Note from the Editor: This article discusses the effect that third-party harms should have on religious accommodations or claims for religious exemptions. The Federalist Society takes no positions on particular legal and public policy matters. Any expressions of opinion are those of the author. Whenever we publish an article that advocates for a particular position, as here, we offer links to other perspectives on the issue, including ones opposed to the position taken in the article. We also invite responses from our readers. To join the debate, please email us at firstname.lastname@example.org.
Introduction: Religious Freedom and Third-Party Harms
In recent years religious accommodation issues have become increasingly contentious, particularly issues concerning religious organizations and individuals who object to being forced to facilitate contraception, abortion, or same-sex marriages and relationships. Increasingly, opposition to religious-freedom claims focuses on harm, or the “shifting of costs,” to third parties. For example, several scholars argued that exempting for-profit employers from the Obama administration’s contraception mandate would violate the Establishment Clause because it would harm employees by denying them the valuable statutory benefit of free insurance coverage for contraception. The federal government likewise argued, in somewhat softer form, that an exemption would harm employees and therefore should be refused under the Religious Freedom Restoration Act (RFRA). The Supreme Court avoided this argument in Burwell v. Hobby Lobby Stores, Inc., by finding that employees could receive identical contraception coverage through insurers and third-party administrators without imposing on objecting employers.
As the contraception-mandate litigation shows, arguments asserting third-party harms take two forms. The first appears when a person or group whose religious practice is substantially restricted by a law makes a claim for an exemption under RFRA, a similar state religious-freedom statute, or a protective state constitutional provision. Under all these provisions, the claimant must first show that applying the law would “substantially burden” religious exercise; if it would, then the government must show that applying the law is the “least restrictive means” of serving a “compelling governmental interest,” which it may show—at least in some cases—on the basis that the religious exercise causes certain harms to third parties.
The second sort of assertion of third-party harms arises when a religious exemption has been declared by a legislative enactment or a judicial ruling. The exemption’s opponents—government officials or private parties—then might argue that it violates the Establishment Clause. Under case law, a court asking whether an accommodation (exemption) of religious exercise amounts to an establishment of religion should consider, among other things, whether the exemption removes a significant burden from religion and takes “adequate account of the burdens on third parties.”
Thus the two arguments, free exercise claims for exemptions and Establishment Clause challenges to exemptions, require looking at similar factors. Both involve examining (1) the nature and seriousness of the burden that the law in question would impose on religious exercise, and (2) the nature and seriousness of the effect on others if the claimant is exempted from the law. But identifying these two considerations does not answer the question how they should be compared with each other. How should burdens on religion and those on others be weighed? And how significant must the third-party harms be to overcome religious claims?
The chief assertion of this article is that harms to others should not be conclusive against religious exemptions under either free exercise or nonestablishment principles. Such harms can certainly be a reason to deny exemption, but they are not the end of the inquiry: a number of factors must be considered. In particular, I argue, Establishment Clause limits on religious exemptions should not be strict. An exemption is not unconstitutional merely because it has negative effects on others: the burdens on others must be significantly disproportionate to the burdens that it removes from religion.
Part I of this article makes general observations about the problem of exemptions and third-party harms. Part II then discusses the scope of accommodation under RFRA or similar state provisions. Part III discusses the limits the Establishment Clause may impose on accommodations that affect others.
I. The Analytical Problem of Third-Party Harms
It may seem obvious that religious freedom does not authorize one person to harm or shift costs to another. Eugene Volokh writes that “religious freedom rights are often articulated as a right to do what your religion motivates you to do, simply because of your religious motivation, but only so long as it doesn’t harm the rights of others.” Obviously religious freedom does not protect killing someone in a ritual sacrifice, or defrauding others because the perpetrator perceives a religious duty.
But the problem comes in defining terms like “causing harms” or “shifting costs.” In earlier eras of smaller government, legal prohibitions generally focused on a limited set of direct harms to another’s body, physical or financial property, or contractual rights. Thus, a number of founding-era figures emphasized that religious freedom gave no one the right to harm others; but the harms they referred to were immediate, concrete, and serious matters like assault and theft. Pierre Bayle defended magistrates’ power and duty “to maintain society and punish all those who destroy the foundations, as murderers and robbers do”; and Thomas Jefferson spoke of religious freedom for actions that “neither pic[k] my pocket nor brea[k] my leg.”
This framework prohibited many harms, but it also left a large zone of freedom in which religious organizations and individuals could act, in ways that affected others but were not defined as a legal harms. For example, before the rise of modern employment regulation—nondiscrimination laws, collective bargaining requirements, and so forth—religious organizations were legally free to set religiously grounded standards for their employees.
This has changed with the rise of the welfare-regulatory state, which declares much broader legal harms. For example, at-will employment has given way to extensive regulation of the employment relationship: government declares it a legal harm when an employee is barred from unionizing or is discriminated against based on a prohibited characteristic. Under post-1937 constitutional jurisprudence, government has broad prima facie power to define, declare, and prohibit such harms. The modern state is not limited to imposing liability for actual harmful effects; it may declare legal rights designed to head off such effects. And it may frame them as benefits or rights for individual third parties. For example, to prevent the ultimate material harms of labor strife and unfair treatment of employees, government can declare rights of employees to unionize and can allow individuals to sue to enforce the right.
But just because government can prima facie regulate does not mean it can do so in ways that substantially burden religious exercise. The very point of the freedoms listed in the Bill of Rights, including religious freedom, is to place limits on actions otherwise within the government’s power. If religious freedom confers no right to harm others, and the government can define anything it wishes as a harm, then the regulatory state will severely constrict religious freedom. For example, once Title VII and analogous laws defined various forms of discrimination as a legal harm to employees, religious organizations faced lawsuits triggering civil court review of their employment decisions concerning their clergy and other leaders. Their ability to choose their leaders was preserved only by a court-ordered religious exemption: the ministerial exception, affirmed in Hosanna-Tabor Evangelical Lutheran Church & School v. EEOC.
The contraception mandate is a prime example of modern government declaring a legal entitlement unknown to the common law: guaranteed insurance coverage (for contraception) without cost-sharing. There may well be good reasons for creating such an entitlement (I personally think there are, in many cases). But it also creates new conflicts with the religious tenets of organizations and individuals. The government should not be able to win such conflicts simply by creating an entitlement and defining its denial as a harm. The Supreme Court recognized this in analyzing the contraception mandate in Hobby Lobby. The extent to which a denial of a benefit materially affects others, the Court said, “will often inform the analysis of the Government’s compelling interest and the availability of a less restrictive means” of advancing it. But it cannot be, the Court added:
that any burden on religious exercise, no matter how onerous and no matter how readily the government interest could be achieved through alternative means, is permissible under RFRA so long as the relevant legal obligation requires the religious adherent to confer a benefit on third parties. [If that were so, then by] framing any Government regulation as benefiting a third party, the Government could turn all regulations into entitlements to which nobody could object on religious grounds, rendering RFRA meaningless.
A number of familiar, accepted religious accommodations involve clear effects on individual third parties. Some of these accommodations are constitutionally required, and all are constitutionally permissible. Draft exemptions shift harm from the pacifist to another person who must be drafted. The clergy-penitent privilege may shift harm to the crime or tort victim who loses the benefit of testimony. The ministerial exception to non-discrimination laws, which Hosanna-Tabor unanimously held was constitutionally required, allows a religious organization to fire a minister for otherwise legally impermissible reasons. The religious-hiring exemption in Title VII, unanimously held permissible in Corporation of Presiding Bishop v. Amos, allows a religious organization to fire or refuse to hire employees outside of its own faith. Protecting faith-based homeless shelters or food pantries from overly restrictive zoning regulations can have some effect on neighbors’ property values. And in cases, like Sherbert v. Verner, where a worker claims unemployment benefits after leaving a job because of a religious conflict, the claim for benefits increases an employer’s rate of assessment for unemployment taxes. These and other examples vindicate Hobby Lobby’s warning that in the era of the active state, many well-accepted protections would be eliminated if it were impermissible for religious freedom to affect the rights of third parties.
If religious freedom is to continue receiving strong weight in an era of greatly expanded government, the existence of some harm to other individuals cannot be enough in itself to deny exemption or accommodation. On the other hand, harms to others certainly are grounds for limiting religious freedom in a number of circumstances. I now discuss when preventing harms to others qualifies as a “compelling governmental interest” under RFRA and similar provisions; then I turn to Establishment Clause limits on accommodations.
II. Third-Party Harms and Compelling Interests: Factors to Consider
When is a harm to others a ground for limiting religious freedom, and what factors go into that determination? Under federal law and the law of more than 30 states, the rule is that substantial restrictions on religious practices should be relieved unless the government interest in the situation is quite strong—“compelling,” in RFRA’s terms—and no less restrictive means can be adopted without significantly compromising the government’s interest. This standard, set forth in RFRA and its state law counterparts, is “a balancing test,” but “with the thumb on the scale in favor of protecting constitutional rights.”
There is no algorithm to tell us what precisely counts as a compelling interest. All that one can do is identify the most common factors and give examples of the roles they play. These various factors should be weighed to produce a balance with the thumb significantly on the side of religious freedom.
A. The Immediacy and Concentrated Nature of the Harm
It is one thing to say that a person cannot rely on religious grounds to assault another or trespass on her property. It is another thing to say that a person cannot ingest drugs at a worship service because some of the supply might be illegally trafficked and end up harming others. Both cases ultimately involve asserted harms to others, but the harms in the drug case are indirect, dependent on contingent chains of events, and diffused throughout society. In the modern state, government can regulate to prevent indirect or diffuse harms. But when application of the regulation substantially burdens religious exercise, the application should be subject to stringent questioning—certainly it should be under the RFRA standard—to ascertain that the harm will be severe and the regulation necessary to prevent it. Protecting religious freedom in these cases is relatively unproblematic because the costs of protection can be borne by the entire society, avoiding concentrated effects on any individual. Even many commentators who support significant limits on religious accommodations acknowledge that religious freedom is a “public good” and that “[t]he costs of permissive accommodations may be imposed on the public or one of its broad subsets.”
In contrast, direct, particularized harms to an individual are more likely to justify denying an exemption. James Madison, a strong defender of free exercise, referred to such harms when he said that free exercise should prevail unless it “trespass[es] on private rights” (or, he added, “the public peace”). Religious freedom gives no one the right to commit direct invasions of another’s life, liberty, or property—the historic framework of criminal or tortious acts.
B. Proximity to Core of Religious Freedom
But even actions with particularized effects on others must be protected in some circumstances, when the actions lie close to the core of religious exercise. We can see this, for example, by looking at employment disputes involving religious organizations. If no action immediately affecting another individual should ever be exempted, then the ministerial exception would be inappropriate, since it allows a religious organization to deny employment to a specific individual. Likewise, it would be inappropriate, in any non-ministerial case, to permit a religious organization to prefer members of its faith in employment, since that would affect applicants of other faiths who were disfavored. But protections for religion-based hiring by religious organizations are well established. Courts have held that RFRA requires exemption in such situations, and in Amos the Supreme Court unanimously upheld a statutory exemption against Establishment Clause challenge as applied to religious non-profits. In those situations, a religious organization’s actions that immediately affect specific individuals should nonetheless be protected because they are part of the organization’s internal governance and self-definition, which are at the core of its religious exercise.
Another way to approach this issue is to ask which persons count as “third parties,” and which by contrast stand in a position internal to the religious community in question. Hosanna-Tabor points Religion Clause jurisprudence in this direction: it gives categorical protection to a religious organization’s selection of leaders on the ground that this is “an internal church decision that affects the faith and mission of the church itself.” Ministers and would-be ministers are not third parties, but rather play or seek to play a role—a core role—in the religious institution in question. On such an internal matter, the Court deems it irrelevant that the law in question is generally applicable and that the religious organization’s action has a negative effect on the individual minister.
One could say, although Hosanna-Tabor does not do so explicitly, that the minister has implicitly consented to the organization’s power to set its internal policies concerning his employment. As the Court said in Watson v. Jones in 1872: “All who unite themselves to [a religious] body do so with an implied consent to [its] government, and are bound to submit to it.” Although this consent may sometimes be constructive, not actual, recognizing it helps support the sphere of freedom for religious organizations—what the Court has commended as “a spirit of freedom . . ., an independence from secular manipulation or control, in short, power to decide for themselves, free from state interference, matters of church government as well as those of faith and doctrine.”
This argument for religious organizational freedom applies not only to houses of worship and to employees who are members of the church. It also applies, presumptively at least, to non-members who agree to work for a religiously affiliated non-profit organization. The organization depends upon them to carry out its “faith and mission”; their loyalty to the mission is a crucial element of the exercise of religion. And they too have chosen to associate with the organization. As even scholars skeptical of accommodation have acknowledged, there is often a “reasonable expectation that employees who work for churches and religious-affiliated non-profits understand that their employers are focused on advancing a religious mission.” It is important to ensure that notice of the organization’s religious nature and policies is clear. But when it is, employees should presumptively be held to have consented, implicitly, to those policies and foundational principles, moving them from third-party to insider status.
On the other hand, employees and customers in the commercial marketplace are certainly third parties, and accordingly exemption from a generally applicable law should be more limited when they are affected. For-profit businesses differ from religious non-profits, as a general matter, for several reasons. First, non-profits that identify themselves as religiously affiliated are generally closer to the core of religious exercise than are for-profit businesses selling ordinary secular products. By their very identity, these non-profits carry out the mission of a religious community. And while religious communities have belief and worship at their core, their exercise cannot be confined to those categories: religious belief and identity have direct implications in service to others. Second, in extending further from the core of religious exercise, for-profit exemptions can affect vastly more persons: the religious non-profit sector covers perhaps 6–7 percent of jobs and wages, but the for-profit sector probably covers ten times that. The state therefore has a heightened interest in regulating the for-profit sector to ensure that all people are able to participate fully in economic life.
Moreover, the state has an increased interest in avoiding unfair commercial advantages for some market actors over others, especially when an exemption claim is less likely to be sincere—and sincerity of religious purpose can be presumed more safely with a religiously affiliated non-profit than with a commercial business. Finally, expectations are different in the two contexts: while people should certainly expect that a religiously affiliated school or social service may run on religious principles, they have less reason to expect this of the ordinary commercial business.
This does not mean that businesses cannot have serious religious interests, or that the pursuit of profit is irreconcilable with religious exercise. The Supreme Court correctly held in Hobby Lobby that for-profit corporations could “exercise religion” and therefore bring claims under RFRA. But it makes sense that there will be broader protections for religious non-profits, and narrower protections for businesses in the commercial market.
C. Severity of the Harm
Of course, a key question ultimately is the severity of the harm. Even a diffuse harm may be very serious: consider, for example, a serious threat to national security or public safety. Conversely, even when a harm is relatively individualized, it will not necessarily be significant enough to implicate a compelling governmental interest and override religious freedom. For example, consider the contraception mandate: the public health benefits of contraception are strong, but employers were not stopping employees from getting contraception. The interest behind the coverage mandate was in ensuring effective access to contraception for each female employee no matter how modest her income or resources. Contraception is often cheap and widely available, which would weaken the government’s case under the compelling-interest component of RFRA. The government’s case for a compelling interest was stronger only because some contraceptives—those preferable or even necessary in some circumstances—cost considerably more, creating a significant expense for modest-income women. I do not mean to adjudicate the government’s interest here; my only point is that facts such as these should be considered under the case-by-case analysis mandated by RFRA.
The nature and severity of the harm is also a crucial question in the growing number of cases involving conflicts between religious freedom and gay rights: small wedding vendors declining to serve same-sex weddings, Catholic adoption agencies declining to place children with same-sex couples, religious colleges applying sexual-conduct policies to faculty, staff, or students. One harm in these cases that anti-discrimination law seeks to avoid is that the protected class might lack access to economic transactions and opportunities. That harm is clearly serious, and a religious provider would receive no exemption if its refusal significantly affected access. But in most cases it does not, because there are ample alternative providers: many adoption agencies in Massachusetts willing to serve same-sex couples, and many colleges for LGBT students to attend that do not have conduct policies that would conflict with their sexuality.
Even in the for-profit sphere, some courts have been willing to exempt a small provider of services—for example, a religiously devout landlord declining to rent to an unmarried, cohabiting male-female couple—if an exemption would not “significantly imped[e] the availability of rental housing for people who are cohabiting.” By contrast, other courts have held that government has a compelling interest in preventing each and every act of discrimination, regardless of its effect on access. The Alaska Supreme Court said that the state had a compelling “transactional interest” in preventing each act of discrimination “based on irrelevant characteristics”—regardless of whether it materially impeded a cohabiting couple’s access to housing—because such discrimination “degrades individuals [and] affronts human dignity.”
The transactional/dignitary harm to same-sex couples in commercial cases can be viewed as serious: it can involve surprise and, in public settings, embarrassment or humiliation in front of others. On the other hand, the dignitary harm from the denial often occurs solely through its “communicative impact”—the impact of the message of disapproval it sends—which in other contexts cannot qualify as a justification for overriding First Amendment rights. And same-sex couples are already aware that some people around them do not approve of their relationships. Under a RFRA, I would protect for-profit objectors in a narrow set of cases: sole proprietors or small-business owners providing personal services to facilitate, in a specific way, a ceremony or relationship to which they object, in cases where there are ample alternatives and thus little effect on access. These cases include the small landlord who objects to renting to unmarried male-female cohabiting couples, the small wedding photographer who declines to provide services for a same-sex commitment ceremony, or the counselor who declines to counsel cohabiting or same-sex couples. These objectors plausibly feel the most direct personal responsibility for their contribution to others’ actions, and the sanctions imposed by anti-discrimination law threaten to drive them from their business.
Whatever courts conclude about those cases, two points should be clear. First, the relatively cautious approach to exemptions in the for-profit sphere means that anti-discrimination exemptions should not extend beyond individuals and small businesses who would otherwise have to provide services directly to facilitate marriages or relationships to which they conscientiously object. The arguments for such a carefully defined small-business exception do not justify exemption for much larger businesses or for those that have market power (for example, in lightly populated areas). Nor should we exempt the objector who refuses service in a context that has no real nexus to the behavior she opposes. Exemption may extend to providers of services specifically tied to the religious objection (the wedding photographer refusing to use her art to sanction what she considers a sinful union), but not to those who seek to avoid dealing with individuals whose unrelated behavior is considered objectionable (the restaurant refusing to provide a table to gay customers). These distinctions are worth making if a jurisdiction wants to value both religious freedom and same-sex family equality.
Second, even if the harm from refusal is deemed per se serious enough to reject exemptions in the for-profit sphere, the same should not apply to religious non-profits, which merit stronger protection. If a religious non-profit’s denial of service is unprotected even when it has no meaningful effect on access, then the organization will be severely restricted in its ability to follow its tenets and identity. As already noted, religious schools and social services are generally closer to the core of religious life than are for-profit businesses, and the element of surprise that may result from a refusal in the for-profit marketplace does not apply when employees or clients deal with a charity that is known to be religious.
We might sum up this section by noting that the seriousness of effects on others can be mitigated by two factors: notice of the religious claimant’s policy and ready alternative providers of services. In the non-profit context, publicly identified religious organizations by nature give notice of their identity, and they may give notice of specific policies as well. Assuming that notice exists, then only in those few cases where religious non-profits hold market power is there a clear argument for a compelling interest in denying exemption. With for-profit businesses, there also may be ample alternatives, but exemptions should be narrower—not nonexistent, but narrower—in part because the lack of inherent notice makes harmful surprise more likely, and in part because of the increased interest in ensuring everyone’s access to the commercial marketplace.
Finally, under RFRA, the government must show not only that the burden it has imposed on religion serves a compelling interest, but also that it does so by the least restrictive means. Hobby Lobby held that the mechanism for coverage by the insurer or third-party administrator was an available, less restrictive means. That mechanism was practicable because, by the government’s own calculations, insurance coverage of contraception saves costs to insurers on net by avoiding costs from pregnancies.
The majority was less clear on whether the option of increasing public funding of contraception would constitute an available less restrictive means: Justice Kennedy, the crucial fifth vote, expressed doubt in his separate opinion that RFRA would mandate that option. Kennedy may have been influenced by the fact that there seemed to be no chance Congress would ever pass such funding. But in many cases, the government could increase access to a good or service by increasing its subsidies or providing tax incentives to encourage manufacturers or distributors to provide it at lower cost. By these mechanisms, the government would take the impact of an employer’s religious-freedom right on a relatively small number of employees and diffuse it among the far larger taxpaying public.
The advantage of focusing on “less restrictive means” is that the government can develop such alternatives based on pragmatic considerations—and the RFRA framework encourages such solutions. Under pressure from lawsuits, the government came up with a creative mechanism to accommodate objections by religious non-profits; in hearing and deciding the Hobby Lobby case under RFRA, the Court likewise turned to this mechanism to accommodate objections by closely-held for-profits. Without RFRA’s mandate to explore means of accommodating religious objections, there would have been little or no legal pressure for the administration, or the Court, to engage in this problem solving.
III. Establishment Clause Limits
When the question is whether the Establishment Clause bars an exemption meant to protect religious exercise, the factors just discussed apply—but they should be weighed with deference to the exemption. The clause places some outside limits on how far a statutory exemption may go, but those limits should be lenient. Because exemptions are crucial to preserving religious freedom in the active state, stringent judicial policing of their permissibility is inappropriate. An exemption should not be struck down unless the direct, immediate burdens it imposes on others are clearly disproportionate to the legal burdens it removes from religious practice.
A. Reasons for a Deferential Establishment Clause Limit
It is clear that an exemption provision is not invalid simply because it singles out religious practice for protection. Two rulings decisively reject that proposition by upholding a statutory accommodation unanimously: Amos, approving Title VII’s exemption of religious organizations from liability for religious discrimination; and Cutter, affirming the provision of the Religious Land Use and Institutionalized Persons Act (RLUIPA) that protects state prisoners’ exercise of religion unless the prison can show a compelling interest in restricting it. In Amos, the Court said that “there is ample room for accommodation of religion,” that a law does not advance or sponsor religion “merely because it allows churches to advance religion,” and that “when government acts with the proper purpose of lifting a regulation that burdens the exercise of religion, we see no reason to require that the exemption comes packaged with benefits to secular entities.”
However, there are Establishment Clause limits on exemptions, and third-party harms figure in those limits. Estate of Thornton v. Caldor, Inc., for example, invalidated a statute imposing an absolute duty on employers to grant an employee’s request for his Sabbath day off. And as already noted, Cutter, while upholding RLUIPA’s prison provisions, laid out a three-part Establishment Clause test that includes whether the accommodation in question takes “adequate account of the burdens [it] may impose on nonbeneficiaries.”
The Cutter test, however, should not be a stringent one. Under it, the burden the accommodation imposes on others is not determinative: it must be weighed, if only in a rough way, against the burden the accommodation removes from sincere religious practice. For several reasons, only a great disparity between the two factors should suffice to disapprove the accommodation.
1. Theoretical/Historical Foundations
First, the theoretical and historical foundations for calling an accommodation an establishment are shaky, and they support only a modest Establishment Clause limit. Historically, exemptions of religious practice from government regulation were not typical components of establishment: exemptions were created to protect minority faiths, not the established majority. “Exemptions protect minority religions,” Douglas Laycock has shown, “and they emerged only in the wake of toleration of dissenting worship,” as part of “a political commitment to free exercise,” not to establishment.
Gedicks and Van Tassell argue that “[p]ermissive accommodations that require unbelievers and nonadherents to bear the costs of someone else’s religious practices constitute a classic Establishment Clause violation.” They point out that classic establishments “imposed legal and other burdens on dissenters and nonmembers that [they] did not impose on members.” But this analogy is weak. Historic establishments pressured dissenters to attend the favored church or required them to pay taxes for its support. Such requirements differ from regulatory exemptions in the very ways that are at issue. Compulsion to attend a church is compulsion to engage in a religious practice, something that no regulatory exemption requires. Required tax support for the favored religion removes no legal burden on that faith and thus serves no free exercise interest. By contrast, most exemptions from regulation serve free exercise interests. To cite forced worship or tax support as analogies to condemn exemptions begs the key questions.
Regulatory exemptions and compulsory tax-generated support are treated very differently in our law. The Court has said there is an especially strong, “historic and substantial,” Establishment Clause interest in preventing tax support for clergy. If regulatory exemptions were like tax support, then the clergy context would be the most problematic one for exemptions. But the law is exactly the opposite: the ministerial exemption was affirmed unanimously in Hosanna-Tabor, and within its domain it is absolute, the strongest religious-freedom exemption in American law. Clergy and worship services present the strongest context for exemption, even as they present the most questionable context for tax support. The reason is that exemptions, unlike tax support, serve interests in religious autonomy, for which clergy and worship are the core contexts.
A more pertinent historical case concerning the constitutionality of religious exemptions is the original “benefit of clergy,” the arrangement by which clerics in the medieval church were immune from civil jurisdiction—triable only in church courts—for any felonies they committed. King Henry II’s attempt to shrink this privilege and prosecute “criminous clerks” in royal courts for rapes, murders, and thefts lay at the core of his confrontation with Archbishop Thomas Becket in the mid-12th century. Although benefit of clergy changed drastically in form before the American colonies were founded, its original form can easily be seen as a feature of establishment. Unlike compelled worship or tax support, benefit of clergy involved the feature relevant to accommodations: exemption of religious actors from secular regulation when they had caused harm to others.
But treating benefit of clergy as a feature of establishment does not mean rejecting most modern exemptions, for there are multiple differences between the two. First, benefit of clergy was for the favored church (the medieval Catholic Church, then the Church of England after the Reformation). Second, it shielded wrongdoers from state jurisdiction even when there was no particularized conflict between the law in question and the demands of faith. Neither clerics nor the church presented any claim that faith or mission called them to engage in felonies. Rather, the church asserted a purely jurisdictional claim: autonomy to resolve cases in its own courts. Such a claim is strong with respect to internal matters of church governance; the ministerial exception, as affirmed in Hosanna-Tabor, essentially gives the church categorical autonomy over the selection of church leaders. But a religious organization cannot have such absolute protection in contexts where third parties are significantly affected. In those contexts, exemptions should—and the vast majority of them do—rest on the existence of a particularized conflict between the civil law and a religious claimant’s tenets or identity.
Finally, benefit of clergy blocked the government from preventing serious, direct harms to the person and property of other individuals: murder, rape, theft. No one argues today that religious freedom shields acts causing such basic harms. Exemptions today concern laws that reflect the far more extensive aims of the modern welfare state. Thus, any analogy to benefit of clergy merely returns to the problem of defining the relative limits of regulation and the countervailing right to free exercise of religion. The proper balance requires recognizing modern government’s expanded power, but not simply deferring to its assertions of what constitutes harm.
2. Deference to Legislative Judgments
Second, when the question is whether a statutory exemption is permissible, the policy of deference to government’s balancing of goals cuts in favor of the exemption. If modern regulators have leeway to define legal harms in order to pursue varying interests, then they should have leeway to protect religious freedom among those interests. It would make little sense, for example, to say that a state that recognized same-sex marriage could not simultaneously exempt religious adoption agencies or counseling organizations, in order to balance the two rights. Why is it any different if the legislature creates exemptions in response to a court decision ordering same-sex marriage than if the legislature enacts the accommodation at the time it recognizes marriage legislatively?
The issue of exemptions from newly-created rights has generated an exchange of accusations of question-begging. Before the Hobby Lobby decision, some commentators argued that exempting employers from the contraception mandate would create no legal burden on employees because RFRA meant that the mandate never gave the employees a right in the first place. Others responded, correctly I think, that such “baseline” arguments begged the question whether applying RFRA would violate the Establishment Clause by making the decision to include someone within a legal right contingent on another person’s religious exercise. But this response also begs a question: Why doesn’t the legislature that creates a legal right have discretion, in either the same statute or a separate one, to balance that right against the religious freedom of others affected by it?
The expansion of regulation in the modern state has narrowed the effective scope of the free exercise of religion, and within some range government clearly has discretion to do so. But the expanded state should likewise narrow the scope of the non-establishment rule. The government should similarly have discretion to reduce the effects that its own expansion has on religious freedom—including effects caused by the declaration of new legal harms to third parties. Otherwise, the expansion of the state’s role would be a one-way ratchet, giving government discretion to shrink free exercise, but no discretion to preserve it.
Establishment Clause review of the balance between religious accommodation and other rights should not be stringent. As Michael McConnell has observed, “when legislatures adjust the benefits and burdens of economic life among the citizens, they regularly impose more than a de minimis burden for the purpose of protecting important interests of the beneficiary class”: consider, for example, the duty of reasonable accommodation of disabilities. The legislature should have as much latitude to protect religion as it has to protect these other important values. Moreover, because “[a]ny comparison of benefits and burdens will admittedly suffer the problem of comparing apples and oranges,” the analysis cannot be highly rigorous: “The courts should be satisfied if they have examined the legislative accommodation and determined that the burden on nonbeneficiaries is not obviously disproportionate. Deference to legislative judgment is appropriate here; secular economic interests are not under-represented in the political process.”
B. Application and the Case Law
The “significantly disproportionate burdens” standard advocated here gives the best account of the Establishment Clause case law. The exemption for religion-based hiring in Amos allowed the organization to discharge an individual from his job—unquestionably a significant, individualized burden. Yet the Court unanimously upheld it because, as Justice Brennan later wrote, it “prevented potentially serious encroachments on protected religious freedoms.” Critics of accommodation concede, as they must, that exemptions for religiously affiliated non-profits are permissible even when they significantly affect identifiable individuals. Courts give—and should give—significant weight to the free exercise interests that support exemptions.
The two Supreme Court decisions invalidating accommodations are consistent with a narrow rule of invalidity. Caldor, which as already noted struck down a state law requiring private employers to give employees their Sabbath day off, involved several features that made it very likely the costs imposed on others would outweigh the burdens removed from religion. First, the statute gave employees an unqualified right regardless of the cost to employers and other employees: it reflected “an unyielding weighting in favor of Sabbath observers over all other interests.” Thus the effects on others were potentially large, and the statute showed no respect for their interests. Second, the case arose in a commercial context where, as already noted, the interest in making room for everyone, without unfair advantages for any, requires special care in the structuring of exemptions. Caldor provides little ground for striking down exemptions protecting non-profit organizations whose religious character is apparent.
The need for limits in the sphere of commercial businesses is shown by the recent federal court decision invalidating Mississippi’s statutory accommodation of objectors to LGBT rights on the ground that the accommodation violated the Equal Protection and Establishment clauses. Under both clauses, the court said that the breadth of the law made it unconstitutional. Section 5 of the act allowed any closely-held business to refuse to provide services to a same-sex wedding, regardless of the business’s size or the effect its refusal would have on same-sex couples’ access to services. Section 6 allowed any such business, again no matter how large, to require that transgender employees use the bathroom of their biological sex at birth.
Some parts of the judge’s opinion were improperly hostile to religious accommodation. But he had a fair point about the breadth of this statute in the commercial sphere. When a group of scholars, including me, proposed “marriage conscience” provisions in various states to accompany recognition of same-sex marriage, we limited the size of the businesses that would be protected, and we included an override for cases where exemption would cause a marrying couple substantial hardship. Our numerical ceiling was five employees, which was surely lower than is constitutionally necessary; but some meaningful ceiling and/or hardship override are necessary. At any rate, the Mississippi decision shows the need to be careful in drafting religious-freedom legislation in the government and commercial spheres.
A third feature of Caldor is that the burden on employees’ religion that the statute removed had been imposed not by the state, but rather by private employers. Thus, as Justice O’Connor put it, the statute “[was] not the sort of accommodation statute specifically contemplated by the Free Exercise Clause.” The strongest case for government to remove a burden on religion is when government itself has created the burden, implicating the Free Exercise Clause’s special concern for religious freedom against the government. In Caldor, with that constitutionally grounded justification absent, the statute was simply reordering interests among private employees, which made the Court more willing to ask whether the balance the statute struck was even-handed.
For the same reasons, it is misplaced to suggest, as some commentators have, that the Establishment Clause should permit only “de minimis” effects on others. The Supreme Court has adopted that standard to limit accommodation of religious employees’ practices under Title VII of the Civil Rights Act; but the analogy to most other exemption cases is entirely inapt. The de minimis standard did not interpret the Constitution; it interpreted (correctly or not) an anti-discrimination statute that does not facially require exempting employees from generally applicable work rules (unlike federal or state RFRAs, which explicitly requires exemptions). When the Court in Caldor actually described what burdens on others render an accommodation unconstitutional, it referred to an “unyielding weighting” of religious over secular interests, which is virtually the opposite of saying that a mere “de minimis” burden on a secular interest outweighs any burden on a religious interest.
Moreover, because the Title VII accommodation provision, like the statute in Caldor, does not promote the constitutionally grounded interest in preventing government-imposed burdens, the Justices may have been more willing to question whether the adjustments the provision makes among employees would be even-handed. Finally, the Title VII accommodation provision operates mostly in the context of ordinary commercial businesses. With respect to religiously affiliated organizations, by contrast, a de minimis standard is irreconcilable with well-established exemptions—like the religious-hiring protection unanimously upheld in Amos—that have significant effects on others.
The other anti-accommodation decision is Texas Monthly v. Bullock,  which struck down a sales-tax exemption for religious publications. There three justices joined a plurality opinion finding that the cost the exemption shifted to others—a higher share of tax liability if tax revenues were to remain constant—outweighed the burdens removed from religion. This weighing could be questioned, because although the incremental burden from a small tax on each sale of a religious periodical is relatively small, so is the cost shifted to any one taxpayer: Texas Monthly involved a diffuse rather than concentrated burden on others. In any event, the concurring justices objected that this rationale for invalidating the exemption was too broad. They focused on the fact that the statute favored religious messages and publications, which among other things implicated content-neutrality principles under the Free Speech and Free Press Clauses. This rationale does not affect most exemptions, because most do not involve cases of speech.
In the past, some judges and commentators have suggested that any exemption specifically for religious interests is invalid favoritism for religion. That analysis has been properly rejected: a distinctive concern for free exercise is part of our constitutional text and national tradition. It is more justifiable to define the limits of religious accommodation on the basis of significant harms that it may cause to nonconsenting third parties—and we can expect continued litigation over these questions in the future. But rules against third-party harms cannot be stringent. Since the modern state can define virtually any effect as a prima facie harm, there must be meaningful limits on these definitions as they apply to religious conduct, or else government’s expansion will crowd out religious exercise in many sectors of life. This is so both when a religious claimant seeks exemption under a RFRA or a state constitutional provision, and when an exemption enacted by the legislature is challenged as unconstitutional. Courts deciding cases, and legislatures considering statutory exemptions, should consider the principles outlined here as a framework for taking religious freedom seriously while recognizing other persons’ interests.
 See, e.g., Frederick Mark Gedicks & Rebecca G. Van Tassell, RFRA Exemptions from the Contraception Mandate: An Unconstitutional Accommodation of Religion, 49 Harv. C.R.-C.L. L. Rev. 343 (2014), available at http://harvardcrcl.org/wp-content/uploads/2011/09/Gedicks-FINAL-website-edits-3-25-2014.pdf; Micah Schwartzman et al., Hobby Lobby and the Establishment Clause, Part III: Reconciling Amos and Cutter, Balkinization (Dec. 9, 2013), available at http://balkin.blogspot.com/2013/12/hobby-lobby-and-establishment-clause_9.html, archived at http://perma.cc/VWZ6-JEA6.
 42 U.S.C. §§ 2000bb-2000bb-4.
 134 S. Ct. 2751 (2014).
 For citations, see Douglas Laycock, Religious Liberty and the Culture Wars, 2014 U. Ill. L. Rev. 839, 845 n.26 (listing 19 state RFRAs); id. at 844 n.22 (listing 12 state constitutional provisions interpreted to require exemptions).
 42 U.S.C. § 2000bb-1.
 Cutter v. Wilkinson, 544 U.S. 709, 720 (2005).
 Eugene Volokh, 5C. RFRA Strict Scrutiny: The Interest in Protecting Newly Created Private Rights, Volokh Conspiracy (Dec. 6 2013), http://volokh.com/2013/12/06/5c-rfra-strict-scrutiny-interest-protecting-newly-created-private-rights/, archived at http://perma.cc/MU4V-JMEC.
 Pierre Bayle, Philosophical Commentary on These Words of Christ: Compel Them To Come In, in Pierre Bayle’s Philosophical Commentary: A Modern Translation and Critical Interpretation 7, 167 (Amie Godman Tannenbaum trans., 1987).
 Thomas Jefferson, Notes on the State of Virginia 159 (1784) (William Peden ed., 1954).
 See Eugene Volokh, A Common-Law Model for Religious Exemptions, 46 UCLA L. Rev. 1465, 1521–26 (1999).
 132 S. Ct. 694, 696 (2012).
 Hobby Lobby, 134 S. Ct. at 2781 n.37.
 See Eugene Volokh, 3B. Would Granting an Exemption from the Employer Mandate Violate the Establishment Clause?, Volokh Conspiracy (Dec. 4 2013) http://volokh.com/2013/12/04/3b-granting-exemption-employer-mandate-violate-establishment-clause/, archived at http://perma.cc/W3N-ZB25.
 483 U.S. 327 (1987).
 The very point of the Religious Land Use and Institutionalized Persons Act (RLUIPA), 42 U.S.C. §§ 2000cc to 2000cc-5, is to protect such activities. See also Chosen 300 Ministries v. City of Philadelphia, 2012 WL 3235317 (E.D. Pa. 2012) (entering preliminary injunction for feeding ministry under Pennsylvania RFRA).
 374 U.S. 398, 399–400 (1963). See, e.g., Volokh, A Common-Law Model, supra note 10, at 1513–14 & 1513 n.154 (“Unemployment compensation is generally experience-rated, so an employer’s unemployment tax payments are tied to the number of claims the employer has had to pay out.”).
 Douglas Laycock, The Religious Exemptions Debate, 11 Rutgers J.L. & Relig. 139, 151–52 (2009).
 These factors overlap with those in Christopher C. Lund, Religious Exemptions, Third-Party Harms, and the Establishment Clause, 91 Notre Dame L. Rev. 1375, 1376-81 (2016).
 Frederick Mark Gedicks and Andrew Koppelman, The Costs of the Public Good of Religion Should Be Borne by the Public, 67 Vand. L. Rev. En Banc 185, 187 (2014), https://www.vanderbiltlawreview.org/wp-content/uploads/sites/89/2014/06/Gedicks-Koppelman-Response.pdf; Alan Brownstein, Taking Free Exercise Rights Seriously, 57 Case W. Res. L. Rev. 55, 129, 130 (2006) (arguing that “[c]onstitutional guarantees, such as freedom of speech or freedom of religion, are public, political goods” and “the state is often required to incur expenses in order to allow other rights such as freedom of speech to be exercised”).
 Letter from James Madison to Edward Livingston (July 10, 1822), reprinted in 9 The Writings of James Madison 98, 100 (Gaillard Hunt ed. 1910).
 The stronger protection for claims at the core of religious exercise applies to individual as well as organizational claims. Protection will be more absolute for an individual’s ability to attend church, or receive a religious education, than for the ability to follow her religion in for-profit employment—although the latter still must receive some protection in the balance.
 See, e.g., Porth v. Roman Catholic Dioc. of Kalamazoo, 532 N.W.2d 195, 199–200 (Mich. Ct. App. 1995) (applying RFRA, when it still applied to states, to exempt a Catholic school from a state religious-discrimination suit by a fifth-grade teacher).
 Hosanna-Tabor, 132 S. Ct. at 707; see id. at 706 (referring to “the internal governance of the church” and “a religious group’s right to shape its own faith and mission through its appointments”).
 Watson v. Jones, 80 U.S. 679, 729 (1872). For perceptive discussion of the “implied consent” doctrine, see Michael A. Helfand, Religious Institutionalism, Implied Consent, and the Value of Voluntarism, 88 S. Cal. L. Rev. 539 (2015).
 Kedroff v. St. Nicholas Cathedral, 344 U.S. 94, 116 (1952).
 Schwartzman et al., supra note 1 (emphasis added).
 See, e.g., Thomas C. Berg, Partly Acculturated Religious Activity: A Case for Accommodating Religious Non-Profits, 91 Notre Dame. L. Rev. 1341, 1371 (2016) (“Religious organizations that do not have explicit religious elements in their programs should make it reasonably apparent to employees—through the employee handbook or contract or some other means—that religious norms may apply.”).
 To speak of situations closer to or further from the core of free exercise is not to deny that religion plays a role even in the non-core situations. Rather, it recognizes, in the words of Elder Lance Wickman of the Mormon Church, that:
in a pluralistic nation where religious people and institutions find themselves competing for influence with others having much different priorities and interests, we . . . have to prioritize. Defenders of religious freedom have to decide what is closer to the essential core of religious freedom and what is more peripheral. To do otherwise risks weakening our defense of what is essential. If everything that could even loosely be considered “religious” is treated as equally important, then effectively nothing religious is important.
Lance B. Wickman, Promoting Religious Freedom in a Secular Age: Fundamental Principles, Practical Priorities, and Fairness for All (July 7, 2016), http://www.mormonnewsroom.org/article/promoting-religious-freedom-secular-age-fundamental-principles-practical-priorities-fairness-for-all.
 For the derivation of this number, see Thomas C. Berg, Religious Accommodation and the Welfare State, 38 Harv. J.L. & Gender 103, 127 n.23 (2015).
 See, e.g., Robert K. Vischer, Do For-Profit Businesses Have Free Exercise Rights?, 21 J. Contemp. Legal Issues 369, 391 (2013) (“The primary concern . . . is that for-profit corporations are so central to our ability to participate in modern life, including our ability to earn a livelihood. They are inescapable conduits for many goods deemed fundamental to our modern existence.”).
These distinctions between for-profits and non-profits are certainly not absolute: some businesses (usually small ones) have a distinctively religious character personal to their owners, while some non-profits, such as hospital chains, act much like large commercial businesses. Non-profit versus for-profit status therefore should be only one among several factors, but it is generally a useful rule of thumb.
 See, e.g., Amos, 483 U.S. at 344 (Brennan, J., concurring) (“The fact that an operation is not organized as a profit-making commercial enterprise makes colorable a claim that it is not purely secular in orientation.”).
 Hobby Lobby, 134 S. Ct. at 2769–72 (noting that corporations often reflect the outlook of their owners, that many for-profit corporations follow moral norms and objectives, and that “there is no apparent reason why they may not further religious objectives as well”).
 Id. at 2799 (Ginsburg, J., dissenting) (“The coverage helps safeguard the health of women for whom pregnancy may be hazardous, even life threatening. And the mandate secures benefits wholly unrelated to pregnancy, preventing certain cancers, menstrual disorders, and pelvic pain.”) (citations omitted).
 See, e.g., id. at 2789.
 See, e.g., Megan McArdle, Sell Birth Control Over-the-Counter, Bloomberg View, Sept. 10, 2014, http://www.bloombergview.com/articles/2014-09-10/sell-birth-control-over-the-counter, archived at http://perma.cc/M6ED-2WHQ (“Generic birth-control pills are a cheap, regular expense used by many millions of people, exactly the sort of thing that insurance is not designed for.”).
 See, e.g., Hobby Lobby, 134 S. Ct. at 2800 (Ginsburg, J., dissenting) (the cost of an IUD, which is “significantly more effective” than other methods, is “nearly equivalent to a month’s full-time pay for workers earning the minimum wage”); McArdle, supra note 36 (noting the higher costs of IUDs).
 See Dale Carpenter, Let Catholics Discriminate, Metro Weekly (Mar. 29, 2006), http://www.metroweekly.com/2006/03/let-catholics-discriminate/ (“Gay couples could still adopt through dozens of other private agencies or through the state child-welfare services department itself, which places most adoptions in the state.”).
 Attorney Gen. v. Desilets, 636 N.E.2d 233, 241 (Mass. 1994) (applying compelling-interest test under state constitution).
 Swanner v. Anchorage Equal Rights Com’n, 874 P.2d 274, 283 (Alaska 1994).
 See, e.g., Texas v. Johnson, 491 U.S. 397, 406 (1989) (“expressive conduct” such as flag-burning may not be prohibited when the law is “‘directed at the communicative nature of [the] conduct’”) (emphasis and quotation omitted); accord United States v. Eichman, 496 U.S. 310, 317 (1990) (government action may not suppress First Amendment conduct “out of concern for its likely communicative impact”).
 See, e.g., Elane Photography LLC v. Willock, 309 P.3d 53 (N.M. 2013).
 See, e.g., In the Matter of: Melissa Elaine Klein, dba Sweetcakes by Melissa, 2015 WL 4868796 (Ore. Bur. Lab. & Ind. 2015) ($135,000 in emotional-distress damages imposed on cakeshop); Elane Photography, 309 P.3d 53 ($6,600 in attorney’s fees imposed on photographer, with no proof of actual damages). The Sweetcakes shop received contributions to defray their costs, and the plaintiff couple in Elane Photography waived the award, see 309 P.3d at 60. But there is no guarantee the same thing will happen in subsequent cases.
 This limit on a for-profit exemption should not apply when the law in question is not neutral or generally applicable: that is, when it singles out religious objections for regulation, like the rules imposed on pharmacies in Stormans, Inc. v. Wiesman, 794 F.3d 1064 (9th Cir. 2015), cert. denied, 136 S. Ct. 2433 (2016). Even with respect to large businesses, there is no constitutional justification for targeting religious reasons for denying service while allowing multiple other reasons, as the state did in Stormans. See Stormans, Inc. v. Selecky, 844 F. Supp. 2d 1172 (W.D. Wash. 2012) (documenting the selective regulation and the targeting of religion).
 Because a small number of states have already begun to require for-profit businesses to cover abortions, it is worth noting that protection of for-profits ought to be broader in that context than in the gay-rights context, for two reasons. First, objections to facilitating abortion unquestionably go only to a particular procedure, while broad objections to serving same-sex couples may go beyond a particular ceremony or activity and become objections to LGBT customers as persons—for example, refusing not just to provide services for a wedding, but to serve a same-sex couple in a restaurant. Second, the pattern of protections in federal and state law for abortion objectors has been uniquely strong, covering a wide range of health-care providers, even large for-profit entities like health insurers—and reflecting the judgment that it is an especially serious burden on conscience to require a person to assist in what he believes to be the unjustified killing of a human being. For documentation of the strength of abortion-conscience protections, see Mark L. Rienzi, The Constitutional Right Not to Kill, 62 Emory L.J. 121, 147–52 (2012); Brief of Democrats for Life of America and Bart Stupak as Amici Curiae in Support of Hobby Lobby and Conestoga et al., Burwell v. Hobby Lobby Stores, Inc., 134 S. Ct. 2751 (2014), at 7–13, available at http://www.americanbar.org/content/dam/aba/publications/supreme_court_preview/briefs-v3/13-354-13-356_amcu_dfla.authcheckdam.pdf.
 Hobby Lobby, 134 S. Ct. at 2759.
 See id. at 2782 n.38.
 See id. at 2786 (Kennedy, J., concurring).
 See, e.g., Korte v. Sebelius, 735 F.3d 654, 686 (7th Cir. 2013) (arguing that “[t]he government can provide a ‘public option’ for contraception insurance; it can give tax incentives to contraception suppliers to provide these medications and services at no cost to consumers; it can give tax incentives to consumers of contraception and sterilization services,” and that “[n]o doubt there are other options”).
 See, e.g., Brownstein, supra note 20, at 128–29 (discussing means of diffusing effects among larger public).
 Amos, 483 U.S. at 339.
 Cutter, 544 U.S. at 726.
 Amos, 483 U.S. at 337–38 (emphasis in original).
 472 U.S. 703 (1985).
 Id. at 708–10.
 Cutter, 544 U.S. at 720.
 Douglas Laycock, Regulatory Exemptions of Religious Behavior and the Original Understanding of the Establishment Clause, 81 Notre Dame L. Rev. 1793, 1796, 1803 (2006); accord Michael W. McConnell, The Origins and Historical Understanding of Free Exercise of Religion, 103 Harv. L. Rev. 1409, 1511 (1990) (“There is no substantial evidence that [religious] exemptions were considered constitutionally questionable.”).
 Gedicks & Van Tassell, supra note 1, at 363.
 Id. at 362 (citing Michael W. McConnell, Establishment and Disestablishment at the Founding, Part I: Establishment of Religion, 44 Wm. & Mary L. Rev. 2105, 2144–46 (2003)).
 McConnell, supra note 59, at 2144–46.
 Locke v. Davey, 540 U.S. 712, 725 (2004); id. at 722 (“[W]e can think of few areas in which a State’s antiestablishment interests come more into play.”).
 See, e.g., George W. Dalzell, Benefit of Clergy in America & Related Matters 9–15 (1955); Theodore F.T. Plucknett, A Concise History of the Common Law 439–41 (5th ed. 1956).
 Harold J. Berman, Law and Revolution: The Formation of the Western Legal Tradition 255–64 (1983); Plucknett, supra note 62, at 439.
 Dalzell, supra note 62, at 16–23; Plucknett, supra note 62, at 441.
 For example, an Indiana court in 1820 rejected a convicted murderer’s claim to a reduced sentence under benefit of clergy, saying: “The benefit of clergy . . . originated with that of sanctuary in the gloomy days of popery. . . . The statutes of England on the subject are local to that kingdom . . . and are certainly not adopted as the laws of our country.” Dalzell, supra note 62, at 238 (citing Fuller v. State, 1 Blackf. 66).
 See id. at 19 (noting that after the Reformation, “[English]-born Catholic priests returning from abroad . . . were hanged without benefit of clergy” unless they took “an oath [renouncing papal loyalty] to which they could not possible subscribe”).
 The three features of medieval benefit of clergy—denominational favoritism, no particularized burden removed from religion, and permitting serious direct harms to others—are the indicia in the Cutter test for an impermissible accommodation. See Cutter, 544 U.S. at 720.
 Kevin C. Walsh, A Baseline Problem for the “Burden on Employees” Argument Against RFRA-Based Exemptions from the Contraceptives Mandate, Mirror of Justice (Jan. 17, 2014), http://mirrorofjustice.blogs.com/mirrorofjustice/ 2014/01/a-baseline-problem-for-the-burden-on-employees-argument-against-rfra-basedexemptions-from-the-contr.html. For a somewhat analogous argument, see Marc DeGirolami, On the Claim that Exemptions from the Mandate Violate the Establishment Clause, Mirror of Justice (Dec. 5, 2013), http://mirrorofjustice.blogs.com/mirrorofjustice/2013/12/exemptions-from-the-mandate-do-not-violate-the-establishment-clause.html.
 Frederick Mark Gedicks & Andrew Koppelman, Invisible Women: Why an Exemption for Hobby Lobby Would Violate the Establishment Clause, 67 Vand. L. Rev. En Banc 51, 59-60 (2014), available at https://www.vanderbiltlawreview.org/wp-content/uploads/sites/89/2014/03/Gedicks-and-Koppelman_Invisible-Women.pdf.
 Michael W. McConnell, Accommodation of Religion: An Update and a Response to the Critics, 60 Geo. Wash. L. Rev. 685, 704 (1992).
 Id. at 705.
 Tex. Monthly, Inc. v. Bullock, 489 U.S. 1, 18–19 n.8 (1989) (Brennan, J., plurality opinion) (citing Amos, 483 U.S. 327).
 See, e.g., Gedicks and Van Tassell, supra note 1, at 368 (acknowledging that the religious-hiring exemption “created a substantial burden [on an employee] where none previously existed”).
 Caldor, 472 U.S. at 709–10.
 See supra notes 29-33 and accompanying text.
 Even Professors Schwartzman, Schragger, and Tebbe—skeptics of accommodation—at one point suggest only that non-profit exemptions raise establishment issues in “special circumstances” such “as where a religious non-profit (e.g., a hospital) monopolizes a local market.” Schwartzman et al., supra note 1.
 Barber v. Bryant, ___ F. Supp. 3d, 2016 WL 3562647 (S.D. Miss. June 30, 2016).
 Id. at *18-*23 (equal protection); id. at *31-32 (Establishment Clause).
 For the latest version, from fall 2013, see Thomas Berg, Archive: Memos/Letters on Religious Liberty and Same-Sex Marriage, Mirror of Justice (Aug. 2, 2009), http://mirrorofjustice.blogs.com/mirrorofjustice/2009/08/memosletters-on-religious-liberty-and-samesex-marriage.html.
 Caldor, 472 U.S. at 712 (O’Connor, J., concurring).
 See, e.g., Nelson Tebbe et al., How Much May Religious Accommodations Burden Others?, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2811815 (forthcoming in Elizabeth Sepper et al., Law, Religion, and Health in the United States (Cambridge Univ. Press, forthcoming 2017)).
 Trans World Airlines v. Hardison, 432 U.S. 63, 84 (1977) (“To require TWA to bear more than a de minimis cost in order to give [a Sabbatarian employee] Saturdays off is an undue hardship.”).
 Caldor, 472 U.S. at 710.
 489 U.S. 1 (1989).
 Id. at 14–20 (Brennan, J., plurality opinion).
 See id. at 25–26 (White, J., concurring in the judgment) (concluding that “the proper basis for reversing the judgment below” was that exemption violated Free Press Clause by discriminating among publications based on content); id. at 28–29, 27 (Blackmun, J., concurring in the judgment) (concluding that exempting only religious publications unconstitutionally gave “preferential support for the communication of religious messages,” but criticizing Brennan’s broader rationale for “subordinating the Free Exercise value”).
 See, e.g., City of Boerne v. Flores, 521 U.S. 507, 536-37 (1997) (Stevens, J., concurring); Mark Tushnet, “Of Church and State and the Supreme Court”: Kurland Revisited, 1989 Sup. Ct. Rev. 373.