Applying Strict Scrutiny: An Empirical Analysis of Free Exercise CasesBy Caleb C. Wolanek & Heidi Liu
78 Mont. L. Rev. 275
Strict scrutiny and the free exercise of religion have had an uneasy relationship in American jurisprudence. In this Article, we trace the history of strict scrutiny in free exercise cases and outline how it applies today. Then, using a unique dataset of cases from a 25-year period, we detail the characteristics of these cases. Finally, we discuss the implications for future cases. Our research indicates that even though claimants currently win a large percentage of cases, those victories might not be durable.
Strict scrutiny is relatively simple in theory: it is a balancing test with a thumb on the scales. The balancing aspect asks whether the asserted government interest is so compelling as to demand that individual liberty take a back seat. If the government interest is strong enough, the next question is whether the government could achieve its interest without burdening liberty as much. If not, the regulation survives.
Yet applying this test raises complex issues. What is a compelling interest? Just how compelling, and how specific, must that interest be? At what point does accommodation become so difficult to implement, or so frustrate the government’s interests, that no accommodation is required? In Employment Division v. Smith, the Supreme Court said judges could not resolve these questions when religious liberty was at issue. The Court was concerned that attempting to resolve them would create a “private right to ignore generally applicable laws,” and it therefore held that free exercise cases did not receive strict scrutiny. But within a few years, strict scrutiny was again required in a large number of cases.
This Article examines whether reinstating strict scrutiny created a “private right to ignore generally applicable laws,” or whether that perennial concern is exaggerated. Put differently, we ask when the government can prevail against religious liberty claimants in a strict scrutiny regime. This Article answers these questions in a comprehensive, empirical manner. Using an original data set of 264 federal cases, we coded each case from 1990 to 2015 that applied strict scrutiny and statistically analyzed the entire set. Ultimately, strict scrutiny is strict but complicated.
After this Introduction, Parts II and III recount the history of strict scrutiny in the free exercise context and describe how courts apply strict scrutiny. Those sections place the empirical results in context. Using an original dataset of cases, Part IV empirically analyzes strict scrutiny free exercise cases. Part V concludes.
II. A Brief Historical Overview
Religious liberty was important to many Framers, so it is no surprise that they protected the “free exercise of religion.” But “free exercise” is not completely self-defining. Despite best efforts, scholars and judges alike have found its original meaning elusive. More specifically, although it is commonly accepted that “targeting religious beliefs as such is never permissible,” there is great disagreement about whether (and to what extent) the Constitution protects the right to act on those beliefs. In Reynolds v. United States, the Court held that the Constitution protected beliefs, but not practices. Otherwise, the Court feared, one might “become a law unto himself.” But this strict distinction between belief and practice is both dangerous and impractical— religious belief often involves action. We are therefore fortunate the Court never carried the distinction to its logical conclusion that religious actions are not protected. But at the same time, Reynolds’s central concern is valid: some restriction on religious practice is needed. At some point, the law must separate permissible from impermissible religious action. This is true despite the text of the Free Exercise Clause; our legal system has not adopted Justice Hugo Black’s First Amendment absolutism.
When religious liberty cases re-emerged some years after Reynolds, the Court (consistent with Reynolds’s concerns) applied a relatively low level of scrutiny. In Braunfeld v. Brown, for instance, a plurality held that because Sunday closing laws were generally applicable (and not covertly aimed at religion), were within state power, and were reasonable, they did not violate Jewish shopkeepers’ right to religious freedom. This was true even though the laws required the shopkeepers to close on both Saturday (for the Sabbath) and Sunday (to comply with the statute). Quite unlike Reynolds, however, the Court’s opinion in Cantwell v. Connecticut said that government could not “unduly” burden religious activity. Although government could regulate the time, place, and manner of preaching, it could not “wholly deny the right to preach” or condition religious speech on an individualized assessment of its content. To be sure, Cantwell overlaps significantly with the Free Speech Clause, but the Court itself reached its decision under the Free Exercise Clause.
Justice Brennan dissented in Braunfeld, arguing that strict scrutiny should apply because religion is a fundamental right. His wish was granted in 1963, when Sherbert v. Verner said religious burdens had to be justified by reference to a “paramount interest” advanced via the least restrictive means—that is, under a strict scrutiny standard. The height of Sherbert-style strict scrutiny was Wisconsin v. Yoder, where the Court exempted the Old Order Amish from compulsory education laws. To be sure, religious claimants did not always win during the so-called “Sherbert-era.” Instead, it is generally accepted that courts actually applied something akin to intermediate scrutiny. The Court also exempted two significant sectors (prisons and the military) from strict scrutiny altogether. Still, from a doctrinal standpoint, strict scrutiny was required.
But in 1990, the Court in Employment Division v. Smith said that heightened scrutiny—whether strict or intermediate—was not required when a neutral and generally applicable law burdened religious exercise. The majority’s core concern echoed Reynolds: requiring the government to demonstrate a compelling interest advanced through the least restrictive means would permit each citizen “to become a law unto himself.” Given America’s religious pluralism, requiring exemptions “would be courting anarchy.” To top it off, the Court said weighing government interests against religious liberty was a task ill-suited for judges. The Court thus swept Sherbert aside, stating it only applied when there was an “individualized . . . assessment” of religious conduct.
Smith was not the final word, however. Both Congress and the Court have since required judges to apply strict scrutiny in a significant number of cases. Although Congress cannot “overturn” Smith, it created statutory protections for religious liberty. Congressional action first came through the almost unanimously enacted Religious Freedom Restoration Act of 1993 (RFRA), which expressly brought back the rule articulated in Sherbert v. Verner and Wisconsin v. Yoder. RFRA still applies to the federal government, and many states have adopted similar statutes. In 1997, City of Boerne v. Flores struck down the federal RFRA as applied to the states, holding it exceeded Congress’ power under Section 5 of the Fourteenth Amendment. Congress responded by unanimously enacting the Religious Land Use and Institutionalized Persons Act of 2000 (RLUIPA). RLUIPA requires states to apply strict scrutiny if a land use or institutional regulation burdens religious beliefs. And to avoid a repeat of Boerne, Congress passed RLUIPA under both the Commerce Clause and the Spending Clause. Though RLUIPA is narrower than RFRA, the two statutes together require strict scrutiny in a large number of cases.
Unlike Congress, the Court actually can change how Smith operates. And in 1993, the Court held that Smith only applies if the challenged regulation is both neutral and generally applicable. This means judges must look closely at the statute to uncover any religious discrimination. If discrimination is found, strict scrutiny applies. Even after Smith then, strict scrutiny exists in many (if not most) religious liberty cases. How courts apply that doctrine is our next inquiry.
III. Strict Scrutiny Applied
Religious claimants prevail under the Free Exercise Clause if they demonstrate that there is a substantial burden on their religious beliefs, if they can invoke strict scrutiny, and if the government cannot demonstrate both a compelling interest and least restrictive means. Conversely, the government wins if it shows that it has a compelling government interest furthered through the least restrictive means, or if the claimant cannot demonstrate either a substantial burden or grounds (statutory or otherwise) for applying strict scrutiny.
A. Substantial Burden
The Free Exercise Clause only bars government actions that coerce someone into violating her sincere religious beliefs. For better or for worse, it is not enough for government action to make religious exercise more difficult—such as when the government builds a road through a National Forest, incidentally making it more difficult for Native Americans to worship there. There must instead be a forced choice between (1) exercising one’s religion and (2) avoiding punishment or securing government benefits. For example, in Sherbert v. Verner, there was a forced choice between working on Saturday and receiving unemployment benefits. In Wisconsin v. Yoder, the choice was between preserving Amish culture and paying a $5 fine. For a Jewish inmate, the choice might be between keeping Kosher (but starving) and eating unclean foods.
Moreover, this burden must be “substantial.” But which half of the choice must be substantial: the hindrance to religious free exercise or the penalty? Even before Smith, Professor Ira Lupu noted this distinction, but the Court has not fully resolved the issue. Sometimes the Court looks at the spiritual harm that comes after one violates a religious commandment, just as the Court did in Wisconsin v. Yoder. But other times, it appears to look at the penalty imposed or the benefit withheld. For example, in Hernandez v. Commissioner, the Church of Scientology argued it should receive tax deductions for its “auditing” fees because paying taxes made its ministry economically more difficult. Under a Yoder-style approach, the Court would have briskly stated that because there was no theological objection, there was no burden on free exercise. But the Court instead went on to consider whether the tax actually pressured the Church or its members by reducing the amount of funds available for religious use. It suggested that the answer was no, and it affirmed that principle three years later in Jimmy Swaggart Ministries v. Board of Equalization: “to the extent that imposition of a generally applicable tax merely decreases the amount of money appellant has to spend on its religious activities, any such burden is not constitutionally significant.” This rule makes clear that the burden is pressure imposed via the decrease in funds—not the religious consequences incurred by paying the tax.
Yoder and Hernandez present distinct approaches. Frustratingly, however, the Court often cites both the Yoder and the Hernandez conceptions of substantial burden. Doctrinally, the better approach is to follow Hernandez and Jimmy Swaggart by looking at the cost of
noncompliance. This means a law might so impede religious exercise that it violates the Constitution even if that religious belief is not central to or mandated by the religion. This is what several lower courts have done, and for good reason. First, Lyng v. Nw. Indian Cemetery Protective Ass’n requires that there be some element of coercion. If there is no penalty for noncompliance, there is no coercion. Second, following Yoder’s approach requires looking into religious beliefs. Because even idiosyncratic religious beliefs are protected, this inquiry is doomed from the start. Quite simply, the Court is incompetent to judge religious beliefs. Third, both Lyng and RLUIPA say that religious beliefs need not be “central” to one’s religion for strict scrutiny to apply. If non-central religious beliefs are protected, then the importance of the religious belief drops out. To maintain the distinction of “substantial” from “unsubstantial” burdens, the penalty must therefore be what is substantial. But demonstrating a substantial burden is only one step on the path to victory for religious claimants. They must somehow invoke strict scrutiny—the inquiry to which we turn next.
B. Strict Scrutiny “Hook”
There are two ways to invoke strict scrutiny. First, either RFRA or RLUIPA might apply. Second, following Church of the Lukumi Babalu Aye, Inc. v. City of Hialeah, the government actor might make an individualized determination of a religious claim, or otherwise fail to have a neutral and generally applicable policy. Courts generally apply RFRA or RLUIPA before looking at neutrality or generality. This is because “if a case can be decided on either of two grounds, one involving a constitutional question, the other a question of statutory construction or general law, the Court will decide only the latter.” We follow the same sequence.
1. RFRA and RLUIPA
If RFRA or RLUIPA applies, strict scrutiny is available. RFRA applies to all federal government actions, and it is relatively straightforward: if there is a substantial burden on religion, the government must demonstrate a compelling governmental interest furthered via the least restrictive means.
RLUIPA applies the same way as RFRA in most cases, but it only extends to institutions (like prisons) and to land use regulations (such as zoning). RLUIPA also says governments may not “treat a religious assembly or institution on less than equal terms with a nonreligious assembly or institution.” The statute’s construction makes clear that a substantial burden is not required in such cases; discriminatory treatment is enough. The crucial inquiry, then, is whether a religious institution has been the victim of unequal treatment. If there is unequal treatment, most circuits hold the regulation invalid per se, but the Eleventh Circuit requires strict scrutiny.
2. Neutrality and Generality
Smith repeatedly emphasized that the laws challenged in that case were neutral and generally applicable. Lukumi made this into a prerequisite (or at least confirmed its existence as such): if a law fails to be neutral or generally applicable, it must undergo strict scrutiny. Yet courts cannot merely examine the text of a given law to determine neutrality and generality; a lawmaker with any degree of political savvy could craft a facially neutral law that in practice only applied to religious adherents. So under Lukumi, a law lacks neutrality if its “object,” whether explicitly or implicitly, “is to infringe upon or restrict practices because of their religious motivation.” General applicability refers to the law’s breadth: one that only restricts religiously motivated conduct is not generally applicable. Both of these concerns are about “religious gerrymandering.” This does not mean that governments must write perfectly tailored laws, but that substantial over-breadth or under-breadth is evidence that the law is actually targeted at religion.
For instance, Lukumi struck down local ordinances that targeted Santería, a religion that requires animal sacrifice at important religious events. The ordinances prohibited “unnecessary” animal killings—but only when in a ceremony and when the purpose was not consumption. Thus, they permitted activities like deer hunting and kosher slaughter. The Court deemed concerns about public health and animal cruelty overblown because the law applied even if the sacrifices were performed under sanitary conditions and because the city could have simply prohibited particular methods of slaughter. This indicated non-neutrality. Similarly, the ordinances were not generally applicable because they failed to prohibit behavior that implicated the government’s concerns. For instance, they did not prohibit hunters from bringing their trophies home. In the end, the Court concluded the ordinances were actually aimed at religion—and at Santería in particular.
After Lukumi, “courts have taken a broad view of what it means for a law to fail the test of neutrality and general applicability.” In Fraternal Order of Police Newark Lodge No. 12 v. City of Newark, a Muslim officer wanted to wear a short beard, but the police department had a rule against facial hair. The department granted medical exemptions, but it denied the officer’s request for a religious one. This failed under Lukumi. Then-Judge Alito wrote: “when the government makes a value judgment in favor of secular motivations, but not religious motivations, the government’s actions must survive heightened scrutiny.” It was discriminatory enough to simply not give religious justifications the same recognition as secular reasons. The policy failed heightened scrutiny for similar reasons: the secular exemption demonstrated there was not a compelling interest in having clean-shaven officers.
C. Compelling Government Interest
Once strict scrutiny is implicated, courts look for a compelling government interest. If this requirement “really means what it says,” then interests are compelling only when they are “of the highest order.” Classic examples include preventing “threat[s] to public safety, peace or order.” But it is not enough to cite a general interest. The government must instead demonstrate how its interest would be furthered in that instance. And the policy’s breadth indicates, at least in part, whether the interest is compelling: “a law cannot be regarded as protecting an interest ‘of the highest order’ . . . when it leaves appreciable damage to that supposedly vital interest unprohibited.” This is a question of law, but it heavily depends on the facts of the individual case.
Numerous articles have argued that Sherbert-era opinions repeatedly diluted the meaning of “compelling government interest” such that strict scrutiny appeared more like intermediate scrutiny. This can make it difficult to predict which interests are actually compelling. RFRA directly references Sherbert, which implies a narrow definition of “compelling interests.” But Lukumi makes no such reference, and the Sherbert-era opinions that arguably weakened the meaning of “compelling” are still good law. In short, then, compelling government interests are required, but it is sometimes unclear what that means. The empirical analysis in Part IV of this Article helps clarify this ambiguity.
D. Least Restrictive Means
At this point in the analysis, the question is not whether the government may burden religious exercise. It may. The question is whether the proposed policy unduly burdens religious exercise. This “exceptionally demanding” comparative approach says that if there is a way to achieve the government’s interest without imposing as great a religious burden, the chosen policy is not the least restrictive alternative. Put differently, the government cannot “burn the house to roast the pig.”
RFRA and RLUIPA clearly invoke the least restrictive means requirement. And despite some confusion on this point, this is the constitutional requirement. On occasion, the Court has said the Constitution imposes a “narrow tailoring” requirement. But as Justice Ginsburg noted in her Burwell v. Hobby Lobby Stores, Inc. dissent, pre-Smith cases did invoke the least restrictive means requirement, sometimes by name. The Hobby Lobby majority was agnostic about this point for strategic—not doctrinal—reasons: to show RFRA protected corporations, it argued that “[b]y enacting RFRA, Congress went far beyond what this Court has held is constitutionally required.” The (purported) move from narrow tailoring to least restrictive means was given as an example of this shift, but it was an example the majority was happy to give up if proven incorrect. And when Lukumi struck down the city’s ordinance because the asserted policy goals “could be achieved by narrower ordinances that burdened religion to a far lesser degree,” the Court was requiring that the city choose the least restrictive means—despite the mention of narrow tailoring.
IV. Empirical Analysis
This Article rides the coattails of Professor Winkler’s pivotal study on strict scrutiny, which argued that strict scrutiny is not “strict in theory, fatal in fact.” It builds on that study by asking how the government is able to survive strict scrutiny. This dovetails with the discussion in Part III by describing how courts actually behave (and not merely what they claim to do). But this is not an attempt to determine whether Smith made matters better or worse. That requires analyzing cases before or after Smith. Our analysis is instead a snapshot that seeks to understand what it takes for the government to succeed on a free exercise claim.
To begin, we compiled a list of published federal cases from 1990 to 2015. The search term for these cases was intentionally broad, as the goal was to find every case that mentioned both strict scrutiny and free exercise. Running this search term in WestlawNext produced a list of 1,578 cases.
Next, we read each case to determine whether it actually applied strict scrutiny in the free exercise context. For example, if a case applied strict scrutiny in a free speech case and then mentioned that the plaintiff’s free exercise claim was irrelevant, the case was set aside. If the court directly considered the free exercise claim but found there was no substantial burden on the plaintiff’s religious beliefs, the case was also set aside (although we briefly consider trends within these cases in Part IV.A). And if the court found there was no strict scrutiny “hook” and therefore applied rational basis under Smith, that case was not considered any further. In each such instance, the case was simply beyond this Article’s scope, which is only concerned with how strict scrutiny applies to free exercise cases.
This winnowing process reduced the number of cases to 264. These cases were coded to quantify the following information: date; District, Circuit, or Supreme Court; geographic region (by Circuit Court); procedural posture; the strict scrutiny “hook” (if indicated); the asserted government interest; the outcome of the case; and the court’s reasoning.
Table 1: Distribution by Court
*This is the circuit in which the case was located; not all cases were heard by circuit courts.
Even though no two cases are identical, we needed some way to group together similar cases. We therefore placed each case into one of eight categories: prison; nature; bodily integrity; institutions; money; family; procedural; and “other.” The prison category encompasses claims brought by inmates. The “nature” category includes cases about the use of natural resources, parks, and the like. “Bodily integrity” includes challenges to policies that raise concerns about a person’s body—such as challenges to contraception, health care, and burial. Monetary claims include tax disputes and bankruptcy proceedings. Institutional claims challenge laws that are either enforced by or enacted on behalf of a larger institution. This includes workplace regulations and school disciplinary proceedings. The “family” category includes religious objections based on family bonds, namely marriages. Procedural issues include policies that address aspects of the court system, such as a speedy trial, search and seizure, bail, and parole.
Table 2: Sectors
We also needed to categorize the government’s asserted interests. In doing so, we undoubtedly lost some of the nuance in each asserted interest. For example, prison safety concerns are different (at least in degree, if not in kind) in maximum-security prisons than they are in county lockup. Nevertheless, the government’s interests fall into ten general categories: prison safety and security; public health; gender equality; preserving Native American heritage; land and wildlife preservation; enforcing the Establishment Clause; enforcing other laws; preserving general operations; public policy; and protecting groups. Most of these are self-explanatory, but a few demand explication. “Enforcing other laws” is an asserted interest when the government wishes to, for example, prevent fraud and punish drug use. “General operations” is a pragmatic government concern that accounts for general public safety, traffic safety, cost minimization, and economic development. “Public policy” is obviously amorphous and depends on the context of the case, but it is invoked when the government wishes to advance an abstract public policy not delineated by statute. This includes preventing litigation and maintaining order and discipline in the schools. Finally, “protecting groups” is the asserted interest when the government seeks to further the well-being of particular constituencies, such as the homeless, minors, and employees.
Table 3: Asserted Interests
|Prison safety and security||99||37.5|
|Preserving Native American heritage||7||2.7|
|Land and wildlife preservation||17||6.4|
|Enforcing the Establishment Clause||7||2.7|
|Enforcing other laws||28||10.6|
*The number of cases is greater than n=264, as several cases may assert multiple interests.
We distinguished between seven religion categories: Christianity, Judaism, Islam, Rastafarianism, Sikhism, Native American religions, and Other (which includes Buddhism, Wicca, Santeria, and Odinism). These categories are broad and do not reflect denominational differences, much less the fact that there is no single “Native American religion.” Still, it draws the necessary distinctions between the largest religions. Groups in the “other” category simply do not bring enough cases to be represented as individual religions.
Table 4: Religions
Unlike many other empirical studies, we account for procedural posture. This is because surviving a motion to dismiss for failure to state a claim is a very different outcome than securing a motion for summary judgment. We therefore noted whether the government was the plaintiff or defendant; whether the opinion followed a motion to dismiss, a motion for a pre-trial injunction, or a decision on the merits (including a motion for summary judgment); and whether the decision was on appeal.
Table 5: Procedural Posture
|Government’s Role||Cases||% Total|
|Appeals||n Cases||% Total|
|Motions**||n Cases||% Total|
|Motion to Dismiss||22||8.3|
*These are typically bankruptcy cases.
**Motions filed by either the Plaintiff or the Defendant. Not all motions are covered, such that the number of motions is less than n=264.
With this dataset in hand, we can trace each prong of the strict scrutiny analysis discussed in Part III. In each subpart, we first look at the summary statistics related to that stage. After that, we gauge the likelihood of succeeding on a free exercise claim under different factors in that stage. We generally find the following:
Table 6: Success Rate
|Prevailing Party||n Cases||% Total|
|Question of fact unresolved||19||7.2|
B. Substantial Burden
Because religious burdens are specific to a religion, we first look to whether there may be significant differences in success rates across burdens and religion. Although this Article is generally only concerned with those cases that pass the substantial burden threshold, we did collect data for those cases in which the court discussed strict scrutiny but ultimately decided there was no substantial burden. We speculate that these are borderline cases: if the absence of a substantial burden had been obvious, no discussion of strict scrutiny would have been needed. The court is therefore looking at strict scrutiny in order to cover its bases and guarantee the legitimacy of its decisions. We found 45 such cases. The plurality related to prison policies (n=16) and to Christianity (n=18).
Table 7 shows the differences between these borderline cases and cases that pass the substantial burden test, using a two-sample z-test. This is a statistical test that incorporates information about the sample size and standard error of the sample to test whether the difference in proportions between two groups is statistically significant. Here, we examine whether those cases that pass the substantial burdens test, compared to those that do not pass, might differ along case characteristics. We found significant differences in several domains. Reassuringly, there are significant differences in whether or not the policy is the least restrictive means. Presumably, the vast majority of cases underlying those non-substantial burden cases should have policies that are the least restrictive means (n=33). Indeed, this statistic demonstrates the relationship between the scope of the policy and the extent of the perceived burden, in turn suggesting that substantial burdens are more about the cost of noncompliance than the centrality of belief.
Similarly, there are significant differences in the discussion of compelling interests. Eighty percent of the non-substantial burden cases are found to have a compelling interest, but that number drops to 43% among the substantial burden cases. Additionally, there are significant differences in the proportion of cases at the motion to dismiss stage or in cases that are appeals. This makes sense because having a non-substantial burden might be grounds for dismissal if there are no additional causes of action. Altogether, the outcome for defendants is less favorable under strict scrutiny: when a substantial burden is not found, defendants ultimately succeed on their motion (or defeat plaintiff’s motion) in 78% of the claims in our dataset, compared to 26% of victories that successfully pass this first stage of free exercise. Similarly, the government’s interest is deemed compelling in 80% of cases without a substantial burden, compared to 43% where there is a substantial burden.
Table 7: Substantial vs. Unsubstantial Burdens
No Substantial Burden
|n Cases||% Total||n Cases||% Total|
|Motion to Dismiss||22||8.3||13||28.9||***|
|Question of fact unresolved||19||7.2||0||0.0|
|Not as presented||120||45.5||5||11.1||***|
|Least Restrictive Means|
|Question of fact unresolved||38||14.4||3||6.3|
*Significant at 5%; **Significant at 1%; ***Significant at 0.1%
These cases are borderline for a reason. It could very well be that cases that are immediately dismissed because of their insubstantial burden on religious exercise have different demographics.
We turn now exclusively to the subset of cases in which burdens have been deemed substantial (n=264). As seen above in Table 2, the plurality of these cases are prison related (48%), followed by institutional pressures on free exercise, such as zoning or education (21%). And as seen in Table 4, the plurality of cases relates to Christianity (33%), followed by other religions (19%) and Islam (17%).
C. Compelling Government Interests
Table 8 shows the “success rate” of asserted compelling interests among various categories—that is, the percentage of cases where the policy appears to further a compelling interest. Based on summary statistics of success rates alone, there appears to be a variety of success rates based on the category of the compelling interest stated and the sector of the case. For instance, cases relating to nature (namely wildlife and national parks) have a 76% compelling interest success rate, whereas cases relating to bodily integrity (for example, contraception and autopsy cases) have a 20% success rate. Similarly, when the government states that its compelling interest is land/wildlife preservation, it has a 77% success rate, but when it claims a compelling interest in public health, that success rate drops to 38%.
To test the statistical validity of these differences, we followed up with a logistic regression analysis with the statistical software STATA, into which the variables in Table 2 were entered. A logistic regression allows us to determine what variables (here, case characteristics) might help to determine a dichotomous outcome (here, whether the stated interest is found compelling or not). Each of the categories was compared to the baseline model, which was set as a prison case impacting Christianity, but with no stated compelling interest. We chose this model because it was the type of case that most frequently appeared in our dataset.
Table 8: Compelling Interests
|n Cases||% Total||Odds Ratio||Standard Error|
|Question of fact unresolved (n=2)||0||0.0|
|Bodily Integrity (n=45)||20||20.0||0.1||0.1||***|
|Native American (n=46)||31||67.4||1.5||0.9|
|No asserted interest (n=28)||2||7.1|
|Prison safety and security (n=99)||56||56.6||55||69||***|
|Public health (n=50)†‡||19||38.0|
|Gender equality (n=30)‡||8||26.7|
|Preserving Native American heritage (n=7)†||6||85.7|
|Land and wildlife preservation (n=17)||13||76.5||109||212||*|
|Enforcing Establishment Clause (n=7)||2||28.6||81||134||**|
|Enforcing other laws (n=28)||13||46.4||103||145||***|
|General operations (n=73)||34||46.6||56||74||**|
|Public policy (n=18)||8||44.4||55||80||**|
|Protecting groups (n=9)||4||44.4||85||167||*|
|Multiple interests (n=93)†||45||48.4||88||112||***|
*Significant at 5%; **Significant at 1%; ***Significant at .1%
†The logistic regression predicted outcomes perfectly here; because of this, the software dropped these automatically from the logistic regression analysis.
‡Because of the nature of the logistic regression analysis, cases with multiple asserted interests were grouped separately. This led to, for example, public health not being represented in the logistical analysis because it was always grouped together with another variable.
Finally, the model perfectly predicted success rates for cases in the family and procedural sectors, for cases involving the Sikh faith, and cases in which the compelling interest was public health, gender equality, or preserving Native American heritage. This suggests that the odds of success are either very high or very low—that is, these factors either predict that the policy will pass or fail the compelling interest prong. As seen in Table 8, courts find a compelling interest for all of the family, procedural, and Sikh cases in our dataset. But this is less clear for the other compelling interests stated. In our dataset, we find that public health and gender equality are nearly always bound together in cases involving contraception funding, and that that is often an insufficiently compelling interest for the court. Similarly, preserving Native American heritage does not appear to be an interest that stands on its own—it is always bound with other interests, most often wildlife preservation. Given that stating multiple interests is generally more likely to yield success on the compelling interest prong, this may explain why the majority of cases citing Native American heritage pass the compelling interest prong.
D. Least Restrictive Means
After a policy passes the compelling interest prong, courts look to the least restrictive means prong. Like Table 8, Table 9 shows the “success rate” of least restrictive means justifications. Here, however, the rate is conditional on having passed the compelling interest analysis. Again, there appear to be differences by criminal case status, sector, and even religion at this stage. But because significantly fewer cases pass the compelling interest analysis (n=107) and the least restrictive means prong, we are limited in our statistical ability to appropriately analyze whether specific sectors, religions, or compelling interests have particularly high success rates.
Table 9: Compelling Interest & Least Restrictive Means
|n Cases||% of n|
|Bodily Integrity (n=9)||0||0.0|
|Native American (n=31)||17||53.3|
|No asserted interest (n=2)||1||50.0|
|Prison safety and security (n=56)||23||41.1|
|Public health (n=19)||8||42.1|
|Gender equality (n=8)||1||12.5|
|Preserving Native American heritage (n=6)||2||33.3|
|Land/wildlife preservation (n=13)||9||69.2|
|Enforcing Establishment Clause (n=2)||2||100.0|
|Enforcing other laws (n=13)||10||76.9|
|General operations (n=34)||15||44.1|
|Public policy (n=8)||3||37.5|
|Protecting groups (n=4)||2||50.0|
|Multiple Interests (n=45)*||19||42.2|
*Includes outcomes from cases analyzed in other compelling interests.Still, some of the summary statistics are of particular interest. For instance, even though cases that cite upholding the Establishment Clause have a 29% success rate on the compelling interest prong, if a case passes that threshold, they are likely to find success on least restrictive means and on strict scrutiny more broadly. Looking to the two cases that pass strict scrutiny because of the Establishment Clause, however, it appears that the cases have unusual facts and unsympathetic plaintiffs. One plaintiff was a teacher fired due to frequent classroom proselytizing, and the other was a student who wanted his public school district to pay for his Catholic school translator. This suggests that the Establishment Clause is an otherwise weak interest for the government to cite.
By contrast, though public policy cases have a 70% success rate as a compelling interest, they only have a 14% success rate when examined for least restrictive means. The only case that succeeded under this scenario is Combs v. Homer Center School District, where the plaintiffs homeschooled their children but refused to turn in a curriculum for review by the district.
 As with any quantitative analysis, we are cautious given the small number of cases we have. Nevertheless, Table 10 gives the most comprehensive and precise analysis of free exercise strict scrutiny cases to date. Specifically, it shows the likelihood of plaintiff success at each stage of litigation. In the end, religious claimants win a significant number of their cases. This is particularly true in two situations. First, the government rarely wins when the court faces a motion for a preliminary injunction. This shows that the religious claimant starts out on top—the government truly bears the burden of proof. Second, and more significantly, when religious claimants are able to demonstrate that a law lacks neutrality, or that it is not generally applicable, they win almost every time. The government has only once successfully articulated a compelling interest furthered by the least restrictive means when the law lacked generality or neutrality. This is reminiscent of free speech jurisprudence, where content-based regulations almost never survive, but where content-neutral regulations might.
But what underlies these statistics? The high success rate is not the product of particularly robust review based on affection for religious liberty but is instead a product of historical accident and government ineptitude. That is, government interests are theoretically compelling but uncertain in their application. Consider two aspects: why the government loses on the compelling interest prong of the analysis, and what it takes to qualify an interest as compelling.
First, when the government’s asserted interest is deemed non-compelling, it is usually because the government’s interest is found not compelling as applied. That is, courts often admit (at least implicitly) that the case could have come out differently had the government’s interest been more directly at stake. Put differently, courts almost never say that an interest could never be compelling. When they do take that step, it is usually because the government failed to assert any interest. The judiciary’s willingness to find an interest theoretically compelling means that more specific interests have a better chance of prevailing. For example, prison administrators are frequently told their interests are not compelling even while game wardens are told their interest in protecting the bald eagle population is compelling: in the former situation, the prison likely asserted a broad interest without stopping to develop the facts behind; in the latter situation, the interest is inherently and obviously specific.
To be sure, courts will sometimes flat-out reject an asserted interest, and the government will inevitably assert an interest unsupported by the facts. Yet when courts invalidate interests only as applied, they concede that the interest is (abstractly speaking) strong enough to overcome religious objections. In this way, the compelling interest prong is like the least restrictive means prong—if the government had articulated its interests more precisely, or if it had developed a more complete record, the result might have been different. This problem is especially acute in RFRA and RLUIPA cases, where the “common-law approach” is particularly vulnerable to judicial reversals.
This leads to the second indication that the review is not robust: the definition of a “compelling interest.” Not all compelling interests are created equal. Some reflect existential threats. Other interests are still strong, but they represent something more like a desire to further policy goals (instead of avoiding a constitutional crisis). And as public policy changes, so too will the outcome of free exercise cases. For example, five justices recently wrote that ensuring cost-free access to contraceptives is a compelling government interest, but the right to access contraceptives is of relatively recent vintage. When Indiana enacted a Religious Freedom Restoration Act, the backlash assumed that nondiscrimination would not be considered a compelling government interest. But one state representative offered an amendment that would have made “preventing discrimination” a compelling interest, and a court could easily find that interest compelling even without a statutory hook. Indeed, the United States Commission on Civil Rights recently promoted this view. Tying acceptable interests to public opinion will likely assuage those concerns (assuming cultural mores develop along their current path), but it should terrify religious claimants who had otherwise planned to take refuge behind state RFRAs. The more those interests are recognized as being “of the highest order,” and the less important we deem religious liberty, the less religious exceptions will be favored.
In the end, it seems that Smith’s central concern—avoiding anarchy—will prove unfounded. Religious claimants will only be given the right to ignore laws that the political culture believes can be ignored. And this will not dilute “compelling” interests because that concern assumes there is a fixed meaning of what is “compelling.” Instead, courts will recognize new interests that they sincerely believe are bona fide compelling interests. But in doing so, it seems that something important about religious liberty—its protection for minority interests—will be lost.
Table 10: Proportion of Cases Succeeding at Each Stage of Litigation
Not Compelling Interest
Claimant Wins at this Stage
And Least Restrictive Means
Not Least Restrictive Means
Not Least Restrictive Means
|n Cases||% Total||n Cases||% Total||n Cases||% Total||n Cases||% Total|
|% of Total||20.1||17||18.6||64.8|
|Prelim. Injunction (n=58)||3||5.3||13||23.2||16||28.6||44||78.6|
|Motion to Dismiss (n=22)||5||22.7||1||4.6||1||4.6||14||60|
|Summary Judgment (n=131)||27||20.6||14||10.7||19||14.5||78||59.5|
|By plaintiff (n=58)||17||29.3||5||8.6||10||17.2||31||53.5|
|By defendant (n=29)||7||24.1||5||17.2||5||17.2||20||69|
|Bodily Integrity (n=45)||0||0||8||17.8||14||31.1||39||86.7|
|Native American (n=46)||17||37||9||19.6||4||8.7||25||54.4|
|Prison safety and security (n=99)||23||23.2||16||16.2||15||15.2||52||52.5|
|Public health (n=50)||8||16||10||20||15||30||39||78|
|Gender equality (n=30)||1||3.3||7||23.3||11||36.7||27||90|
|Preserving Native American Heritage (n=7)||2||28.6||3||42.9||1||14.3||5||71.4|
|Land and Wildlife Preservation (n=17)||9||52.9||3||17.7||2||11.8||8||47.1|
|Enforcing Establishment Clause (n=7)||2||28.6||0||0||2||28.6||5||71.4|
|Enforcing other laws (n=28)||10||35.7||2||7.1||6||21.4||18||64.3|
|General operations (n=73)||15||20.6||11||15.1||15||20.6||48||65.8|
|Public policy (n=18)||3||16.7||4||22.2||2||11.1||11||61.1|
|Protecting groups (n=9)||2||22.2||2||22.2||1||11.1||3||33.3|
*The victory statistic is different than “neither compelling interest, nor least restrictive means” in that we include cases that advance to the next stage on the high likelihood of these factors being met (e.g., restraining orders). For instance, these include cases in which least restrictive means have not yet been decided. For this reason, the number of cases in a row will not sum up to the total n.
†Filed by either plaintiff or defendant.
‡Victory by plaintiff is on non-free exercise grounds, such as Equal Protection.
How to properly regulate religious activity has been a long and contentious struggle. Even though Smith eliminated strict scrutiny for neutral and generally applicable laws, Lukumi (along with RFRA and RLUIPA) revived it. Despite Smith’s concerns, therefore, courts still often require the government to demonstrate a compelling interest furthered through the least restrictive means.
We looked at what characteristics allow free exercise claims through each prong of strict scrutiny: substantial burden, compelling interest, and least restrictive means. We found that a majority (78%) of cases with free exercise claims succeed on a motion or judgment if they pass the substantial burden test, as opposed to 26% if they do not. This jump in success rates suggests that the substantial burden test may be the tallest hurdle in this process. But there appear to be few differences in the relative distribution of sectors, religion, and compelling interest between cases that passed or failed the substantial burden test. This suggests that, for the most part, no set of facts is more sympathetic to judges.
Whether a law passes the compelling interest and least restrictive means prongs varies on the type of case and the issues addressed. Though that seems obvious, our dataset allowed us to see what cases have been most represented at each step. Prison-related cases—in which the government invokes prison safety and security as a concern—make up the plurality of cases represented in a compelling interest analysis. But the sectors and interests that passed the compelling interest test at the highest rates (for example, family and procedural sectors, or preserving Native American heritage) were relatively small. This suggests that success turns on the facts of the individual case. Yet the compelling interest and least restrictive means prongs do not always run together. Moreover, courts usually defer to the religious claimant when a question of fact is presented, allowing the case to continue.
This area is ripe for future research. Although we limited our scope to religious liberty and free exercise, it is likely that the standards set by the court may spill over into other areas—particularly in cases with hybrid claims. Similarly, some are hesitant to apply strict scrutiny in free exercise cases because it might weaken strict scrutiny elsewhere. A comprehensive study of other contexts, such as free speech, is needed before the truth of this claim can be determined.
The use of strict scrutiny has been well documented in other domains, but occasionally that means statements about strict scrutiny are distilled into catchall analyses and flowcharts. Yet as we see here, strict scrutiny is tailored to the facts. Government can—and does—win cases against religious claimants. But when it does win, it does so only in the instant case. Put simply, strict scrutiny makes the government prove its case in every case. For that reason, it is a valuable and necessary protection for religious liberty. The current approach to strict scrutiny is precariously tied to popular opinion and government formulations, however, meaning that the outcome of cases may very well change in the near future.
* J.D. 2017, Harvard Law School; B.A. 2014, Auburn University (Political Science).
** Ph.D. Student, Harvard Kennedy School (Public Policy); J.D. 2017, Harvard Law School; B.A. 2011, Harvard College (Economics).
We contributed equally to this Article. Professor Adam Winkler and Michael Francus provided valuable insight for which we are profoundly grateful. Of course, any remaining errors are our own.
 Daniel A. Farber, The Categorical Approach to Protecting Speech in American Constitutional Law, 84 Ind. L. J. 917, 919 (2009). This characterization may not apply in other contexts. According to Professor Fallon, there are three interpretations of strict scrutiny: a “nearly categorical prohibition”; a “weighted balancing test”; and an “‘illicit motives’ test, aimed at ‘smoking out’ forbidden government purposes.” Richard H. Fallon, Jr., Strict Judicial Scrutiny, 54 UCLA L. Rev. 1267, 1302–03 (2007). But at least in the free exercise context, strict scrutiny essentially equals balancing.
 See Reed v. Town of Gilbert, 135 S. Ct. 2218, 2226 (2015) (citing R.A.V. v. City of St. Paul, 505 U.S. 377, 395 (1992); Simon & Schuster, Inc. v. Members of N.Y. State Crime Victims Bd., 502 U.S. 105, 115–18 (1991)). For the history of strict scrutiny, see Fallon, supra note 1, at 1273–85, 1297–1302; Adam Winkler, Fatal in Theory, Strict in Fact: An Empirical Analysis of Strict Scrutiny in the Federal Courts, 59 Vand. L. Rev. 793, 798–801, 805–09 (2006).
 These questions describe the nature of our society. Cf. Alasdair MacIntyre, After Virtue 254 (3d ed. 2007) (“The nature of any society . . . is not to be deciphered from its laws alone, but from those understood as an index of its conflicts.”).
 494 U.S. 872, 885–86 (1990).
 Smith, 494 U.S. at 886.
 See Christopher C. Lund, Keeping Hobby Lobby in Perspective, in The Rise of Corporate Religious Liberty 285, 298 (Micah Schwartzman et al. eds., 2016) (“No academic or judicial discussion of the compelling-interest test ever seems complete without some reference to a likely parade of horribles. . . . But such hypothetical claims seem less scary when one realizes that they are rarely brought and do not win.”); Christopher C. Lund, Exploring Free Exercise Doctrine: Equal Liberty and Religious Exemptions, 77 Tenn. L. Rev. 351, 373–75 (2010).
 See, e.g., James Madison, Memorial and Remonstrance against Religious Assessments ¶ 1 (June 20, 1785), in 8 The Papers of James Madison, 10 March 1784–28 March 1786, at 295, 299 (Robert A. Rutland & William M.E. Rachel eds., 1973); Virginia Declaration of Rights, art. 16 (June 12, 1776), in 1 Kate M. Rowland, The Life of George Mason, 1725–1792, at 438, 441 (1892).
 U.S. Const. amend. I; see also Cantwell v. Connecticut, 310 U.S. 296, 303–04 (1940) (incorporating the Free Exercise Clause to the states).
 Some believe the free exercise clause was intended to protect exemptions from neutral, generally applicable laws. See City of Boerne v. Flores, 521 U.S. 507, 564 (1997) (O’Connor, J., dissenting) (concluding “the Free Exercise Clause is properly understood as an affirmative guarantee” of religious exemptions); Michael W. McConnell, The Origins and Historical Understanding of Free Exercise of Religion, 103 Harv. L. Rev. 1409, 1511–12 (1990) (arguing a right to exemptions “is more consistent with the original understanding than is a position that leads only to the facial neutrality of legislation.”); see also Kurt T. Lash, The Second Adoption of the Free Exercise Clause: Religious Exemptions Under the Fourteenth Amendment, 88 Nw. U. L. Rev. 1106, 1156 (1994) (arguing “the original intent of the framers of the Fourteenth Amendment” allows religious claimants to seek “exemption from the unjustified impact of generally applicable law.”). Others, however, respond that neutral, generally applicable laws satisfy the First Amendment’s demands. See City of Boerne, 521 U.S. at 539 (Scalia, J., concurring in part) (responding that the better reading of history was that no exemption from neutral, generally applicable laws is required); Philip A. Hamburger, A Constitutional Right of Religious Exemption: An Historical Perspective, 60 Geo. Wash. L. Rev. 915, 948 (1992) (same).
 See, e.g., Church of the Lukumi Babalu Aye, Inc. v. City of Hialeah, 508 U.S. 520, 533 (1993) (citing McDaniel v. Paty, 435 U.S. 618, 626 (1978) (plurality)).
 98 U.S. 145, 166–67 (1878).
 Id. at 167.
 Id. at 166 (positing human sacrifice as an example of impermissible religious action); see also Davíd Carrasco, Human Sacrifice: Aztec Rites, in 6 Encyclopedia of Religion 4185, 4187 (Lindsay Jones ed., 2d ed. 2005) (describing an Aztec human sacrifice ritual). Professor Amar raises an interesting question: what if those being sacrificed consented to the ritual? See Akhil Reed Amar, The Bill of Rights 256 n.100 (1998). But as Professor Volokh reminds us, there are often government interests in preventing so-called “voluntary” actions. Eugene Volokh, Intermediate Questions of Religious Exemptions—A Research Agenda with Test Suites, 21 Cardozo L. Rev. 595, 625–28 (1999) (describing legal interests in guarding against irreversible choices, pressured choices, and negative externalities).
 U.S. Const. amend. I (“Congress shall make no law respecting an establishment of religion, or prohibiting the free exercise thereof . . . .”); see also Everson v. Bd. of Ed., 330 U.S. 1, 15 (1947) (incorporating the First Amendment to the states).
 See Smith v. California, 361 U.S. 147, 157 (1959) (Black, J., concurring) (“I read ‘no law . . . abridging’ to mean no law abridging.”) (free speech case); Hugo L. Black, The Bill of Rights, 35 N.Y.U. L. Rev. 865, 874–76 (1960).
 366 U.S. 599, 605–09 (1961) (plurality).
 Id. at 606–09.
 310 U.S. 296, 304 (1940).
 Id. at 304.
 Id. at 303–04.
 See Braunfeld, 366 U.S. at 612–14 (Brennan, J., dissenting). See also Prince v. Massachusetts, 321 U.S. 158, 175 (1944) (Murphy, J., dissenting).
 374 U.S. 398, 406 (1963). See also Fallon, supra note 1, at 1281 (stating Sherbert “does not employ the precise language now associated with strict scrutiny, but it includes the modem test’s central conceptual elements.”). For a discussion of Sherbert’s relationship to earlier cases, compare Laurence H. Tribe, American Constitutional Law 1255 (2d ed. 1988) (finding continuity between Sherbert and earlier cases), with Jesse H. Choper, The Rise & Decline of the Constitutional Protection of Religious Liberty, 70 Neb. L. Rev. 651, 655 (1991) (finding discontinuity).
 406 U.S. 205, 215 (1972); see also, e.g., Hobbie v. Unemp’t Appeals Comm’n, 480 U.S. 136, 141–42 (1987); Bob Jones Univ. v. United States, 461 U.S. 574, 603 (1983); United States v. Lee, 455 U.S. 252, 259 (1982); Thomas v. Review Bd., 450 U.S. 707, 713–14 (1981).
 See, e.g., Paul Horwitz, The Agnostic Age 178 (2010); Tribe, supra note 22, at 1261–62; Eugene Volokh, A Common-Law Model for Religious Exemptions, 46 UCLA L. Rev. 1465, 1500 n.106 (1999) (collecting authorities).
 O’Lone v. Estate of Shabazz, 482 U.S. 342, 353 (1987) (prisons); Goldman v. Weinberger, 475 U.S. 503, 507 (1986) (military).
 494 U.S. at 885. Smith did not specify whether rational basis would apply instead of strict scrutiny, but lower courts have almost unquestioningly interpreted Smith as applying rational basis. See, e.g., Korte v. Sebelius, 735 F.3d 654, 671 (7th Cir. 2013); Miller v. Reed, 176 F.3d 1202, 1206 (9th Cir. 1999).
 Smith, 494 U.S. at 885 (quoting Reynolds, 98 U.S. at 167).
 Id. at 888; see Richard F. Duncan, Free Exercise is Dead, Long Live Free Exercise: Smith, Lukumi & the General Applicability Requirement, 3 U. Pa. J. Const. L. 850, 853 (2001) (“[Smith] appears to have been determined by a formula that can be stated as follows: Religious Pluralism plus Religious Liberty equals Anarchy.”).
 Smith, 494 U.S. at 889; see Ira C. Lupu, The Trouble with Accommodation, 60 Geo. Wash. L. Rev. 743, 759 (1992) (describing one of Smith’s “central objectives” as “freeing courts from the federal constitutional obligation to weigh state interests against the impact upon religion worked by state policies.”); see also Volokh, Religious Exemptions, supra note 24, at 1521–29, 1565 (noting connection to substantive due process). Along these lines, Professor Smith postulates that “the low win percentage for religious objectors [during the Sherbert era] might be the product of two factors: (a) the fact that democratic legislatures had already acquired the habit of trying to accommodate religious minorities; and (b) the deference that courts properly pay to judgments by other, more electorally accountable branches. . . . And so it might seem that the repudiation of the Sherbert doctrine in Smith represented no great loss.” Steven D. Smith, Religious Freedom and its Enemies, 32 Cardozo L. Rev. 2033, 2042 (2011).
 Smith, 494 U.S. at 884. This distinction blends into the Court’s Lukumi decision. See infra notes 81–89 and accompanying text; Duncan, supra note 28, at 860–61, 882. The Court also said Sherbert was limited to the unemployment context, Smith, 494 U.S. at 884, but this was almost certainly an incorrect reading of Sherbert, see Michael W. McConnell, Free Exercise Revisionism & the Smith Decision, 57 U. Chi. L. Rev. 1109, 1122–23 (1990).
 See City of Boerne, 521 U.S. at 524–29 (defining Congress’s powers under § 5 of the Fourteenth Amendment).
 Pub. L. No. 103-141, 107 Stat. 1488 (1993) (codified at 42 U.S.C. §§ 2000bb–2000bb-4). For a historical account of RFRA’s passage, see generally Robert F. Drinan & Jennifer I. Huffmann, Religious Freedom Restoration Act: A Legislative History, 10 J.L. & Religion 531 (1993).
 42 U.S.C. § 2000bb(b)(1) (2012); see also S. Rep. No. 103-111, at 7–8 (1993), reprinted in 1993 U.S.C.C.A.N. 1892, 1897–98.
 See Burwell v. Hobby Lobby Stores, Inc., 134 S. Ct. 2751, 2761 (2014) (applying RFRA); Gonzales v. O Centro Espirita Beneficente Uniao do Vegetal, 546 U.S. 418, 439 (2006) (applying RFRA even though it “was not required as a matter of constitutional law.”).
 See State Religious Freedom Restoration Acts, Nat’l. Conf. of State Legs. (Oct. 15, 2015) (collecting statutes), perma.cc/SWU8-R8UM. But cf. generally Christopher C. Lund, Religious Liberty After Gonzales: A Look at State RFRAs, 55 S.D. L. Rev. 466 (2010) (noting multiple problems with state RFRAs). Some states also impose strict scrutiny as a matter of state constitutional law. See Kara Loewentheil, The Satanic Temple, Scott Walker, and Contraception: A Partial Account of Hobby Lobby’s Implications for State Law, 9 Harv. L. & Pol’y Rev. 89, 104 n.75 (2015) (collecting cases).
 521 U.S. at 529–36.
 Pub. L. No. 106-274, 114 Stat. 803 (2000) (codified at 42 U.S.C. §§ 2000cc–2000cc-5). See Douglas Laycock & Luke W. Goodrich, RLUIPA: Necessary, Modest, and Under-Enforced, 39 Fordham Urb. L.J. 1021, 1021–41 (2012) (overviewing the history of, and need for, RLUIPA); David W. Dunlap, God, Caesar & Zoning, N.Y. Times, Aug. 27, 2000, at RE1. Cutter v. Wilkinson upheld RLUIPA against an Establishment Clause challenge. 544 U.S. 709, 714 (2005).
 42 U.S.C. § 2000cc(a) (2012) (land use regulation); 42 U.S.C. § 2000cc-1(a) (2012) (prisons).
 42 U.S.C. §§ 2000cc(b)(2), 2000cc-1(b)(2) (2012). The Second Circuit has affirmed RLUIPA against an as-applied Commerce Clause challenge. See Westchester Day Sch. v. Vill. of Mamaroneck, 504 F.3d 338, 354 (2d Cir. 2007). Most circuits have avoided this question. See, e.g., Smith v. Allen, 502 F.3d 1255, 1274 n.9 (11th Cir. 2007) (doubting Commerce Clause applicability); Charles v. Verhagen, 348 F.3d 601, 609 (7th Cir. 2003) (not reaching the question).
 42 U.S.C. §§ 2000cc(b)(1), 2000cc-1(b)(1) (2012). Several lower courts have upheld RLUIPA under the Spending Clause. See, e.g., Sossamon v. Lone Star State of Tex., 560 F.3d 316, 328 n.34 (5th Cir. 2009); Madison v. Virginia, 474 F.3d 118, 124 (4th Cir. 2006).
 RLUIPA only applies in two contexts: land use regulations and institutionalized persons (e.g., prisons). RFRA, by contrast, applies to all federal government actions. Together, they cover a large (but unknown) percentage of religious liberty claims.
 Lukumi, 508 U.S. at 531. Smith implied these limits, 494 U.S. at 879–91, but Lukumi brought them to the fore and dispelled any doubt as to their existence. Those doubts were not unfounded given that Smith cited the Lukumi district court opinion as an example of religious claimants seeking an impermissible exemption. Smith, 494 U.S. at 888–89 (citing Lukumi Babalu Aye Inc. v. City of Hialeah, 723 F. Supp. 1467 (S.D. Fla. 1989)).
 Commack Self-Serv. Kosher Meats, Inc. v. Hooker, 680 F.3d 194, 211 (2d Cir. 2012) (citing Gillette v. United States, 401 U.S. 437, 452 (1971)).
 Lukumi, 508 U.S. at 546.
 The religious claimant must demonstrate the existence of a substantial burden. Navajo Nation v. U.S. Forest Serv., 535 F.3d 1058, 1069 (9th Cir. 2008) (en banc). See Gonzales, 546 U.S. at 428 (noting church met the burden); Jimmy Swaggart Ministries v. Bd. of Equalization, 493 U.S. 378, 384–85, 389–92 (1990) (holding religious claimants had not demonstrated a substantial burden). If RFRA or RLUIPA applies, the government must meet strict scrutiny. If neither applies, Lukumi allows the policy to survive if it is neutral and generally applicable. Some lower courts seem to place the burden of proving neutrality or generality on religious claimants. See Stormans, Inc. v. Wiesman, 794 F.3d 1064, 1079 (9th Cir. 2015); Grace United Methodist Church v. City of Cheyenne, 451 F.3d 643, 653 (10th Cir. 2006); cf. Miller v. Johnson, 515 U.S. 900, 916 (1995) (placing burden on plaintiff in political gerrymandering cases). But the Second Circuit treats neutrality and generality as a defense, meaning the government must justify imposing the burden by pointing to a neutral and generally applicable law. See Fifth Ave. Presbyterian Church v. City of N.Y., 293 F.3d 570, 575 (2d Cir. 2002); cf. Axson-Flynn v. Johnson, 356 F.3d 1277, 1294 (10th Cir. 2004) (reversing grant of summary judgment for government, remanding for government to establish generality and neutrality). In the end, this ambiguity is inconsequential: most courts simply review the facts for themselves without discussing the burden of proof. See, e.g., Lukumi, 508 U.S. at 531–46; Ward v. Polite, 667 F.3d 727, 740 (6th Cir. 2012); Blackhawk v. Pennsylvania, 381 F.3d 202, 209 (3d Cir. 2004) (Alito, J.).
 The government bears the burden of proof here. 42 U.S.C. §§ 2000bb-1(c), 2000cc(a)(1), 2000cc-1(a) (2012); Thomas, 450 U.S. at 718.
 RFRA protects corporations. See Hobby Lobby, 134 S. Ct. at 2768. For a discussion of corporate religious liberty, see generally The Rise of Corporate Religious Liberty (Micah Schwartzman et al. eds., 2016). Hobby Lobby did not address whether the Constitution also protects corporations.
 Religious beliefs are not determined by referencing theological texts. Instead, the Court asks whether the beliefs are sincerely held and subjectively religious. United States v. Ballard, 322 U.S. 78, 85–88 (1944).
 See Lyng v. Nw. Indian Cemetery Protective Ass’n, 485 U.S. 439, 450–51 (1988); see also Oklevueha Native Am. Church of Haw., Inc. v. Lynch, 828 F.3d 1012, 1017 (9th Cir. 2016) (rejecting a RFRA claim where the claimants could not demonstrate the presence of a forced choice).
 These requirements need not be in the criminal code; requirements for optional benefits count for these purposes. See Thomas, 450 U.S. at 717–18; Sherbert, 374 U.S. at 404–05 & n.6; see also William W. Van Alstyne, The Demise of the Right–Privilege Distinction in Constitutional Law, 81 Harv. L. Rev. 1439, 1446 & n.26 (1968) (citing Sherbert as one of many cases breaking down the right-privilege distinction).
 Sherbert, 374 U.S. at 404.
 406 U.S. at 208–09.
 See Baranowski v. Hart, 486 F.3d 112, 125 (5th Cir. 2007); see also Shilling v. Crawford, 536 F. Supp. 2d. 1227, 1233 (D. Nev. 2008) (conditioning receiving Kosher meals on being transferred to a higher security facility).
 Swaggart, 493 U.S. at 391–92; Hernandez v. Comm’r, 490 U.S. 680, 699 (1989) (citing Hobbie, 480 U.S. at 141–42; Thomas, 450 U.S. at 717–19; Yoder, 406 U.S. at 220–21).
 Ira Lupu, Where Rights Begin: The Problem of Burdens on The Free Exercise of Religion, 102 Harv. L. Rev. 933, 961 (1989) (distinguishing the “coercion theory” from the “substantial impact” theory of burdens); cf. Michael C. Dorf, Incidental Burdens of Fundamental Rights, 109 Harv. L. Rev. 1175, 1213 (1996) (“[N]either Sherbert nor Yoder gives a satisfactory explanation of the substantiality threshold of RFRA.”).
 See Yoder, 406 U.S. at 218 (“[Wisconsin’s compulsory attendance law] carries with it precisely the kind of objective danger to the free exercise of religion that the First Amendment was designed to prevent. As the record shows, [the law] carries with it a very real threat of undermining the Amish community and religious practice as they exist today; they must either abandon belief and be assimilated into society at large or be forced to migrate to some other and more tolerant region.”). But see Navajo Nation v. U.S. Forest Serv., 535 F.3d 1058, 1070 n.12 (9th Cir. 2008) (en banc) (arguing that “the undue burden [in Yoder] was the penalty of criminal sanctions on the parents for refusing to enroll their children in such school.”).
 490 U.S. at 699.
 Swaggart, 493 U.S. at 391.
 True, in Hernandez and Jimmy Swaggart Ministries there was no theological objection. See id. at 391–92 (also stating that Sherbert did not apply because there was no religious-based objection); Hernandez, 490 U.S. at 699. But the critical lesson to learn from Hernandez and Jimmy Swaggart Ministries is that the court will look at the pressure imposed on religious adherents. Had the taxes in those cases been discriminatory, for instance, the Court could very well have found the existence of a substantial burden—even in the absence of a theological objection. Cf. generally Lukumi, 508 U.S. 520 (not asking whether there was a substantial burden). This is because the pressure imposed on religious free exercise would have been substantial.
 See Holt v. Hobbs, 135 S. Ct. 853, 859 (2015); Hobby Lobby, 134 S. Ct. at 2775–76.
 See Michael A. Helfand, Identifying Substantial Burdens, 2016 U. Ill. L. Rev. 1771, 1805 (advocating this approach). We thank Trenton Van Oss for suggesting the “compliance” and “noncompliance” phrasing.
 See, e.g., E. Tex. Baptist Univ. v. Burwell, 793 F.3d 449, 456 (5th Cir.), vacated and remanded sub nom., Zubik v. Burwell, 136 S. Ct. 1557 (2016) (per curiam) (“Is the penalty for noncompliance substantial?”); Korte, 735 F.3d at 682–94; Hobby Lobby Stores, Inc. v. Sebelius, 723 F.3d 1114, 1140–41 (10th Cir. 2013), aff’d sub nom. Hobby Lobby, 134 S. Ct. 2751 (2014); Kaemmerling v. Lappin, 553 F.3d 669, 678–79 (D.C. Cir. 2008) (citing Thomas, 450 U.S. at 718).
 485 U.S. 439, 450–51 (1988); see also Russell W. Galloway, Basic Free Exercise Clause Analysis, 29 Santa Clara L. Rev. 865, 871 (1989).
 See, e.g., Thomas, 450 U.S. at 714.
 Both sides in the recent contraception mandate controversy putatively agree. Compare Catholic Health Care Sys. v. Burwell, 796 F.3d 207, 219–26 (2d Cir. 2015), vacated and remanded sub nom., and Zubik, 136 S. Ct. 1557 (2016) (per curiam) (agreeing that courts could not gauge religious beliefs, but holding the challenged requirement did not actually trigger the religiously prohibited conduct), with Sharpe Holdings, Inc. v. U.S. Dep’t of Health & Human Servs., 801 F.3d 927, 941–43 (8th Cir. 2015), vacated and remanded sub nom., and Zubik, 136 S. Ct. 1557 (holding the exemption scheme actually did impose a substantial burden). See also Recent Case, Seventh Circuit Denies Preliminary Injunction to Wheaton College, 129 Harv. L. Rev. 851, 855 n.48 (2016) (discussing circuit split).
 Thomas, 450 U.S. at 714; see also Frederick Mark Gedicks, “Substantial” Burdens: How Courts May (and Why They Must) Judge Burdens on Religion under RFRA, 85 Geo. Wash. L. Rev. 94, 106–09 (2017) (discussing the “Religious-Question Doctrine” and citing authorities). But see Mack v. O’Leary, 80 F.3d 1175, 1179 (7th Cir. 1996) (Posner, J.), vacated by O’Leary v. Mack, 522 U.S. 801 (1997) (vacating and remanding for further consideration in light of City of Boerne, 521 U.S. 507 (1997)); Marc O. DeGirolami, Substantial Burdens Imply Central Beliefs, 2016 U. Ill. L. Rev. 19, 20–21 (arguing religious beliefs must be considered, at least to some extent).
 42 U.S.C. § 2000cc-5(7)(A) (2012) (“The term ‘religious exercise’ includes any exercise of religion, whether or not compelled by, or central to, a system of religious belief.”); Lyng, 485 U.S. at 457–58.
 To be sure, this approach is not perfect. Religions often require obedience to civil laws. See, e.g., Romans 13:1 (English Standard Version) (“Let every person be subject to the governing authorities.”). Although these religions also require following conscience instead of the law, it would be bizarre if doctrine required religious claimants to be willing to break the law and then said such lawbreaking was not substantial enough. Following Yoder in addition to Hernandez and Jimmy Swaggart prevents this perverse outcome. Moreover, a necessary proviso would be that discriminatory treatment is always a substantial burden. This essentially says that a burden becomes more onerous because it is intentionally applied against religious actors. Cf. Lukumi, 508 U.S. at 531 (not asking whether there was a substantial burden); Lighthouse Inst. for Evangelism, Inc. v. City of Long Branch, 510 F.3d 253, 263 (3d Cir. 2007) (citing Tenafly Eruv Ass’n v. Borough of Tenafly, 309 F.3d 144, 170 (3d Cir. 2002)); Brown v. Borough of Mahaffey, 35 F.3d 846, 849–50 (3d Cir. 1994)); Westchester Day Sch., 504 F.3d at 351 (holding that “arbitrary and capricious” zoning practices are substantial burdens).
 508 U.S. 520.
 See id. at 537 (quoting Smith, 494 U.S. at 884) (“As we noted in Smith, in circumstances in which individualized exemptions from a general requirement are available, the government ‘may not refuse to extend that system to cases of religious hardship without compelling reason.’”); id. at 531–32 (“A law failing to satisfy the requirements [of neutrality and general applicability] must be justified by a compelling governmental interest and must be narrowly tailored to advance that interest.”).
 Ashwander v. TVA, 297 U.S. 288, 347 (1936) (Brandeis, J., concurring) (citations omitted); see Lisa A. Kloppenberg, Avoiding Constitutional Questions, 35 B.C. L. Rev. 1003, 1025–35 (1994) (discussing the “last resort” canon).
 Several cases apply RFRA in bankruptcy cases, where creditors seek to recover religious donations. See, e.g., In re Young, 82 F.3d 1407, 1417 (8th Cir. 1996), vacated sub nom. Christians v. Crystal Evangelical Free Church, 521 U.S. 1114 (1997) (vacating and remanding for further consideration in light of City of Boerne, 521 U.S. 507 (1997)). But see Listecki v. Official Comm. of Unsecured Creditors, 780 F.3d 731, 736 (7th Cir. 2015) (holding RFRA did not apply). We do not determine whether RFRA (or RLUIPA, for that matter) should apply in these quasi-governmental cases.
 See Gonzales, 546 U.S. at 439.
 42 U.S.C. § 1997e(a) (2012) (The Prison Litigation Reform Act requires inmates to exhaust administrative remedies before seeking redress in court).
 42 U.S.C. § 2000cc(b)(1) (2012). There are other equality and nondiscrimination rules in 42 U.S.C. § 2000cc(b)(2)–(3) (2012).
 See Centro Familiar Cristiano Buenas Nuevas v. City of Yuma, 651 F.3d 1163, 1172 (9th Cir. 2011); River of Life Kingdom Ministries v. Vill. of Hazel Crest, 611 F.3d 367, 381 (7th Cir. 2010) (en banc).
 Courts are split on how to define unequal treatment. See Centro Familiar, 651 F.3d at 1169 n.25 (noting circuit split); see also Andrew Cleves, Comment, Equal Terms: What Does It Mean and How Does It Work: Interpreting the Equal Terms Provision of the Religious Land Use and Institutionalized Persons Act (RLUIPA), 80 U. Cin. L. Rev. 179, 187 (2011) (same). According to one writer, “courts have made it very challenging for a plaintiff to successfully establish a prima facie equal terms violation under RLUIPA.” Bram Alden, Comment, Reconsidering RLUIPA: Do Religious Land Use Protections Really Benefit Religious Land Users? 57 UCLA L. Rev. 1779, 1802 (2010).
 Centro Familiar, 651 F.3d at 1171 n.37; River of Life, 611 F.3d at 370–71; Lighthouse Inst. for Evangelism, Inc. v. City of Long Branch, 510 F.3d 253, 269 (3d Cir. 2007).
 Primera Iglesia Bautista Hispana of Boca Raton, Inc. v. Broward Cnty., 450 F.3d 1295, 1308 (11th Cir. 2006); Midrash Sephardi, Inc. v. Town of Surfside, 366 F.3d 1214, 1232 (11th Cir. 2004).
 Smith, 494 U.S. at 879–81.
 Lukumi, 508 U.S. at 531–32.
 Id. at 533; see also Cent. Rabbinical Cong. of U.S. & Canada v. N.Y. City Dep’t of Health & Mental Hygiene, 763 F.3d 183, 193–96 (2d Cir. 2014); Stormans, Inc. v. Wiseman, 794 F.3d 1064, 1077 (9th Cir. 2015).
 See Lukumi, 508 U.S. at 543; see also Stormans, 794 F.3d at 1077–78; Tenafly, 309 F.3d at 165.
 Commack Self-Service Kosher Meats, Inc. v. Hooker, 680 F.3d 194, 211 (2d Cir. 2012) (quoting Gillette, 401 U.S. at 452).
 Lukumi, 508 U.S. at 526.
 Id. at 535–38.
 Id. at 538–39.
 Id. at 544.
 Horwitz, supra note 24, at 183.
 170 F.3d 359, 366 (3d Cir. 1999).
 Id. at 366.
 Id. (“[W]hen the government makes a value judgment in favor of secular motivations, but not religious motivations, the government’s actions must survive heightened scrutiny.”) (footnote omitted).
 Id. at 366. Other cases have reached similar conclusions. See, e.g., Rader v. Johnston, 924 F. Supp. 1540, 1556–58 (D. Neb. 1996). Not all exemptions indicate religious gerrymandering, however. See, e.g., Stormans, 794 F.3d at 1080–82 (holding that regulation was generally applicable where the secular exemptions furthered policy goals while the requested religious exemptions frustrated them); Am. Fam. Ass’n, Inc. v. FCC, 365 F.3d 1156, 1171 (D.C. Cir. 2004) (finding rule neutral and generally applicable because it did not single out religion). These differences result from competing approaches to defining neutrality and generality. See Douglas Laycock, The Religious Exemption Debate, 11 Rutgers J.L. & Religion 139, 143–44 (2009) (“Some courts have said that all laws are generally applicable unless they were enacted with anti-religious motive or single out religion for uniquely disadvantageous treatment. Other courts … have said that a law that is generally applicable is a law that applies to everybody. If a law has a secular exception that undermines its purpose, then it must also have a religious exception—or a compelling reason why not.”) (citations omitted). It is beyond this Article’s scope to resolve those differences.
 Smith, 494 U.S. at 888.
 Lukumi, 508 U.S. at 546 (quoting McDaniel v. Paty, 435 U.S. 618, 628 (1978)); see also Brown v. Entm’t Merch. Ass’n, 564 U.S. 786, 804 (2011) (describing strict scrutiny as requiring a “high degree of necessity”); Sherbert, 374 U.S. at 406 (“paramount interests”); Korte, 735 F.3d at 685–86. But see Michael A. Helfand, Religious Institutionalism, Implied Consent, and the Value of Voluntarism, 88 S. Cal. L. Rev. 539, 582 (2015) (arguing for a broad reading of “compelling interest.”).
 Sherbert, 374 U.S. at 403.
 Gonzales, 546 U.S. at 431–32; Yoder, 406 U.S. at 236. As Gonzales noted, this is similar to how strict scrutiny is applied elsewhere. See Grutter v. Bollinger, 539 U.S. 306, 327 (2003); Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 228 (1995).
 Lukumi, 508 U.S. at 547 (quoting Fla. Star v. B.J.F., 491 U.S. 524, 541–42 (1989) (Scalia, J., concurring in part & concurring in judgment)). But see Yellowbear v. Lampert, 741 F.3d 48, 61 (10th Cir. 2014) (“A government can rebut an argument from underinclusion by showing that it hasn’t acted in a logically inconsistent way …”).
 See United States v. Friday, 525 F.3d 938, 949 (10th Cir. 2008) (McConnell, J.) (“We now conclude, as other circuits have, that both prongs of RFRA’s strict scrutiny test are legal questions.”). In the prison context, courts should defer to prison administrators on certain factual questions. See Cutter v. Wilkinson, 544 U.S. 709, 723 (2005) (citing 146 Cong. Rec. S7774, S7775 (July 27, 2000) (joint statement of Sen. Hatch and Sen. Kennedy on RLUIPA)); cf. Overton v. Bazzetta, 539 U.S. 126, 132 (2003) (“We must accord substantial deference to the professional judgment of prison administrators, who bear a significant responsibility for defining the legitimate goals of a corrections system and for determining the most appropriate means to accomplish them.”). For example, if an inmate wants to grow a beard but the prison objects on safety grounds, courts should defer to the prison’s safety assessment even though the legal question—whether the safety interest is compelling in that particular instance—is for the courts.
 Supra note 24; John P. Forren, Revisiting Four Myths About the Peyote Case, 8 U. Pa. J. Const. L. 209, 214 (2006); cf. Whole Woman’s Health v. Hellerstedt, 136 S. Ct. 2292, 2328 (2016) (Thomas, J., dissenting) (stating that “recent decisions reflect the Court’s tendency to relax purportedly higher standards of review for less-preferred rights.”). Empirical data supports this position. See Winkler, supra note 2, at 815, 857–62 (noting government policies survive in 59% of religious liberty cases, compared to 22% of free speech cases and 27% of discrimination cases). But the mere fact that religious claimants lose more often than free speech advocates does not necessarily mean strict scrutiny is diluted. It might instead be free speech is given “something beyond strict scrutiny.” Eugene Volokh, Freedom of Speech, Permissible Tailoring, and Transcending Strict Scrutiny, 144 U. Pa. L. Rev. 2417, 2452 (1996). Or it might be that there are more occasions for regulating religious conduct than regulating free speech. Finally, the free speech survival rate might be inflated because certain “low-value” categories of speech (such as fighting words) are unprotected whereas all religious exercises—no matter how harmful—are examined under the same framework.
 See Douglas Laycock & Oliver S. Thomas, Interpreting the Religious Freedom Restoration Act, 73 Tex. L. Rev. 209, 222 (1994) (“The compelling interest test has fallen into disarray, especially in the lower courts.”).
 See Laycock & Thomas, supra note 102, at 224, 226 (arguing RFRA courts should adopt Sherbert’s stringent view of compelling interests).
 See Swaggart, 493 U.S. at 384 (citing Yoder, 406 U.S. at 220).
 Holt, 135 S. Ct. at 864 (quoting Hobby Lobby, 134 S. Ct. at 2780).
 Id. at 864 (citing Hobby Lobby, 134 S. Ct. at 2780; United States v. Playboy Entm’t Grp., 529 U.S. 803, 815 (2000)) (internal alterations omitted).
 Butler v. Michigan, 352 U.S. 380, 383 (1957). Put differently still, the government cannot use a hammer to kill a flea. See The Emperor’s New Groove (Walt Disney Pictures 2000).
 42 U.S.C. §§ 2000bb-1(b)(2), 2000cc(a)(1)(B), 2000cc-1(a)(2) (2012).
 See, e.g., Lukumi, 508 U.S. at 546; see also City of Boerne, 521 U.S. at 535 (stating, without citation, that pre-Smith cases did not use a least restrictive means test); N. Arapaho Tribe v. Ashe, 92 F. Supp. 3d 1160, 1182 (D. Wyo. 2015) (stating the Free Exercise Clause requires narrow tailoring, not the least restrictive means); Martin S. Lederman, Reconstructing RFRA: The Contested Legacy of Religious Freedom Restoration, 125 Yale L.J. Forum 416, 429 (2016) (stating RFRA “had the potential, if read literally, to be much more restrictive than anything the Court had insisted upon in its pre-Smith jurisprudence”). Obviously, RFRA and RLUIPA go further than Sherbert-era jurisprudence to the extent they require strict scrutiny in prison and military cases. But to mean much, claims that RFRA and RLUIPA go beyond Sherbert need to assert far more. Narrow tailoring only requires that the policy be no more extensive than necessary; it does not require that the policy impose the least burden possible. See Bd. of Trustees of State Univ. of N.Y. v. Fox, 492 U.S. 469, 477–78 (1989); Members of City Council v. Taxpayers for Vincent, 466 U.S. 789, 810 (1984). “Least restrictive means” is a subset of “narrowly tailored.” See Ian Ayres, Narrow Tailoring, 43 UCLA L. Rev. 1781, 1788–89 (1996) (citing Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 280 n.6 (1986)); see also Volokh, Freedom of Speech, supra note 101, at 2422–23 (identifying the “least restrictive means” test as a prong of “narrow tailoring,” which also asks whether the interest is actually furthered and whether the regulation is over-inclusive or under-inclusive).
 Hobby Lobby, 134 S. Ct. at 2793 (Ginsburg, J., dissenting) (citing Thomas, 450 U.S. at 718; Sherbert, 374 U.S. at 407; Thomas C. Berg, The New Attacks on Religious Freedom Legislation and Why They Are Wrong, 21 Cardozo L. Rev. 415, 424 (1999)).
 Id. at 2767 & n.18 (majority opinion).
 Id. at 2767 n.18.
 Lukumi, 508 U.S. at 546.
 See generally Winkler, supra note 2. Another article considered how the U.S. Circuit Court of Appeals treated religious exemption claims from 1980 to 1990. James E. Ryan, Note, Smith & the Religious Freedom Restoration Act: An Iconoclastic Assessment, 78 Va. L. Rev. 1407, 1416–37 (1992). Our Article differs in substantial respects, however: it considers cases from 1990 to 2015, it includes all federal courts, and it is more explicitly empirical. This is not the first empirical study of free exercise cases, but most others have looked at judicial ideology and decisionmaking. See, e.g., Michael Heise & Gregory C. Sisk, Free Exercise of Religion Before the Bench: Empirical Evidence from the Federal Courts, 88 Notre Dame L. Rev. 1371 (2013); Gregory C. Sisk, Michael Heise, & Andrew P. Morris, Searching for the Soul of Judicial Decisionmaking: An Empirical Study of Religious Freedom Decisions, 65 Ohio St. L.J. 491 (2004). To the extent studies did not focus on judicial decision-making, they tended to be limited in scope. See, e.g., Gregory C. Sisk & Michael Heise, Muslims and Religious Liberty in the Era of 9/11: Empirical Evidence from the Federal Courts, 98 Iowa L. Rev. 231, 251–52 (2012) (finding that Muslim claimants won about 22.2% of cases, compared to the 38.0% success rate for other religions); Amy Adamczyk et al., Religious Regulation and the Courts: Documenting the Effects of Smith and RFRA, 46 J. Church & State 237, 239 (2004); John Wybraniec & Roger Finke, Religious Regulation and the Courts: The Judiciary’s Changing Rule in Protecting Minority Religions from Majoritarian Rule, 40 J. Scientific Study of Religion 427, 433–40 (2001); Ira C. Lupu, The Failure of RFRA, 20 U. Ark. Little Rock L.J. 575, 585 (1998) (arguing “RFRA failed to produce any substantial improvement in the legal atmosphere surrounding religious liberty.”).
 Technically, we started in April 1990, as we excluded cases decided before Smith.
 We entered this search term (for published federal opinions) into WestlawNext: adv: DA(1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015) & “strict scrutiny” “strictest scrutiny” “exacting scrutiny” ((compelling /2 interest) /s (narrow! “least restrictive”)) & “free exercise” “religious liberty” “religious freedom” “freedom #of religion” See Winkler, supra note 2, at 810 n.101 (using a similar search term). We first ran this search in October 2015. We ran it again in January 2016 to collect opinions issued between October and December 2015.
 This excluded a number of religious liberty cases that were brought under the free speech clause. Those cases are quite common. See Steven P. Brown, Trumping Religion 58 (2002) (arguing that “[m]ore than any other constitutional principle, the free speech clause of the First Amendment has come to characterize the religious liberty efforts” of certain groups); see also Patrick M. Garry, Inequality Among Equals: Disparities in the Judicial Treatment of Free Speech and Religious Exercise Claims, 39 Wake Forest L. Rev. 361, 384–93 (2004) (criticizing this strategy).
 Sixteen cases appeared twice in the dataset because of appeals and fourteen cases appear twice for different motions. Twenty cases appear multiple times, totaling fifty-two observations, because they have multiple claims that the court analyzed individually.
 See, e.g., United States v. Epstein, 91 F. Supp. 3d 573, 585 (D.N.J. 2015) (“[T]he Government unquestionably has a compelling interest in uniformly applying kidnapping laws to prevent serious crimes of violence”); Bronson v. Swensen, 394 F. Supp. 2d 1329, 1332 (D. Utah 2005), vacated for lack of standing, 500 F.3d 1099 (10th Cir. 2007) (“[T]he State of Utah has a compelling state interest in banning plural marriage”).
 See, e.g., Chance v. Tex. Dep’t of Criminal Justice, 730 F.3d 404, 408 (5th Cir. 2013) (accepting prison’s asserted interest in prison administration); McTernan v. City of York, 564 F.3d 636, 651 (3d Cir. 2009) (finding question of material fact existed as to whether “traffic safety” was compelling as applied); Muhammad v. N.Y.C. Transit Auth., 52 F. Supp. 3d 468, 489 (E.D.N.Y. 2014) (rejecting asserted interest in a uniform workforce).
 See, e.g., A.A. ex rel. Betenbaugh v. Needville Indep. Sch. Dist., 611 F.3d 248, 268 (5th Cir. 2010) (holding school district’s asserted interest in enforcing its grooming policy—including related concerns about hygiene and classroom order—was not compelling); Hudson v. Dennehy, 538 F. Supp. 2d 400, 412 (D. Mass. 2008) (accepting asserted interest in prisoner rehabilitation in addition to interest in prison order); Hankins v. N.Y. Annual Conf. of United Methodist Church, 516 F. Supp. 2d 225, 237 (E.D.N.Y. 2007) (rejecting, as applied, asserted interest in preventing age discrimination).
 See, e.g., Listecki, 780 F.3d at 745 (“The Committee’s asserted compelling governmental interest is the protection of creditors”); Fifth Ave. Presbyterian Church, 293 F.3d at 576 (noting the city asserted an interest in ensuring the homeless slept in shelters).
 See Fed. R. Civ. P. 12(b). Most of these were motions under Rule 12(b)(6) for failure to state a claim. See Ashcroft v. Iqbal, 556 U.S. 662, 678–79 (2009); Conley v. Gibson, 355 U.S. 41, 45–46 (1957).
 See Fed. R. Civ. P. 65(a)–(b); Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 20 (2008).
 See Fed. R. Civ. P. 56(a); Celotex Corp. v. Catrett, 477 U.S. 317, 322–24 (1986).
 See Roxy Peck, Chris Olsen & Jay L. Devore, Introduction to Statistics and Data Analysis 594–99 (5th ed. 2015).
 The challenged policies are also more likely to be found to be the least restrictive means. This demonstrates a hesitation to say that a burden on a religious belief is insubstantial.
 Note that these rates are not identical because a case involving nature, for example, may not necessarily name land or wildlife preservation as a compelling interest, although it often does. For example, in one case the government cited complying with the multi-use mandate for national parks, ensuring parkgoers’ safety, and avoiding an Establishment Clause violation as reasons to use artificial snow made from treated sewage on a ski area atop a sacred mountain. Navajo Nation, 479 F.3d at 1044–46.
 See Jeremy J. Foster, Emma Barkus & Christian Yavorsky, Understanding and Using Advanced Statistics: A Practical Guide for Students 57–58 (2006).
 With regard to overfitting concerns, we tested subsets of variables (for example, sector only, compelling interest only) and found that results were similar to the full model. Due to this concern, we hesitate to interpret the odds ratios and focus our discussion on the significance of each factor only.
 See, e.g., Zubik v. Sebelius, 983 F. Supp. 2d 576 (W.D. Pa. 2009), rev’d sub nom. on other grounds, Geneva Coll. v. Sec’y U.S. Dep’t of Health & Human Servs., 778 F.3d 422 (3d Cir. 2015), vacated and remanded sub nom. Zubik, 136 S. Ct. 1557.
 See, e.g., United States v. Winddancer, 435 F. Supp. 2d 687, 695 (M.D. Tenn. 2006); United States v. Lundquist, 932 F. Supp. 1237, 1243 (D. Or. 1996).
 See Helland v. S. Bend Cmty. Sch. Corp., 93 F.3d 327, 329 (7th Cir. 1996).
 See Zobrest v. Catalina Foothills Sch. Dist., 963 F.2d 1190, 1192 (9th Cir. 1992), rev’d on other grounds, 509 U.S. 1 (1993).
 468 F. Supp. 2d 738, 739–40 (W.D. Pa. 2006). The strict scrutiny analysis was tied in with a discussion of hybrid rights, but the court also considered free exercise standing alone. See Combs, 468 F. Supp. 2d at 776–77.
 As previously mentioned, we take our lead from Professor Winkler’s 2006 article. We followed his methodology in many respects. See Winkler, supra note 2, at 809–12 (describing the methodology). Winkler found that religious claimants only win strict scrutiny cases 41% of the time. Id. at 815. Our data tells a different story: religious claimants win much more often than Winkler reports. This difference may be explained in several ways. First, and most simply, we have more cases: 264 compared to 73. See id. at 857. Second, Winkler looked at cases from 1990 to 2003; our data spans 1990 to 2015. Winker wrote that strict scrutiny was “apparently becoming more fatal” as the years passed by, and we could expect that trend to continue. Id. at 824–25. Finally, we did not follow his methodology in every respect. For example, he excluded both preliminary injunction cases and “decisions that were subsequently reversed or affirmed on appeal.” Id. at 811. We included all of these cases, and Winkler’s decision to exclude them is subject to question. He ignores that preliminary injunctions are an essential litigation strategy. Secure the injunction, and the case might settle. By excluding these cases, Winkler focuses only on cases that are so contested that parties do not relent. His decision to exclude affirmed cases obscures the fact that two separate panels reached the same conclusion. Litigants could—and do—cite both opinions, so it makes little sense to ignore the lower opinion. Finally, although on first thought one might choose to exclude reversed cases, that decision obscures what judges actually do. This Article attempts to discern how judges actually apply strict scrutiny. From this perspective, it makes little sense to exclude a district court merits decision simply because a circuit court thought there was a question of fact precluding summary judgment. After all, on remand the district judge may issue the same opinion. Similarly, decisions that were vacated after City of Boerne are presumably still helpful guides to applying strict scrutiny.
 It is difficult to establish causality with the few cases that pass both prongs of strict scrutiny. Nonetheless, to our knowledge, this is the first quantitative analysis of its kind in the post-Smith era.
 See Tabbaa v. Chertoff, 509 F.3d 89, 106 (2d Cir. 2007) (holding that “protecting against terrorism” allowed U.S. Customs and Border Protection to single out a group of Muslim men for heightened security screening). Professor Winkler found similar results. See Winkler, supra note 2, at 861. This is not to say Lukumi is a silver bullet. The law must first be found non-neutral or not generally applicable—and that is where the real battle occurs. See Lukumi, 508 U.S. at 531–32.
 See Volokh, Religious Exemptions, supra note 24, at 1505, 1508 (comparing certain forms of religious discrimination with content-based speech restrictions).
 Restrictions on free speech come in two flavors. “The first form of abridgement may be summarized as encompassing government actions aimed at communicative impact; the second, as encompassing government actions aimed at noncommunicative impact but nonetheless having adverse effects on communicative opportunity.” Tribe, supra note 22, at 790. The first is “presumptively at odds with the first amendment” and requires strict scrutiny. Id. The second is not presumptively invalid, but the Court “requires a ‘thumb’ on the scale to assure that the balance struck in any particular situation properly reflects the central position of free expression in the constitutional scheme.” Id. at 791. Therefore, an ordinance differentiating between “political,” “ideological,” and “directional” signs requires strict scrutiny. See Reed, 135 S. Ct. at 2226–27. But a blanket ban on all signs in public right-of-ways is on subject to intermediate scrutiny. See City Council of L.A. v. Taxpayers for Vincent, 466 U.S. 789, 817 (1984).
 The government’s asserted interest is found “not compelling as presented” in 45.5% of cases, compared to “never compelling” in 11% of cases. Supra Table 7; see also Jova v. Smith, 582 F.3d 410, 415 (2d Cir. 2009) (“Where a circuit court has found that a state’s identified interest was insufficiently compelling, this was often based on the court’s observation that the state failed to provide sufficient evidence explaining how the practices at issue furthered the stated interest”) (discussing prison cases).
 See supra Table 7.
 See, e.g., Salahuddin v. Goord, 467 F.3d 263, 279 (2d Cir. 2006) (“One can imagine that a prison official could supply [a government interest] (time constraints, the cost of personnel, etc.), but we are provided with none.”); Irshad Learning Ctr. v. Cnty. of Dupage, 937 F. Supp. 2d 910, 948 (N.D. Ill. 2013) (“The County has not argued that its decision . . . was justified by a compelling government interest.”); Falwell v. Miller, 203 F. Supp. 2d 624, 632 (W.D. Va. 2002) (stating the government “presented no governmental interest—compelling or otherwise”); W. Presbyterian Church v. Bd. of Zoning Adjust., 862 F. Supp. 538, 545–46 (D.D.C. 1994) (noting the government conceded that it had no compelling interest). Of course, the government has also occasionally stated interests that were deemed insufficient as a matter of law. See, e.g., In re Young, 82 F.3d 1407, 1420 (8th Cir. 1996) (holding that furthering “the interests advanced by the bankruptcy system are not compelling under the RFRA,” and therefore that tithes were protected from recovery in bankruptcy proceedings), vacated sub nom. Christians v. Crystal Evangelical Free Church, 521 U.S. 1114 (1997) (vacating and remanding for further consideration in light of City of Boerne, 521 U.S. 507 (1997)); Rocky Mountain Christian Church v. Bd. of Cnty. Comm’rs of Boulder Cnty., 612 F. Supp. 2d 1163, 1175 (D. Colo. 2009) (holding that the county’s asserted interests in preserving the current zoning system, “although legitimate in many senses, [did] not constitute compelling governmental interest.”).
 For example, there could never be a compelling interest in merely following a state statute: “Allowing state actors to escape liability by claiming that they have a ‘compelling state interest’ in implementing a state law that violates federal law [i.e., RFRA] would make the Supremacy Clause hollow indeed.” Bessard v. Cal. Cmty. Colls., 867 F. Supp. 1454, 1464 (E.D. Cal. 1994).
 See Volokh, Religious Exemptions, supra note 24, at 1488 (“A court might grant an exception based on its tentative conclusion that the exemption . . . won’t do much harm to the government interest; but a few years later, after evaluating the exemption in action, the court might change its view.”).
 See Fowler v. Crawford, 534 F.3d 931, 939 (8th Cir. 2008) (affirming that prison security is a compelling interest); cf. Holder v. Humanitarian Law Project, 561 U.S. 1, 28 (2010) (“Everyone agrees that the Government’s interest in combating terrorism is an urgent objective of the highest order.”) (free speech case).
 See Michael Stokes Paulsen, The Priority of God: A Theory of Religious Liberty, 39 Pepp. L. Rev. 1159, 1209 (2013) (“[A]s a society changes from a liberal religious society to a modern (or post-modern) irreligious society, intuitions shift too as to what are sufficiently compelling interests to trump religious exercise”); see also Yuval Levin, The Perils of Religious Liberty, First Things, Feb. 2016, at 29, 33 (arguing “the compulsive state religion, or at least our new civil religion, is supposed to be progressive liberalism.”). This change in religious affiliation can be seen in the increasing number of Americans who claim no religious affiliation. See Pew Research Center, America’s Changing Religious Landscape 4 (May 12, 2015) (reporting that 22.8% of Americans in 2014 were atheist, agnostic, or “nothing in particular”—up from 16.1% in 2007).
 In Hobby Lobby, both Justice Kennedy (in a concurring opinion) and Justice Ginsburg (in a dissent joined by Justices Breyer, Sotomayor, and Kagan) wrote that access to contraception is a compelling interest. 134 S. Ct. at 2786 (Kennedy, J., concurring) (“It is important to confirm that a premise of the Court’s opinion is its assumption that the HHS regulation here at issue furthers a legitimate and compelling interest in the health of female employees”); Hobby Lobby, 134 S. Ct. at 2799 (Ginsburg, J., dissenting) (“[T]he Government has shown that the contraceptive coverage for which the ACA provides furthers compelling interests in public health and women’s well-being. Those interests are concrete, specific, and demonstrated by a wealth of empirical evidence.”). Even the majority could not conclusively reject “guaranteeing cost-free access to . . . contraceptive methods” as a compelling interest. See Hobby Lobby, 134 S. Ct. at 2780. But see Catholic Benefits Ass’n LCA v. Burwell, 81 F. Supp. 3d 1269, 1276 (W.D. Okla. 2014) (concluding the Kennedy and Ginsburg opinions did not conclusively find this interest compelling).
 See Carey v. Population Servs. Int’l, 431 U.S. 678, 696 (1977) (plurality) (invalidating a ban on distributing contraceptives to minors); Eisenstadt v. Baird, 405 U.S. 438, 453 (1972) (same for unmarried persons); Griswold v. Connecticut, 381 U.S. 479, 485–86 (1965) (same for married couples).
 See, e.g., Rick Ungar, Understanding Why Indiana’s RFRA Clears The Way To Discriminating Against LGBT Americans, Forbes (Mar. 30, 2015, 12:51 P.M.), https://perma.cc/L3CR-X5RJ; Michael Barbaro & Erik Eckholm, Right to Deny Service to Gays Stirs Broad Uproar, N.Y. Times, Mar. 28, 2015, at A1.
 See Jonathan Cohn, Why Indiana’s Religious Freedom Law Is Such A Big Deal, The Huffington Post (updated Apr. 2, 2015), http://perma.cc/8GH9-LPXC (noting that the impact of Indiana’s RFRA depends, in part, on the definition of “compelling interest”).
 Andrew Koppleman, Gay Rights, Religious Accommodations, and the Purposes of Antidiscrimination Law, 88 S. Cal. L. Rev. 619, 636 (2015) (citing H. Amend. 5 to S.B. 101, 119th Gen. Assemb., Reg. Sess. (Ind. 2015)).
 See Roberts v. U.S. Jaycees, 468 U.S. 609, 623 (1984) (finding a “compelling interest in eradicating discrimination against . . . female citizens”); Bob Jones Univ., 461 U.S. at 604 (stating that “the Government has a fundamental, overriding interest in eradicating racial discrimination in education”); Grote v. Sebelius, 708 F.3d 850, 860 n.4 (7th Cir. 2013) (Rovner, J., dissenting) (stating there is a “compelling interest in eradicating discrimination.”). Bob Jones is the most applicable precedent, but it emphasized the special status of racial discrimination in education. See Bob Jones Univ., 461 U.S. at 595 (“Few social or political issues in our history have been more vigorously debated and more extensively ventilated than the issue of racial discrimination, particularly in education.”). It could therefore be distinguished on those lines, but that distinction is not likely to stick given that many compare protections for religious claimants to Jim Crow laws. See, e.g., Adam Winkler, Will the Supreme Court License Anti-Gay Discrimination?, The Huffington Post (Mar. 24, 2014 9:28 A.M.), http://perma.cc/5FEA-SM59 (“We’re already seeing a spate of proposals at the state level for Jim Crow-like laws that would permit discrimination against LGBT people.”).
 See U.S. Comm’n on Civil Rights, Peaceful Coexistence: Reconciling Nondiscrimination Principles with Civil Liberties 25, at 25–27 (September 2016), available at https://perma.cc/8ANP-ZTTY (asserting that religious exemptions to nondiscrimination statutes “significantly infringe upon . . . civil rights” and that RFRAs should be amended or interpreted to ensure they do not “unduly burden civil liberties and civil rights protections against status-based discrimination”); id. at 26, available at https://perma.cc/8ANP-ZTTY (“Overly broad religious exeptions unduly burden nondiscrimination laws and policies.”).
 Cf. Paul Horwitz, The Hobby Lobby Moment, 128 Harv. L. Rev. 154, 170 (2014) (“The consensus in favor of accommodation of religion . . . seems to have weakened, if not collapsed.”).
 Note, however, that the stated interest of “preserving Native American heritage” represented a statistically higher proportion of cases rejected on substantial burden grounds relative to other interests.
 See Volokh, Religious Exemptions, supra note 24, at 1500.
 See Lund, Perspective, supra note 6, at 299 n.62 (doubting the existence of such evidence). This study should specially consider the role of religious speech cases. See supra note 109.
 See Callahan v. Woods, 736 F.2d 1269, 1273 (9th Cir. 1984) (“A synthesis of the [compelling interest and least restrictive means] prongs is . . . the question whether the government has a compelling interest in not exempting a religious individual from a particular regulation.”) (emphasis added).