit); Minn. Stat. § 290.0674 (Supp. 2001).

Most of these tax policies are well established, see, e. g., Mueller v. Allen, 463 U. S. 388 (1983) (upholding Minnesota tax deduction for educational expenses); Walz v. Tax Comm’n of City of New York, 397 U. S. 664 (1970) (upholding an exemption for religious organizations from New York property tax), yet confer a significant relative benefit on religious institutions. The state property tax exemptions for religious institutions alone amount to very large sums annually. For example, available data suggest that Colorado’s exemption lowers that State’s tax revenues by more than $40 million annually, see Rabey, Exemptions a Matter of Faith: No Proof Required of Tax-Free Churches, Colorado Springs Gazette Telegraph, Oct. 26, 1992, p. Bl; Colorado Debates Church, Nonprofit Tax-Exempt Status, Philadelphia Enquirer, Oct. 4, 1996, p. 8; Maryland’s exemption lowers revenues by more than $60 million, see Maryland Dept, of Assessment and Taxation, 2001 SDAT Annual Report (Apr. 25, 2002), http://www.dat.state.md.us/sdatweb/stats/ *66601ar_rpt.html (Internet sources available in Clerk of Court’s case file); Wisconsin’s exemption lowers revenues by approximately $122 million, see Wisconsin Dept, of Revenue, Division of Research and Analysis, Summary of Tax Exemption Devices 2001, Property Tax (Apr. 25, 2002), http://www.dor. state.wi.us/ra/sumOOpro.html ($5,688 billion in exempt religious property; statewide average property tax rate of $21.46 per $1,000 of property); and Louisiana’s exemption, looking just at the city of New Orleans, lowers revenues by over $36 million, see Bureau of Governmental Research, Property Tax Exemptions and Assessment Administration in Orleans Parish: Summary and Recommendations 2 (Dec. 1999) ($22.6 million for houses of worship and $14.1 million for religious schools). As for the Federal Government, the tax deduction for charitable contributions reduces federal tax revenues by nearly $25 billion annually, see U. S. Dept, of Commerce, Bureau of Census, Statistical Abstract of the United States 344 (2000) (hereinafter Statistical Abstract), and it is reported that over 60 percent of household charitable contributions go to religious charities, id., at 397. Even the relatively minor exemptions lower federal tax receipts by substantial amounts. The parsonage exemption, for example, lowers revenues by around $500 million. See Diaz, Ramstad Prepares Bill to Retain Tax Break for Clergy’s Housing, Star Tribune (Minneapolis-St. Paul), Mar. 30, 2002, p. 4A.

These tax exemptions, which have “much the same effect as [cash grants] ... of the amount of tax [avoided],” Regan v. Taxation With Representation of Wash., 461 U. S. 540, 544 (1983); see also Rosenberger v. Rector and Visitors of Univ. of Va., 515 U. S. 819, 859-860, esp. n. 4 (1995) (THOMAS, J., concurring), are just part of the picture. Federal dollars also reach religiously affiliated organizations through public health programs such as Medicare, 42 U. S. C. §§ 1395-1395ggg, and Medicaid, §1396 et seq., through educational programs such as the Pell Grant program, 20 U. S. C. § 1070a, and the G. I. Bill of Rights, 38 U. S. C. §§3451, 3698; and *667through childcare programs such as the Child Care and Development Block Grant Program (CCDBG), 42 U. S. C. § 9858 (1994 ed., Supp. V). Medicare and Medicaid provide federal funds to pay for the healthcare of the elderly and the poor, respectively, see 1 B. Furrow, T. Greaney, S. Johnson, T. Jost, & R. Schwartz, Health Law 545-546 (2d ed. 2000); 2 id., at 2; the Pell Grant program and the G. I. Bill subsidize higher education of low-income individuals and veterans, respectively, see Mulleneaux, The Failure to Provide Adequate Higher Education Tax Incentives for Lower-Income Individuals, 14 Akron Tax J. 27, 31 (1999); and the CCDBG program finances child care for low-income parents, see Pitegoff, Child Care Policy and the Welfare Reform Act, 6 J. Affordable Housing & Community Dev. L. 113, 121-122 (1997). These programs are well-established parts of our social welfare system, see, e. g., Committee for Public Ed. & Religious Liberty v. Nyquist, 413 U. S. 756, 782-783, n. 38 (1973), and can be quite substantial, see Statistical Abstract 92 (Table 120) ($211.4 billion spent on Medicare and nearly $176.9 billion on Medicaid in 1998), id., at 135 (Table 208) ($9.1 billion in financial aid provided by the Department of Education and $280.5 million by the Department of Defense in 1999); Bush On Welfare: Tougher Work Rules, More State Control, Congress Daily, Feb. 26, 2002, p. 8 ($4.8 billion for the CCDBG program in 2001).

A significant portion of the funds appropriated for these programs reach religiously affiliated institutions, typically without restrictions on its subsequent use. For example, it has been reported that religious hospitals, which account for 18 percent of all hospital beds nationwide, rely on Medicare funds for 36 percent of their revenue. Merger-Watch, New Study Details Public Funding of Religious Hospitals (Jan. 2002), http://www.mergerwatch.org/inthenews/ publicfunding.html. Moreover, taking into account both Medicare and Medicaid, religious hospitals received nearly $45 billion from the federal fisc in 1998. Ibid. Federal aid *668to religious schools is also substantial. Although data for all States are not available, data from Minnesota, for example, suggest that a substantial share of Pell Grant and other federal funds for college tuition reach religious schools. Roughly one-third or $27.1 million of the federal tuition dollars spent on students at schools in Minnesota were used at private 4-year colleges. Minnesota Higher Education Services Office, Financial Aid Awarded, Fiscal Year 1999: Grants, Loans, and Student Earning from Institution Jobs (Jan. 24, 2001). The vast majority of these funds — $23.5 million— flowed to religiously affiliated institutions. Ibid.

Against this background, the support that the Cleveland voucher program provides religious institutions is neither substantial nor atypical of existing government programs. While this observation is not intended to justify the Cleveland voucher program under the Establishment Clause, see post, at 709-710, n. 19 (Souter, J., dissenting), it places in broader perspective alarmist claims about implications of the Cleveland program and the Court’s decision in these cases. See post, at 685-686 (Stevens, J., dissenting); post, at 715-716 (Souter, J., dissenting); post, p. 717 (Breyer, J., dissenting).

II

Nor does today’s decision signal a major departure from this Court’s prior Establishment Clause jurisprudence. A central tool in our analysis of cases in this area has been the Lemon test. As originally formulated, a statute passed this test only if it had “a secular legislative purpose,” if its “principal or primary effect” was one that “neither advance[d] nor inhibited] religion,” and if it did “not foster an excessive government entanglement with religion.” Lemon v. Kurtzman, 403 U. S. 602, 612-613 (1971) (internal quotation marks omitted). In Agostini v. Felton, 521 U. S. 203, 218, 232-233 (1997), we folded the entanglement inquiry into the primary effect inquiry. This made sense because both inquiries rely on the same evidence, see ibid., and the degree of entangle*669ment has implications for whether a statute advances or inhibits religion, see Lynch v. Donnelly, 465 U. S. 668, 688 (1984) (O’Connor, J., concurring). The test today is basically the same as that set forth in School Dish of Abington Township v. Schempp, 374 U. S. 203, 222 (1963) (citing Everson v. Board of Ed. of Ewing, 330 U. S. 1 (1947); McGowan v. Maryland, 366 U. S. 420, 442 (1961)), over 40 years ago.

The Court’s opinion in these cases focuses on a narrow question related to the Lemon test: how to apply the primary effects prong in indirect aid cases? Specifically, it clarifies the basic inquiry when trying to determine whether a program that distributes aid to beneficiaries, rather than directly to service providers, has the primary effect of advancing or inhibiting religion, Lemon v. Kurtzman, supra, at 613-614, or, as I have put it, of “endors[ing] or disapproving] . . . religion,” Lynch v. Donnelly, supra, at 691-692 (concurring opinion); see also Wallace v. Jaffree, 472 U. S. 38, 69-70 (1985) (O’Connor, J., concurring in judgment). See also ante, at 652. Courts are instructed to consider two factors: first, whether the program administers aid in a neutral fashion, without differentiation based on the religious status of beneficiaries or providers of services; second, and more importantly, whether beneficiaries of indirect aid have a genuine choice among religious and nonreligious organizations when determining the organization to which they will direct that aid. If the answer to either query is “no,” the program should be struck down under the Establishment Clause. See ante, at 652-653.

Justice Souter portrays this inquiry as a departure from Everson. See post, at 687-688 (dissenting opinion). A fair reading of the holding in that case suggests quite the opposite. Justice Black’s opinion for the Court held that the “[First] Amendment requires the state to be a neutral in its relations with groups of religious believers and nonbelievers; it does not require the state to be their adversary.” Everson, supra, at 18; see also Schempp, supra, at 218, 222. *670How else could the Court have upheld a state program to provide students transportation to public and religious schools alike? What the Court clarifies in these cases is that the Establishment Clause also requires that state aid flowing to religious organizations through the hands of beneficiaries must do so only at the direction of those beneficiaries. Such a refinement of the Lemon test surely does not betray Everson.

Ill

There is little question in my mind that the Cleveland voucher program is neutral as between religious schools and nonreligious schools. See ante, at 653-654. Justice Sou-ter rejects the Court’s notion of neutrality, proposing that the neutrality of a program should be gauged not by the opportunities it presents but rather by its effects. In particular, a “neutrality test... [should] focus on a category of aid that may be directed to religious as well as secular schools, and ask whether the scheme favors a religious direction.” Post, at 697 (dissenting opinion). JUSTICE SOUTER doubts that the Cleveland program is neutral under this view. He surmises that the cap on tuition that voucher schools may charge low-income students encourages these students to attend religious rather than nonreligious private voucher schools. See post, at 704-705. But Justice Souter’s notion of neutrality is inconsistent with that in our case law. As we put it in Agostini, government aid must be “made available to both religious and secular beneficiaries on a nondiscriminatory basis.” 521 U. S., at 231.

I do not agree that the nonreligious schools have failed to provide Cleveland parents reasonable alternatives to religious schools in the voucher program. For nonreligious schools to qualify as genuine options for parents, they need not be superior to religious schools in every respect. They need only be adequate substitutes for religious schools in the eyes of parents. The District Court record demonstrates that nonreligious schools were able to compete effectively *671with Catholic and other religious schools in the Cleveland voucher program. See ante, at 656-657, n. 4. The best evidence of this is that many parents with vouchers selected nonreligious private schools over religious alternatives and an even larger number of parents send their children to community and magnet schools rather than seeking vouchers at all. Supra, at 663-664. Moreover, there is no record evidence that any voucher-eligible student was turned away from a nonreligious private school in the voucher program, let alone a community or magnet school. See 234 F. 3d 945, 969 (CA6 2000) (Ryan, J., concurring in part and dissenting in part); Affidavit of David L. Brennan ¶ 8, App. 147a.

To support his hunch about the effect of the cap on tuition under the voucher program, Justice Soutee cites national data to suggest that, on average, Catholic schools have a cost advantage over other types of schools. See post, at 705-706, n. 15 (dissenting opinion). Even if national statistics were relevant for evaluating the Cleveland program, Justice Sou-tee ignores evidence which suggests that, at a national level, nonreligious private schools may target a market for a different, if not a higher, quality of education. For example, nonreligious private schools are smaller, see U. S. Dept. of Ed., National Center for Education Statistics, Private School Universe Survey, 1997-1998 (Oct. 1999) (Table 60) (87 and 269 students per private nonreligious and Catholic elementary school, respectively); have smaller class sizes, see ibid. (9.4 and 18.8 students per teacher at private nonreligious and Catholic elementary schools, respectively); have more highly educated teachers, see U. S. Dept. of Ed., National Center for Education Statistics, Private Schools in the United States: A Statistical Profile, 1993-1994 (NCES 97-459, July 1997) (Table 3.4) (37.9 percent of nonreligious private school teachers but only 29.9 percent of Catholic school teachers have Master’s degrees); and have principals with longer job tenure than Catholic schools, see ibid. (Table 3.7) (average ten*672ure of principals at private nonreligious and Catholic schools is 8.2 and 4.7 years, respectively).

Additionally, Justice Souter’s theory that the Cleveland voucher program’s cap on the tuition encourages low-income students to attend religious schools ignores that these students receive nearly double the amount of tuition assistance under the community schools program than under the voucher program and that none of the community schools is religious. See ante, at 647.

In my view the more significant finding in these cases is that Cleveland parents who use vouchers to send their children to religious private schools do so as a result of true private choice. The Court rejects, correctly, the notion that the high percentage of voucher recipients who enroll in religious private schools necessarily demonstrates that parents do not actually have the option to send their children to nonreligious schools. Ante, at 656-660. Likewise, the mere fact that some parents enrolled their children in religious schools associated with a different faith than their own, see post, at 704 (Souter, J., dissenting), says little about whether these parents had reasonable nonreligious options. Indeed, no voucher student has been known to be turned away from a nonreligious private school participating in the voucher program. Supra this page. This is impressive given evidence in the record that the present litigation has discouraged the entry of some nonreligious private schools into the voucher program. Declaration of David P. Zanotti ¶¶5, 10, App. 225a, 227a. Finally, as demonstrated above, the Cleveland program does not establish financial incentives to undertake a religious education.

I find the Court’s answer to the question whether parents of students eligible for vouchers have a genuine choice between religious and nonreligious schools persuasive. In looking at the voucher program, all the choices available to potential beneficiaries of the government program should be considered. In these cases, parents who were eligible to *673apply for a voucher also had the option, at a minimum, to send their children to community schools. Yet the Court of Appeals chose not to look at community schools, let alone magnet schools, when evaluating the Cleveland voucher program. See 234 F. 3d, at 958. That decision was incorrect. Focusing in these cases only on the program challenged by respondents ignores how the educational system in Cleveland actually functions. The record indicates that, in 1999, two nonreligious private schools that had previously served 15 percent of the students in the voucher program were prompted to convert to community schools because parents were concerned about the litigation surrounding the program, and because a new community schools program provided more per-pupil financial aid. Many of the students that enrolled in the two schools under the voucher program transferred to the community schools program and continued to attend these schools. See Affidavit of David L. Brennan ¶¶3, 10, App. 145a, 147a; Declaration of David P. Zanotti ¶¶ 4 — 10, id., at 225a-227a. This incident provides strong evidence that both parents and nonreligious schools view the voucher program and the community schools program as reasonable alternatives.

Considering all the educational options available to parents whose children are eligible for vouchers, including community and magnet schools, the Court finds that parents in the Cleveland schools have an array of nonreligious options. Ante, at 655. Not surprisingly, respondents present no evidence that any students who were candidates for a voucher were denied slots in a community school or a magnet school. Indeed, the record suggests the opposite with respect to community schools.' See Affidavit of David L. Brennan ¶ 8, App. 147 a.

Justice Souter nonetheless claims that, of the 10 community schools operating in Cleveland during the 1999-2000 school year, 4 were unavailable to students with vouchers and 4 others reported poor test scores. See post, at 702-*674703, n. 10 (dissenting opinion). But that analysis unreasonably limits the choices available to Cleveland parents. It is undisputed that Cleveland’s 24 magnet schools are reasonable alternatives to voucher schools. See post, at 701-702, n. 9 (Souter, J., dissenting); http://www.cmsdnet.net/ administration/EducationalServices/magnet.htm (June 20, 2002). And of the four community schools Justice Souter claims are unavailable to voucher students, he is correct only about one (Life Skills Center of Cleveland). Affidavit of Steven M. Puckett ¶ 12, App. 162a. JUSTICE Souter rejects the three other community schools (Horizon Science Academy, Cleveland Alternative Learning, and International Preparatory School) because they did not offer primary school classes, were targeted toward poor students or students with disciplinary or academic problems, or were not in operation for a year. See post, at 702-703, n. 10. But a community school need not offer primary school classes to be an alternative to religious middle schools, and catering to impoverished or otherwise challenged students may make a school more attractive to certain inner-city parents. Moreover, the one community school that was closed in 1999-2000 was merely looking for a new location and was operational in other years. See Affidavit of Steven M. Puckett ¶ 12, App. 162a; Ohio Dept, of Ed., Office of School Options, Community Schools, Ohio’s Community School Directory' (June 22, 2002), http://www.ode.state.oh.us/community_ schools/community_school_directory/default.asp. Two more community schools were scheduled to open after the 1999-2000 school year. See Affidavit of Steven M. Puckett ¶ 13, App. 163a.

Of the six community schools that Justice Souter admits as alternatives to the voucher program in 1999-2000, he notes that four (the Broadway, Cathedral, Chapelside, and Lincoln Park campuses of the Hope Academy) reported lower test scores than public schools during the school year after the District Court’s grant of summary judgment to re*675spondents, according to report cards prepared by the Ohio Department of Education. See post, at 702-703, n. 10 (dissenting opinion). (One, Old Brooklyn Montessori School, performed better than public schools. Ibid.; see also Ohio Dept, of Ed., 2001 Community School Report Card, Old Brooklyn Montessori School 5 (community school scored higher than public schools in four of five subjects in 1999-2000).) These report cards underestimate the value of the four Hope Academy schools. Before they entered the community school program, two of them participated in the voucher program. Although they received far less state funding in that capacity, they had among the highest rates of parental satisfaction of all voucher schools, religious or nonreligious. See P. Peterson, W. Howell, & J. Greene, An Evaluation of the Cleveland Voucher Program after Two Years 6, Table 4 (June 1999) (hereinafter Peterson). This is particularly impressive given that a Harvard University study found that the Hope Academy schools attracted the “poorest and most educationally disadvantaged students.” J. Greene, W. Howell, P. Peterson, Lessons from the Cleveland Scholarship Program 22, 24 (Oct. 15, 1997). Moreover, Justice Souter’s evaluation of the Hope Academy schools assumes that the only relevant measure of school quality is academic performance. It is reasonable to suppose, however, that parents in the inner city also choose schools that provide discipline and a safe environment for their children. On these dimensions some of the schools that Justice Sou-TER derides have performed quite ably. See Peterson, Table 7.

Ultimately, Justice Souter relies on very narrow data to draw rather broad conclusions. One year of poor test scores at four community schools targeted at the most challenged students from the inner city says little about the value of those schools, let alone the quality of the 6 other community schools and 24 magnet schools in Cleveland. Justice Sou-ter’s use of statistics confirms the Court’s wisdom in refus*676ing to consider them when assessing the Cleveland program’s constitutionality. See ante, at 658. What appears to motivate Justice Souter’s analysis is a desire for a limiting principle to rule out certain nonreligious schools as alternatives to religious schools in the voucher program. See post, at 700,701-702, n. 9 (dissenting opinion). But the goal of the Court’s Establishment Clause jurisprudence is to determine whether, after the Cleveland voucher program was enacted, parents were free to direct state educational aid in either a nonreligious or religious direction. See ante, at 655-656. That inquiry requires an evaluation of all reasonable educational options Ohio provides the Cleveland school system, regardless of whether they are formally made available in the same section of the Ohio Code as the voucher program.

Based on the reasoning in the Court’s opinion, which is consistent with the realities of the Cleveland educational system, I am persuaded that the Cleveland voucher program affords parents of eligible children genuine nonreligious options and is consistent with the Establishment Clause.

concurring.

Frederick Douglass once said that “[education .. . means emancipation. ■ It means light and liberty. It means the uplifting of the soul of man into the glorious light of truth, the light by which men can only be made free.”1 Today many of our inner-city public schools deny emancipation to urban minority students. Despite this Court’s observation nearly 50 years ago in Brown v. Board of Education, 347 U. S. 483, 493 (1954), that “it is doubtful that any child may reasonably be expected to succeed in life if he is denied the opportunity of an education,” urban children have been forced into a system that continually fails them. These cases present an *677example of such failures. Besieged by escalating financial problems and declining academic achievement, the Cleveland City School District was in the midst of an academic emergency when Ohio enacted its scholarship program.

The dissents and respondents wish to invoke the Establishment Clause of the First Amendment, as incorporated through the Fourteenth, to constrain a State’s neutral efforts to provide greater educational opportunity for underprivileged minority students. Today’s decision properly upholds the program as constitutional, and I join it in full.

I

This Court has often considered whether efforts to provide children with the best educational resources conflict with constitutional limitations. Attempts to provide aid to religious schools or to allow some degree of religious involvement in public schools have generated significant controversy and litigation as States try to navigate the line between the secular and the religious in education. See generally Illinois ex rel. McCollum v. Board of Ed. of School Dist. No. 71, Champaign Cty., 333 U. S. 203, 237-238 (1948) (Jackson, J., concurring) (noting that the Constitution does not tell judges “where the secular ends and the sectarian begins in education”). We have recently decided several cases challenging federal aid programs that include religious schools. See, e. g., Mitchell v. Helms, 530 U. S. 793 (2000); Agostini v. Felton, 521 U. S. 203 (1997). To determine whether a federal program survives scrutiny under the Establishment Clause, we have considered whether it has a secular purpose and whether it has the primary effect of advancing or inhibiting religion. See Mitchell, supra, at 807-808. I agree with the Court that Ohio’s program easily passes muster under our stringent test, but, as a matter of first principles, I question whether this test should be applied to the States.

*678The Establishment Clause of the First Amendment states that “Congress shall make no law respecting an establishment of religion.” On its face, this provision places no limit on the States with regard to religion. The Establishment Clause originally protected States, and by extension their citizens, from the imposition of an established religion by the Federal Government.2 Whether and how this Clause should constrain state action under the Fourteenth Amendment is a more difficult question.

The Fourteenth Amendment fundamentally restructured the relationship between individuals and the States and ensured that States would not deprive citizens of liberty without due process of law. It guarantees citizenship to all individuals born or naturalized in the United States and provides that “[n]o State shall make or enforce any law which shall abridge the privileges or immunities of citizens of the United States; nor shall any State deprive any person of life, liberty, or property, without due process of law; nor deny to any person within its jurisdiction the equal protection of the laws.” As Justice Harlan noted, the Fourteenth Amendment “added greatly to the dignity and glory of American citizenship, and to the security of personal liberty.” Plessy v. Ferguson, 163 U. S. 537, 555 (1896) (dissenting opinion). When rights are incorporated against the States through the Fourteenth Amendment they should advance, not constrain, individual liberty.

Consequently, in the context of the Establishment Clause, it may well be that state action should be evaluated on different terms than similar action by the Federal Government. “States, while bound to observe strict neutrality, should be freer to experiment with involvement [in religion] — on a neu*679tral basis — than the Federal Government.” Walz v. Tax Comm'n of City of New York, 397 U. S. 664, 699 (1970) (Harlan, J., concurring). Thus, while the Federal Government may “make no law respecting an establishment of religion,” the States may pass laws that include or touch on religious matters so long as these laws do not impede free exercise rights or any other individual religious liberty interest. By considering the particular religious liberty right alleged to be invaded by a State, federal courts can strike a proper balance between the demands of the Fourteenth Amendment on the one hand and the federalism prerogatives of States on the other.3

Whatever the textual and historical merits of incorporating the Establishment Clause, I can accept that the Fourteenth Amendment protects religious liberty rights.4 But I *680cannot accept its use to oppose neutral programs of school choice through the incorporation of the Establishment Clause. There would be a tragic irony in converting the Fourteenth Amendment’s guarantee of individual liberty into a prohibition on the exercise of educational choice.

II

The wisdom of allowing States greater latitude in dealing with matters of religion and education can be easily appreciated in this context. Respondents advocate using the Fourteenth Amendment to handcuff the State’s ability to experiment with education. But without education one can hardly exercise the civic, political, and personal freedoms conferred by the Fourteenth Amendment. Faced, with a severe educational crisis, the State of Ohio enacted wide-ranging educational reform that allows voluntary participation of private and religious schools in educating poor urban children otherwise condemned to failing public schools. The program does not force any individual to submit to religious indoctrination or education. It simply gives parents a greater choice as to where and in what manner to educate their children.5 This is a choice that those with greater means have routinely exercised.

*681Cleveland parents now have a variety of educational choices. There are traditional public schools, magnet schools, and privately run community schools, in addition to the scholarship program. Currently, 46 of the 66 private schools participating in the scholarship program are church affiliated (35 are Catholic), and 96 percent of students in the program attend religious schools. See App. 281a-286a; 234 F. 3d 945, 949 (CA6 2000). Thus, were the Court to disallow the inclusion of religious schools, Cleveland children could use their scholarships at only 10 private schools.

In addition to expanding the reach of the scholarship program, the inclusion of religious schools makes sense given Ohio’s purpose of increasing educational performance and opportunities. Religious schools, like other private schools, achieve far better educational results than their public counterparts. For example, the students at Cleveland’s Catholic schools score significantly higher on Ohio proficiency tests than students at Cleveland public schools. Of Cleveland eighth graders taking the 1999 Ohio proficiency test, 95 percent in Catholic schools passed the reading test, whereas only 57 percent in public schools passed. And 75 percent of Catholic school students passed the math proficiency test, compared to only 22 percent of public school students. See Brief for Petitioners in No. 00-1777, p. 10. But the success of religious and private schools is in the end beside the point, because the State has a constitutional right to experiment with a variety of different programs to promote educational opportunity. That Ohio’s program includes successful schools simply indicates that such reform can in fact provide improved education to underprivileged urban children.

Although one of the purposes of public schools was to promote democracy and a more egalitarian culture,6 failing urban public schools disproportionately affect minority children most in need of educational opportunity. At the time *682of Reconstruction, blacks considered public education “a matter of personal liberation and a necessary function of a free society.” J. Anderson, Education of Blacks in the South, 1860-1935, p. 18 (1988). Today, however, the promise of public school education has failed poor inner-city blacks. While in theory providing education to everyone, the quality of public schools varies significantly across districts. Just as blacks supported public education during Reconstruction, many blacks and other minorities now support school choice programs because they provide the greatest educational opportunities for their children in struggling communities.7 Opponents of the program raise formalistic concerns about the Establishment Clause but ignore the core purposes of the Fourteenth Amendment.

While the romanticized ideal of universal public education resonates with the cognoscenti who oppose vouchers, poor urban families just want the best education for their children, who will certainly need it to function in our high-tech and advanced society. As Thomas Sowell noted 30 years ago: “Most black people have faced too many grim, concrete problems to be romantics. They want and need certain tangible results, which can be achieved only by developing certain specific abilities.” Black Education: Myths and Tragedies 228 (1972). The same is true today. An individual’s life prospects increase dramatically with each successfully completed phase of education. For instance, a black high *683school dropout earns just over $13,500, but with a high school degree the average income is almost $21,000. Blacks with a bachelor’s degree have an average annual income of about $37,500, and $75,500 with a professional degree. See U. S. Dept, of Commerce, Bureau of Census, Statistical Abstract of the United States 140 (2001) (Table 218). Staying in school and earning a degree generates real and tangible financial benefits, whereas failure to obtain even a high school degree essentially relegates students to a life of poverty and, all too often, of crime.8 The failure to provide education to poor urban children perpetuates a vicious cycle of poverty, dependence, criminality, and alienation that continues for the remainder of their lives. If society cannot end racial discrimination, at least it can arm minorities with the education to defend themselves from some of discrimination’s effects.

* * *

Ten States have enacted some form of publicly funded private school choice as one means of raising the quality of education provided to underprivileged urban children.9 These programs address the root of the problem with failing urban public schools that disproportionately affect minority students. Society’s other solution to these educational failures is often to provide racial preferences in higher education. Such preferences, however, run afoul of the Fourteenth Amendment’s prohibition against distinctions based on race. See Plessy, 163 U. S., at 555 (Harlan, J., dissenting). By contrast, school choice programs that involve religious schools *684appear unconstitutional only to those who would twist the Fourteenth Amendment against itself by expansively incorporating the Establishment Clause. Converting the Fourteenth Amendment from a guarantee of opportunity to an obstacle against education reform distorts our constitutional values and disserves those in the greatest need.

As Frederick Douglass poignantly noted, “no greater benefit can be bestowed upon a long benighted people, than giving to them, as we are here earnestly this day endeavoring to do, the means of an education.”10

dissenting.

Is a law that authorizes the use of public funds to pay for the indoctrination of thousands of grammar school children in particular religious faiths a “law respecting an establishment of religion” within the meaning of the First Amendment? In answering that question, I think we should ignore three factual matters that are discussed at length by my colleagues.

First, the severe educational crisis that confronted the Cleveland City School District when Ohio enacted its voucher program is not a matter that should affect our appraisal of its constitutionality. In the 1999-2000 school year, that program provided relief to less than five percent of the students enrolled in the district's schools. The solution to the disastrous conditions that prevented over 90 percent of the student body from meeting basic proficiency standards obviously required massive improvements unrelated to the voucher program.1 Of course, the emergency may have *685given some families a powerful motivation to leave the public school system and accept religious indoctrination that they would otherwise have avoided, but that is not a valid reason for upholding the program.

Second, the wide range of choices that have been made available to students within the public school system has no bearing on the question whether the State may pay the tuition for students who wish to reject public education entirely and attend private schools that will provide them with a sectarian education. The fact that the vast majority of the voucher recipients who have entirely rejected public education receive religious indoctrination at state expense does, however, support the claim that the law is one “respecting an establishment of religion.” The State may choose to divide up its public schools into a dozen different options and label them magnet schools, community schools, or whatever else it decides to call them, but the State is still required to provide a public education and it is the State’s decision to fund private school education over and above its traditional obligation that is at issue in these cases.2

Third, the voluntary character of the private choice to prefer a parochial education over an education in the public school system seems to me quite irrelevant to the question whether the government’s choice to pay for religious indoctrination is constitutionally permissible. Today, however, the Court seems to have decided that the mere fact that a family that cannot afford a private education wants its children educated in a parochial school is a sufficient justification for this use of public funds.

For the reasons stated by Justice Souter and Justice Breyer, I am convinced that the Court’s decision is profoundly misguided. Admittedly, in reaching that conclusion *686I have been influenced by my understanding of the impact of religious strife on the decisions of our forbears to migrate to this continent, and on the decisions of neighbors in the Balkans, Northern Ireland, and the Middle East to mistrust one another. Whenever we remove a brick from the wall that was designed to separate religion and government, we increase the risk of religious strife and weaken the foundation of our democracy.

I respectfully dissent.

with whom Justice Stevens, Justice Ginsburg, and Justice Breyer join, dissenting.

The Court's majority holds that the Establishment Clause is no bar to Ohio’s payment of tuition at private religious elementary and middle schools under a scheme that systematically provides tax money to support the schools’ religious missions. The occasion for the legislation thus upheld is the condition of public education in the city of Cleveland. The record indicates that the schools are failing to serve their objective, and the vouchers in issue here are said to be needed to provide adequate alternatives to them. If there were an excuse for giving short shrift to the Establishment Clause, it would probably apply here. But there is no excuse. Constitutional limitations are placed on government to preserve constitutional values in hard cases, like these. “ [Constitutional lines have to be drawn, and on one side of every one of them is an otherwise sympathetic case that provokes impatience with the Constitution and with the line. But constitutional lines are the price of constitutional government.” Agostini v. Felton, 521 U. S. 203, 254 (1997) (Souter, J., dissenting). I therefore respectfully dissent.

The applicability of the Establishment Clause1 to public funding of benefits to religious schools was settled in Everson v. Board of Ed. of Ewing, 330 U. S. 1 (1947), which inau*687gurated the modern era of establishment doctrine. The Court stated the principle in words from which there was no dissent:

“No tax in any amount, large or small, can be levied to support any religious activities or institutions, whatever they may be called, or whatever form they may adopt to teach or practice religion.” Id., at 16.

The Court has never in so many words repudiated this statement, let alone, in so many words, overruled Everson.

Today, however, the majority holds that the Establishment Clause is not offended by Ohio’s Pilot Project Scholarship Program, under which students may be eligible to receive as much as $2,250 in the form of tuition vouchers transferable to religious schools. In the city of Cleveland the overwhelming proportion of large appropriations for voucher money must be spent on religious schools if it is to be spent at all, and will be spent in amounts that cover almost all of tuition. The money will thus pay for eligible students’ instruction not only in secular subjects but in religion as well, in schools that can fairly be characterized as founded to teach religious doctrine and to imbue teaching in all subjects with a religious dimension.2 Public tax money will pay at a systemic level for teaching the covenant with Israel and Mosaic law in Jewish schools, the primacy of the Apostle Peter and the Papacy in Catholic schools, the truth of reformed Christianity in Protestant schools, and the revelation to the Prophet in Muslim schools, to speak only of major religious groupings in the Republic.

*688How can a Court consistently leave Everson on the books and approve the Ohio vouchers? The answer is that it cannot. It is only by ignoring Everson that the majority can claim to rest on traditional law in its invocation of neutral aid provisions and private choice to sanction the Ohio law. It is, moreover, only by ignoring the meaning of neutrality and private choice themselves that the majority can even pretend to rest today’s decision on those criteria.

H-Í

The majority’s statements of Establishment Clause doctrine cannot be appreciated without some historical perspective on the Court’s announced limitations on government aid to religious education, and its repeated repudiation of limits previously set. My object here is not to give any nuanced exposition of the cases, which I tried to classify in some detail in an earlier opinion, see Mitchell v. Helms, 530 U. S. 793, 873-899 (2000) (dissenting opinion), but to set out the broad doctrinal stages covered in the modern era, and to show that doctrinal bankruptcy has been reached today.

Viewed with the necessary generality, the cases can be categorized in three groups. In the period from 1947 to 1968, the basic principle of no aid to religion through school benefits was unquestioned. Thereafter for some 15 years, the Court termed its efforts as attempts to draw a line against aid that would be divertible to support the religious, as distinct from the secular, activity of an institutional beneficiary. Then, starting in 1983, concern with divertibility was gradually lost in favor of approving aid in amounts unlikely to afford substantial benefits to religious schools, when offered evenhandedly without regard to a recipient’s religious character, and when channeled to a religious institution only by the genuinely free choice of some private individual. Now, the three stages are succeeded by a fourth, in which the substantial character of government aid is held to have no constitutional significance, and the espoused criteria *689of neutrality in offering aid, and private choice in directing it, are shown to be nothing but examples of verbal formalism.

A

Everson v. Board of Ed. of Ewing inaugurated the modern development of Establishment Clause doctrine at the behest of a taxpayer challenging state provision of “tax-raised funds to pay the bus fares of parochial school pupils” on regular city buses as part of a general scheme to reimburse the public-transportation costs of children attending both public and private nonprofit schools. 330 U. S., at 17. Although the Court split, no Justice disagreed with the basic doctrinal principle already quoted, that “[n]o tax in any amount . . . can be levied to support any religious activities or institutions, . . . whatever form they may adopt to teach . . . religion.” Id., at 16. Nor did any Member of the Court deny the tension between the New Jersey program and the aims of the Establishment Clause. The majority upheld the state law on the strength of rights of religious-school students under the Free Exercise Clause, id., at 17-18, which was thought to entitle them to free public transportation when offered as a “general government servie[e]” to all schoolchildren, id., at 17. Despite the indirect benefit to religious education, the transportation was simply treated like “ordinary police and fire protection, connections for sewage disposal, public highways and sidewalks,” id., at 17-18, and, most significantly, “state-paid policemen, detailed to protect children going to and from church schools from the very real hazards of traffic,” id., at 17. The dissenters, however, found the benefit to religion too pronounced to survive the general principle of no establishment, no aid, and they described it as running counter to every objective served by the establishment ban: New Jersey’s use of tax-raised funds foi'ced a taxpayer to “contribute] to the propagation of opinions which he disbelieves in so far as... religions differ,” id., at 45 (internal quotation marks omitted); it exposed religious *690liberty to the threat of dependence on state money, id., at 53; and it had already sparked political conflicts with opponents of public funding, id., at 54.3

The difficulty of drawing a line that preserved the basic principle of no aid was no less obvious some 20 years later in Board of Ed. of Central School Dist. No. 1 v. Allen, 392 U. S. 236 (1968), which upheld a New York law authorizing local school boards to lend textbooks in secular subjects to children attending religious schools, a result not self-evident from Everson’s “general government services” rationale. The Court relied instead on the theory that the in-kind aid could only be used for secular educational purposes, 392 U. S., at 243, and found it relevant that “no funds or books are furnished [directly] to parochial schools, and the financial benefit is to parents and children, not to schools,” id., at 243-244.4 Justice Black, who wrote Everson, led the dissenters. Textbooks, even when “‘secular,’ realistically will in some way inevitably tend to propagate the religious views of the favored sect,” 392 U. S., at 252, he wrote, and Justice Douglas raised other objections underlying the establishment ban, id., at 254-266. Religious schools would request those books most in keeping with their faiths, and public boards would have final approval power: “If the board of education supinely submits by approving and supplying the sectarian or sectarian-oriented textbooks, the struggle to keep church *691and state separate has been lost. If the board resists, then the battle line between church and state will have been drawn . . . Id., at 256 (Douglas, J., dissenting). The scheme was sure to fuel strife among religions as well: “we can rest assured that a contest will be on to provide those books for religious schools which the dominant religious group concludes best reflect the theocentric or other philosophy of the particular church.” Id., at 265.

Transcending even the sharp disagreement, however, was

“the consistency in the way the Justices went about deciding the case .... Neither side rested on any facile application of the ‘test’ or any simplistic reliance on the generality or evenhandedness of the state law. Disagreement concentrated on the true intent inferrable behind the law, the feasibility of distinguishing in fact between religious and secular teaching in church schools, and the reality or sham of lending books to pupils instead of supplying books to schools.... [T]he stress was on the practical significance of the actual benefits received by the schools.” Mitchell, 530 U. S., at 876 (Souter, J., dissenting).

B

Allen recognized the reality that “religious schools pursue two goals, religious instruction and secular education,” 392 U. S., at 245; if state aid could be restricted to serve the second, it might be permissible under the Establishment Clause. But in the retrenchment that followed, the Court saw that the two educational functions were so intertwined in religious primary and secondary schools that aid to secular education could not readily be segregated, and the intrusive monitoring required to enforce the line itself raised Establishment Clause concerns about the entanglement of church and state. See Lemon v. Kurtzman, 403 U. S. 602, 620 (1971) (striking down program supplementing salaries for teachers of secular subjects in private schools). To avoid *692the entanglement, the Court's focus in the post -Allen cases was on the principle of divertibility, on discerning when ostensibly secular government aid to religious schools was susceptible to religious uses. The greater the risk of diversion to religion (and the monitoring necessary to avoid it), the less legitimate the aid scheme was under the no-aid principle. On the one hand, the Court tried to be practical, and when the aid recipients were not so “pervasively sectarian” that their secular and religious functions were inextricably intertwined, the Court generally upheld aid earmarked for secular use. See, e. g., Roemer v. Board of Public Works of Md., 426 U. S. 736 (1976); Hunt v. McNair, 413 U. S. 734 (1973); Tilton v. Richardson, 403 U. S. 672 (1971). But otherwise the principle of nondivertibility was enforced strictly, with its violation being presumed in most cases, even when state aid seemed secular on its face. Compare, e.g., Levitt v. Committee for Public Ed. & Religious Liberty, 413 U. S. 472, 480 (1973) (striking down state program reimbursing private schools’ administrative costs for teacher-prepared tests in compulsory secular subjects), with Wolman v. Walter, 433 U. S. 229, 255 (1977) (upholding similar program using standardized tests); and Meek v. Pittenger, 421 U. S. 349, 369-372 (1975) (no public funding for staff and materials for “auxiliary services” like guidance counseling and speech and hearing services), with Wolman, supra, at 244 (permitting state aid for diagnostic speech, hearing, and psychological testing).

The fact that the Court’s suspicion of divertibility reflected a concern with the substance of the no-aid principle is apparent in its rejection of stratagems invented to dodge it. In Committee for Public Ed. & Religious Liberty v. Nyquist, 413 U. S. 756 (1973), for example, the Court struck down a New York program of tuition grants for poor parents and tax deductions for more affluent ones who sent their children to private schools. The Nyquist Court dismissed warranties of a “statistical guarantee,” that the scheme provided at most 15% of the total cost of an education at a religious school, *693id., at 787-788, which could presumably be matched to a secular 15% of a child’s education at the school. And it rejected the idea that the path of state aid to religious schools might be dispositive: “far from providing a per se immunity from examination of the substance of the State’s program, the fact that aid is disbursed to parents rather than to the schools is only one among many factors to be considered.” Id., at 781. The point was that “the effect of the aid is unmistakably to provide desired financial support for nonpublic, sectarian institutions.” Id., at 783.5 Nyquist thus held that aid to parents through tax deductions was no different from forbidden direct aid to religious schools for religious uses. The focus remained on what the public money bought when it reached the end point of its disbursement.

C

Like all criteria requiring judicial assessment of risk, diver tibility is an invitation to argument, but the object of the arguments provoked has always been a realistic assessment of facts aimed at respecting the principle of no aid. In Mueller v. Allen, 463 U. S. 388 (1983), however, that object began to fade, for Mueller started down the road from realism to formalism.

*694The aid in Mueller was in substance indistinguishable from that in Nyquist, see 463 U. S., at 396-397, n. 6, and both were substantively difficult to distinguish from aid directly to religious schools, id., at 399. But the Court upheld the Minnesota tax deductions in Mueller, emphasizing their neutral availability for religious and secular educational expenses and the role of private choice in taking them. Id., at 397-398. The Court relied on the same two principles in Witters v. Washington Dept. of Servs. for Blind, 474 U. S. 481 (1986), approving one student’s use of a vocational training subsidy for the blind at a religious college, characterizing it as aid to individuals from which religious schools could derive no “large” benefit: “the full benefits of the program [are not] limited, in large part or in whole, to students at sectarian institutions.” Id., at 488.

School Dist. of Grand Rapids v. Ball, 473 U. S. 373, 395-396, and n. 13 (1985), overruled in part by Agostini v. Felton, 521 U. S. 203 (1997), clarified that the notions of evenhandedness neutrality and private choice in Mueller did not apply to cases involving direct aid to religious schools, which were still subject to the divertibility test. But in Agostini, where the substance of the aid was identical to that in Ball, public employees teaching remedial secular classes in private schools, the Court rejected the 30-year-old presumption of divertibility, and instead found it sufficient that the aid “supplemented]” but did not “supplant” existing educational services, 521 U. S., at 210, 230. The Court, contrary to Ball, viewed the aid as aid “directly to the eligible students . . . no matter where they choose to attend school.” 521 U. S., at 229.

In the 12 years between Ball and Agostini, the Court decided not only Witters, but two other cases emphasizing the form of neutrality and private choice over the substance of aid to religious uses, but always in circumstances where any aid to religion was isolated and insubstantial. Zobrest v. Catalina Foothills School Dist., 509 U. S. 1 (1993), like Wit*695ters, involved one student’s choice to spend funds from a general public program at a religious school (to pay for a sign-language interpreter). As in Witters, the Court reasoned that “[disabled children, not sectarian schools, [were] the primary beneficiaries . . . ; to the extent sectarian schools benefit at all ... , they are only incidental beneficiaries.” 509 U. S., at 12. Rosenberger v. Rector and Visitors of Univ. of Va., 515 U. S. 819 (1995), like Zobrest and Witters, involved an individual and insubstantial use of neutrally available public funds for a religious purpose (to print an evangelical magazine).

To be sure, the aid in Agostini was systemic and arguably substantial, but, as I have said, the majority there chose to view it as a bare “supplement.” 521 U. S., at 229. And this was how the controlling opinion described the systemic aid in our most recent case, Mitchell v. Helms, 530 U. S. 793 (2000), as aid going merely to a “portion” of the religious schools’ budgets, id., at 860 (O’Connor, J., concurring in judgment). The plurality in that case did not feel so uncomfortable about jettisoning substance entirely in favor of form, finding it sufficient that the aid was neutral and that there was virtual private choice, since any aid “first passes through the hands (literally or figuratively) of numerous private citizens who are free to direct the aid elsewhere.” Id., at 816. But that was only the plurality view.

Hence it seems fair to say that it was not until today that substantiality of aid has clearly been rejected as irrelevant by a majority of this Court, just as it has not been until today that a majority, not a plurality, has held purely formal criteria to suffice for scrutinizing aid that ends up in the coffers of religious schools. Today’s cases are notable for their stark illustration of the inadequacy of the majority’s chosen formal analysis.

II

Although it has taken half a century since Everson to reach the majority’s twin standards of neutrality and *696free choice, the facts show that, in the majority’s hands, even these criteria cannot convincingly legitimize the Ohio scheme.

A

Consider first the criterion of neutrality. As recently as two Terms ago, a majority of the Court recognized that neutrality conceived of as evenhandedness toward aid recipients had never been treated as alone sufficient to satisfy the Establishment Clause, Mitchell, 530 U. S., at 838-839 (O’Con-nor, J., concurring in judgment); id., at 884 (Souter, J., dissenting). But at least in its limited significance, formal neutrality seemed to serve some purpose. Today, however, the majority employs the neutrality criterion in a way that renders it impossible to understand.

Neutrality in this sense refers, of course, to evenhandedness in setting eligibility as between potential religious and secular recipients of public money. Id., at 809-810 (plurality opinion); id., at 878-884 (Souter, J., dissenting) (three senses of “neutrality”).6 Thus, for example, the aid scheme in Witters provided an eligible recipient with a scholarship to be used at any institution within a practically unlimited universe of schools, 474 U. S., at 488; it did not tend to provide more or less aid depending on which one the scholarship recipient chose, and there was no indication that the maximum scholarship amount would be insufficient at secular *697schools. Neither did any condition of Zobrest’s interpreter’s subsidy favor religious education. See 509 U. S., at 10.

In order to apply the neutrality test, then, it makes sense to focus on a category of aid that may be directed to religious as well as secular schools, and ask whether the scheme favors a religious direction. Here, one would ask whether the voucher provisions, allowing for as much as $2,250 toward private school tuition (or a grant to a public school in an adjacent district), were written in a way that skewed the scheme toward benefiting religious schools.

This, however, is not what the majority asks. The majority looks not to the provisions for tuition vouchers, Ohio Rev. Code Ann. § 3313.976 (West Supp. 2002), but to every provision for educational opportunity: “The program permits the participation of all schools within the district, [as well as public schools in adjacent districts], religious or nonreligious.” Ante, at 653 (emphasis in original). The majority then finds confirmation that “participation of all schools” satisfies neutrality by noting that the better part of total state educational expenditure goes to public schools, ante, at 654, thus showing there is no favor of religion.

The illogic is patent. If regular, public schools (which can get no voucher payments) “participate” in a voucher scheme with schools that can, and public expenditure is still predominantly on public schools, then the majority’s reasoning would find neutrality in a scheme of vouchers available for private tuition in districts with no secular private schools at all. “Neutrality” as the majority employs the term is, literally, verbal and nothing more. This, indeed, is the only way the majority can gloss over the very nonneutral feature of the total scheme covering “all schools”: public tutors may receive from the State no more than $324 per child to support extra tutoring (that is, the State’s 90% of a total amount of $360), App. 166a, whereas the tuition voucher schools (which *698turn out to be mostly religious) can receive up to $2,250, id., at 56a.7

Why the majority does not simply accept the fact that the challenge here is to the more generous voucher scheme and judge its neutrality in relation to religious use of voucher money seems very odd. It seems odd, that is, until one recognizes that comparable schools for applying the criterion of neutrality are also the comparable schools for applying the other majority criterion, whether the immediate recipients of voucher aid have a genuinely free choice of religious and secular schools to receive the voucher money. And in applying this second criterion, the consideration of “all schools” is ostensibly helpful to the majority position.

B

The majority addresses the issue of choice the same way it addresses neutrality, by asking whether recipients or potential recipients of voucher aid have a choice of public schools among secular alternatives to religious schools. Again, however, the majority asks the wrong question and misapplies the criterion. The majority has confused choice in spending scholarships with choice from the entire menu of *699possible educational placements, most of them open to anyone willing to attend a public school. I say “confused” because the majority’s new use of the choice criterion, which it frames negatively as “whether Ohio is coercing parents into sending their children to religious schools/’ ante, at 655-656, ignores the reason for having a private choice enquiry in the first place. Cases since Mueller have found private choice relevant under a rule that aid to religious schools can be permissible so long as it first passes through the hands of students or parents.8 The majority’s view that all educational choices are comparable for purposes of choice thus ignores the whole point of the choice test: it is a criterion for deciding whether indirect aid to a religious school is legitimate because it passes through private hands that can spend or use the aid in a secular school. The question is whether the private hand is genuinely free to send the money in either a secular direction or a religious one. The majority now has transformed this question about private choice in channeling aid into a question about selecting from examples of state spending (on education) including direct spending on magnet and community public schools that goes through no private hands and could never reach a religious school under any circumstance. When the choice test is transformed from where to spend the money to where to go to school, it is cut loose from its very purpose.

*700Defining choice as choice in spending the money or channeling the aid is, moreover, necessary if the choice criterion is to function as a limiting principle at all. If “choice” is present whenever there is any educational alternative to the religious school to which vouchers can be endorsed, then there will always be a choice and the voucher can always be constitutional, even in a system in which there is not a single private secular school as an alternative to the religious school. See supra, at 697 (noting the same result under the majority’s formulation of the neutrality criterion). And because it is unlikely that any participating private religious school will enroll more pupils than the generally available public system, it will be easy to generate numbers suggesting that aid to religion is not the significant intent or effect of the voucher scheme.

That is, in fact, just the kind of rhetorical argument that the majority accepts in these cases. In addition to secular private schools (129 students), the majority considers public schools with tuition assistance (roughly 1,400 students), magnet schools (13,000 students), and community schools (1,900 students), and concludes that fewer than 20% of pupils receive state vouchers to attend religious schools. Ante, at 659. (In fact, the numbers would seem even more favorable to the majority’s argument if enrollment in traditional public schools without tutoring were considered, an alternative the majority thinks relevant to the private choice en-quiry, ante, at 655.) Justice O’Connor focuses on how much money is spent on each educational option and notes that at most $8.2 million is spent on vouchers for students attending religious schools, ante, at 664 (concurring opinion), which is only 6% of the State’s expenditure if one includes separate funding for Cleveland’s community ($9.4 million) and magnet ($114.8 million) public schools. The variations show how results may shift when a judge can pick and choose the alternatives to use in the comparisons, and they also show what dependably comfortable results the choice crite*701rion will yield if the identification of relevant choices is wide open. If the choice of relevant alternatives is an open one, proponents of voucher aid will always win, because they will always be able to find a “choice” somewhere that will show the bulk of public spending to be secular. The choice en-quiry will be diluted to the point that it can screen out nothing, and the result will always be determined by selecting the alternatives to be treated as choices.

Confining the relevant choices to spending choices, on the other hand, is not vulnerable to comparable criticism. Although leaving the selection of alternatives for choice wide open, as the majority would, virtually guarantees the availability of a “choice” that will satisfy the criterion, limiting the choices to spending choices will not guarantee a negative result in every case. There may, after all, be cases in which a voucher recipient will have a real choice, with enough secular private school desks in relation to the number of religious ones, and a voucher amount high enough to meet secular private school tuition levels. See infra, at 704-707. But, even to the extent that choice-to-spend does tend to limit the number of religious funding options that pass muster, the choice criterion has to be understood this way in order, as I have said, for it to function as a limiting principle.9 Otherwise *702there is surely no point in requiring the choice to be a true or real or genuine one.10

*703It is not, of course, that I think even a genuine choice criterion is up to the task of the Establishment Clause when substantial state funds go to religious teaching; the discussion in Part III, infra, shows that it is not. The point is simply that if the majority wishes to claim that choice is a criterion, it must define choice in a way that can function as a criterion with a practical capacity to screen something out.

If, contrary to the majority, we ask the right question about genuine choice to use the vouchers, the answer shows that something is influencing choices in a way that aims the money in a religious direction: of 56 private schools in the district participating in the voucher program (only 53 of which accepted voucher students in 1999-2000), 46 of them are religious; 96.6% of all voucher recipients go to religious schools, only 3.4% to nonreligious ones. See App. 281a-286a. Unfortunately for the majority position, there is no explanation for this that suggests the religious direction results simply from free choices by parents. One answer to these statistics, for example, which would be consistent with the genuine choice claimed to be operating, might be that 96.6% of families choosing to avail themselves of vouchers choose to educate their children in schools of their own religion. This would not, in my view, render the scheme constitutional, but it would speak to the majority’s choice criterion. *704Evidence shows, however, that almost two out of three families using vouchers to send their children to religious schools did not embrace the religion of those schools. App. to Pet. for Cert, in No. 00-1777, p. 147a.11 The families made it clear they had not chosen the schools because they wished their children to be proselytized in a religion not their own, or in any religion, but because of educational opportunity.12

Even so, the fact that some 2,270 students chose to apply their vouchers to schools of other religions, App. 281a-286a, might be consistent with true choice if the students “chose” their religious schools over a wide array of private nonreligious options, or if it could be shown generally that Ohio’s program had no effect on educational choices and thus no impermissible effect of advancing religious education. But both possibilities are contrary to fact. First, even if all existing nonreligious private schools in Cleveland were willing to accept large numbers of voucher students, only a few more than the 129 currently enrolled in such schools would be able to attend, as the total enrollment at all nonreligious private schools in Cleveland for kindergarten through eighth grade is only 510 children, see Brief for California Alliance for Public Schools as Amicus Curiae 15, and there is no indication that these schools have many open seats.13 Second, the *705$2,500 cap that the program places on tuition for participating low-income pupils has the effect of curtailing the participation of nonreligious schools: “nonreligious schools with higher tuition (about $4,000) stated that they could afford to accommodate just a few voucher students.”14 By comparison, the average tuition at participating Catholic schools in Cleveland in 1999-2000 was $1,592, almost $1,000 below the cap.15

*706Of course, the obvious fix would be to increase the value of vouchers so that existing nonreligious private and non-Catholic religious schools would be able to enroll more voucher students, and to provide incentives for educators to create new such schools given that few presently exist. Private choice, if as robust as that available to the seminarian in Witters, would then be “true private choice” under the majority’s criterion. But it is simply unrealistic to presume that parents of elementary and middle school students in Cleveland will have a range of secular and religious choices even arguably comparable to the statewide program for vocational and higher education in Witters. And to get to that hypothetical point would require that such massive financial support be made available to religion as to disserve every objective of the Establishment Clause even more than the present scheme does. See Part III — B, infra.16

*707There is, in any case, no way to interpret the 96.6% of current voucher money going to religious schools as reflecting a free and genuine choice by the families that apply for vouchers. The 96.6% reflects, instead, the fact that too few nonreligious school desks are available and few but religious schools can afford to accept more than a handful of voucher students. And contrary to the majority’s assertion, ante, at 654, public schools in adjacent districts hardly have a financial incentive to participate in the Ohio voucher program, and none has.17 For the overwhelming number of children in the voucher scheme, the only alternative to the public schools is religious. And it is entirely irrelevant that the State did not deliberately design the network of private schools for the sake of channeling money into religious institutions. The criterion is one of genuinely free choice on the part of the private individuals who choose, and a Hobson’s choice is not a choice, whatever the reason for being Hobsonian.

Ill

I do not dissent merely because the majority has misapplied its own law, for even if I assumed arguendo that the *708majority’s formal criteria were satisfied on the facts, today’s conclusion would be profoundly at odds with the Constitution. Proof of this is clear on two levels. The first is circumstantial, in the now discarded symptom of violation, the substantial dimension of the aid. The second is direct, in the defiance of every objective supposed to be served by the bar against establishment.

A

The scale of the aid to religious schools approved today is unprecedented, both in the number of dollars and in the proportion of systemic school expenditure supported. Each measure has received attention in previous cases. On one hand, the sheer quantity of aid, when delivered to a class of religious primary and secondary schools, was suspect on the theory that the greater the aid, the greater its proportion to a religious school’s existing expenditures, and the greater the likelihood that public money was supporting religious as well as secular instruction. As we said in Meek, “it would simply ignore reality to attempt to separate'secular educational functions from the predominantly religious role” as the object of aid that comes in “substantial amounts.” 421 U. S., at 365. Cf. Nyquist, 413 U. S., at 787-788 (rejecting argument that tuition assistance covered only 15% of education costs, presumably secular, at religious schools). Conversely, the more “attenuated [the] financial benefit. . . that eventually flows to parochial schools,” the more the Court has been willing to find a form of state aid permissible. Mueller, 463 U. S., at 400.18

*709On the other hand, the Court has found the gross amount unhelpful for Establishment Clause analysis when the aid afforded a benefit solely to one individual, however substantial as to him, but only an incidental benefit to the religious school at which the individual chose to spend the State’s money. See Witters, 474 U. S., at 488; cf. Zobrest, 509 U. S., at 12. When neither the design nor the implementation of an aid scheme channels a series of individual students’ subsidies toward religious recipients, the relevant beneficiaries for establishment purposes, the Establishment Clause is unlikely to be implicated. The majority’s reliance on the observations of five Members of the Court in Witters as to the irrelevance of substantiality of aid in that case, see ante, at 651, is therefore beside the point in the matter before us, which involves considerable sums of public funds systematically distributed through thousands of students attending religious elementary and middle schools in the city of Cleveland.19

*710The Cleveland voucher program has cost Ohio taxpayers $33 million since its implementation in 1996 ($28 million in voucher payments, $5 million in administrative costs), and its cost was expected to exceed $8 million in the 2001-2002 school year. People for the American Way Foundation, Five Years and Counting: A Closer Look at the Cleveland Voucher Program 1-2 (Sept. 25, 2001) (hereinafter Cleveland Voucher Program) (cited in Brief for National School Boards Association et al. as Amici Curiae 9). These tax-raised funds are on top of the textbooks, reading and math tutors, laboratory equipment, and the like that Ohio provides to private schools, worth roughly $600 per child. Cleveland Voucher Program 2.20

The gross amounts of public money contributed are symptomatic of the scope of what the taxpayers’ money buys for a broad class of religious-school students. In paying for practically the full amount of tuition for thousands of qualifying students,21 cf. Nyquist, supra, at 781-783 .(state aid amounting to 50% of tuition was unconstitutional), the scholarships purchase everything that tuition purchases, be it instruction in math or indoctrination in faith. The eonse-*711quenees of “substantial” aid hypothesized in Meek are realized here: the majority makes no pretense that substantial amounts of tax money are not systematically underwriting religious practice and indoctrination.

B

It is virtually superfluous to point out that every objective underlying the prohibition of religious establishment is betrayed by this scheme, but something has to be said about the enormity of the violation. I anticipated these objectives earlier, supra, at 689-690, in discussing Everson, which cataloged them, the first being respect for freedom of conscience. Jefferson described it as the idea that no one “shall be compelled to... support any religious worship, place, or ministry whatsoever,” A Bill for Establishing Religious Freedom, in 5 The Founders’ Constitution 84 (P. Kurland & R. Lerner eds. 1987), even a “teacher of his own religious persuasion,” ibid., and Madison thought it violated by any “‘authority which can force a citizen to contribute three pence ... of his property for the support of any ... establishment.’ ” Memorial and Remonstrance ¶ 3, reprinted in Everson, 330 U. S., at 65-66. “Any tax to establish religion is antithetical to the command that the minds of men always be wholly free,” Mitchell, 530 U. S., at 871 (Souter, J., dissenting) (internal quotation marks and citations omitted).22 Madison’s objection to three pence has simply been lost in the majority’s formalism.

As for the second objective, to save religion from its own corruption, Madison wrote of the “ ‘experience ... that eccle*712siastical establishments, instead of maintaining the purity and efficacy of Religion, have had a contrary operation.’” Memorial and Remonstrance ¶ 7, reprinted in Everson, 330 U. S., at 67. In Madison’s time, the manifestations were “pride and indolence in the Clergy; ignorance and servility in the laity[,] in both, superstition, bigotry and persecution,” ibid.; in the 21st century, the risk is one of “corrosive secularism” to religious schools, Ball, 473 U. S., at 385, and the specific threat is to the primacy of the schools’ mission to educate the children of the faithful according to the unaltered precepts of their faith. Even “[t]he favored religion may be compromised as political figures reshape the religion’s beliefs for their own purposes; it may be reformed as government largesse brings government regulation.” Lee v. Weisman, 505 U. S. 577, 608 (1992) (Blackmun, J., concurring).

The risk is already being realized. In Ohio, for example, a condition of receiving government money under the program is that participating religious schools may not “discriminate on the basis of . . . religion,” Ohio Rev. Code Ann. § 3313.976(A)(4) (West Supp. 2002), which means the school may not give admission preferences to children who are members of the patron faith; children of a parish are generally consigned to the same admission lotteries as nonbelievers, §§3313.977(A)(l)(c)-(d). This indeed was the exact object of a 1999 amendment repealing the portion of a predecessor statute that had allowed an admission preference for “[cjhildren ... whose parents are affiliated with any organization that provides financial support to the school, at the discretion of the school.” § 3313.977(A)(1)(d) (West 1999). Nor is the State’s religious antidiscrimination restriction limited to student admission policies: by its terms, a participating religious school may well be forbidden to choose a member of its own clergy to serve as teacher or principal over a layperson of a different religion claiming *713equal qualification for the job.23 Cf. National Catholic Educational Association, Balance Sheet for Catholic Elementary Schools: 2001 Income and Expenses 25 (2001) (“31% of [reporting Catholic elementary and middle] schools had at least one full-time teacher who was a religious sister”)- Indeed, a separate condition that “[t]he school. . . not . . . teach hatred of any person or group on the basis of . . . religion,” § 3313.976(A)(6) (West Supp. 2002), could be understood (or subsequently broadened) to prohibit religions from teaching traditionally legitimate articles of faith as to the error, sinfulness, or ignorance of others,24 if they want government money for their schools.

*714For perspective on this foot-in-the-door of religious regulation, it is well to remember that the money has barely begun to flow. Prior examples of aid, whether grants through individuals or in-kind assistance, were never significant enough to alter the basic fiscal structure of religious schools; state aid was welcome, but not indispensable. See, e. g., Mitchell, 530 U. S., at 802 (federal funds could only supplement funds from nonfederal sources); Agostini, 521 U. S., at 210 (federally funded services could “ ‘supplement, and in no case supplant, the level of services’” already provided). But given the figures already involved here, there is no question that religious schools in Ohio are on the way to becoming bigger businesses with budgets enhanced to fit their new stream of tax-raised income. See, e. g., People for the American Way Foundation, A Painful Price 5, 9, 11 (Feb. 14, 2002) (of 91 schools participating in the Milwaukee program, 75 received voucher payments in excess of tuition, 61 of those were religious and averaged $185,000 worth of overpayment per school, justified in part to “raise low salaries”). The administrators of those same schools are also no doubt following the politics of a move in the Ohio State Senate to raise the current maximum value of a school voucher from $2,250 to the base amount of current state spending on each public school student ($4,814 for the 2001 fiscal year). See Bloedel, Bill Analysis of S. B. No. 89,124th Ohio Gen. Assembly, regular session 2001-2002 (Ohio Legislative Service Commission). Ohio, in fact, is merely replicating the experience in Wisconsin, where a similar increase in the value of educational vouchers in Milwaukee has induced the creation of some 23 new private schools, Public Policy Forum, Research Brief, vol. 90, no. 1, p. 3 (Jan. 23, 2002), some of which, we may safely surmise, are religious. New schools have presumably *715pegged their financial prospects to the government from the start, and the odds are that increases in government aid will bring the threshold voucher amount closer to the tuition at even more expensive religious schools.

When government aid goes up, so does reliance on it; the only thing likely to go down is independence. If Justice Douglas in Allen was concerned with state agencies, influenced by powerful religious groups, choosing the textbooks that parochial schools would use, 392 U. S., at 265 (dissenting opinion), how much more is there reason to wonder when dependence will become great enough to give the State of Ohio an effective veto over basic decisions on the content of currículums? A day will come when religious schools will learn what political leverage can do, just as Ohio’s politicians are now getting a lesson in the leverage exercised by religion.

Increased voucher spending is not, however, the sole portent of growing regulation of religious practice in the school, for state mandates to moderate religious teaching may well be the most obvious response to the third concern behind the ban on establishment, its inextricable link with social conflict. See Mitchell, supra, at 872 (Souter, J., dissenting); Everson, 330 U. S., at 8-11. As appropriations for religious subsidy rise, competition for the money will tap sectarian religion’s capacity for discord. “Public money devoted to payment of religious costs, educational or other, brings the quest for more. It brings too the struggle of sect against sect for the larger share or for any. Here one by numbers alone will benefit most, there another.” Id., at 53. (Rutledge, J., dissenting).

Justice Breyer has addressed this issue in his own dissenting opinion, which I join, and here it is enough to say that the intensity of the expectable friction can be gauged by realizing that the scramble for money will energize not only contending sectarians, but taxpayers who take their liberty of conscience seriously. Religious teaching at taxpayer *716expense simply cannot be cordoned from taxpayer politics, and every major religion currently espouses social positions that provoke intense opposition. Not all taxpaying Protestant citizens, for example, will be content to underwrite the teaching of the Roman Catholic Church condemning the death penalty.25 Nor will all of America’s Muslims acquiesce in paying for the endorsement of the religious Zionism taught in many religious Jewish schools, which combines “a nationalistic sentiment” in support of Israel with a “deeply religious” element.26 Nor will every secular taxpayer be content to support Muslim views on differential treatment of the sexes,27 or, for that matter, to fund the espousal of a wife’s obligation of obedience to her husband, presumably taught in any schools adopting the articles of faith of the Southern Baptist Convention.28 Views like these, and innumerable others, have been safe in the sectarian pulpits and classrooms of this Nation not only because the Free Exercise Clause protects them directly, but because the ban on supporting religious establishment has protected free exercise, by keeping it relatively private. With the arrival of vouchers in religious schools, that privacy will go, and along with it will go confidence that religious disagreement will stay moderate.

* * *

If the divisiveness permitted by today’s majority is to be avoided in the short term, it will be avoided only by action *717of the political branches at the state and national levels. Legislatures not driven to desperation by the problems of public education may be able to see the threat in vouchers negotiable in sectarian schools. Perhaps even cities with problems like Cleveland’s will perceive the danger, now that they know a federal court will not save them from it.

My own course as a judge on the Court cannot, however, simply be to hope that the political branches will save us from the consequences of the majority’s decision. Everson’s statement is still the touchstone of sound law, even though the reality is that in the matter of educational aid the Establishment Clause has largely been read away. True, the majority has not approved vouchers for religious schools alone, or aid earmarked for religious instruction. But no scheme so clumsy will ever get before us, and in the cases that we may see, like these, the Establishment Clause is largely silenced. I do not have the option to leave it silent, and I hope that a future Court will reconsider today’s dramatic departure from basic Establishment Clause principle.