This Court has traditionally required a heightened showing for preliminary injunctions in three “disfavored” categories: injunctions that disturb the status quo, mandatory injunctions, and injunctions that afford the movant substantially all the relief it may recover at the conclusion of a full trial on the merits. SCFC ILC, Inc. v. Visa USA Inc., 936 F.2d 1096, 1098-99 (10th Cir.1991). We heard this case en banc to consider whether to jettison the heightened standard for preliminary injunctions that disturb the status quo. A majority of this Court has concluded that there are reasons — not fully accounted for in the balance of harms analysis — for courts to disfavor preliminary injunctions that disturb the status quo, and thus reaffirms our traditional rule (with slight modification and clarification). See Opinion of Murphy, J., at 976-980. A different majority has concluded that, even under the heightened standard, Appellee O Centro Espirita Beneficíente Uniao do Vegetal (“UDV”) is entitled to a preliminary injunction against enforcement of laws against the possession and use of its sacramental substance, hoasca. Opinion of Sey*1012mour, J., at 28. I write separately to explain why both halves of this holding, in my opinion, are correct.1
1. A Heightened Standard Should Apply to Preliminary Injunctions That Disturb the Status Quo
The Supreme Court has stated that preliminary injunctions have the “limited purpose” of “merely [preserving] the relative positions of the parties until a trial on the merits can be held.” University of Texas v. Camenisch, 451 U.S. 390, 395, 101 S.Ct. 1830, 68 L.Ed.2d 175 (1981). This emphasis on preserving the status quo is not the same as, and cannot be reduced to, minimizing irreparable harm to the parties during the pendency of litigation, as suggested by the dissent. See Opinion of Seymour, J., at 1001-1002. At the preliminary injunction stage, before there has been a trial on the merits, the function of the court is not to take whatever steps are necessary to prevent irreparable harm, but primarily to keep things as they were, until the court is able to determine the parties’ respective legal rights. That is why, in addition to the four preliminary injunction factors of harm to the movant, balance of harm, public interest, and likelihood of success on the merits, this Court has required district courts to take into account whether preliminary relief would preserve or disturb the status quo. The burden of justifying preliminary relief is higher if it would disturb the status quo. SCFC ILC, Inc., 936 F.2d at 1098-99.
There is no reason to think that the “general maxim” that “the purpose of a preliminary injunction is to preserve the status quo between the parties pending a full trial on the merits” is one that “should not be taken merely at face value” or disregarded except insofar as it “impacts the balance of harms between the parties and the public interest.” Opinion of Seymour, J., at 1002, 1003. A judicial version of Hippocrates’ ancient injunction to physicians — above all, to do no harm — counsels against forcing changes before there has been a determination of the parties’ legal rights. The settled rule of our tradition is that losses should remain where they fall until an adequate legal or equitable justification for shifting them has been demonstrated.
Traditional equity practice held that the sole purpose of a preliminary injunction was to preserve the status quo during the pendency of litigation. See, e.g., Farmers’ R.R. Co. v. Reno, Oil Creek & Pithole Ry. Co., 53 Pa. 224 (Pa.1866) (dissolving an injunction that blocked defendants from continuing to use certain land in their possession because the sole purpose of a preliminary injunction is to preserve the status quo); Chicago, St. Paul & Kansas City R.R. Co. v. Kansas City, St. Joseph & Council Bluffs R.R. Co., 38 F. 58, 60 (C.C.W.D.Mo.1889) (noting that a higher standard applies to mandatory injunctions that disrupt the status quo); New Orleans & North Eastern R.R. Co. v. Mississippi, Terre-aux Boeufs & Lake R.R. Co., 36 La.Ann. 561 (La.1884) (maintaining an injunction insofar as it maintained the status quo, but dissolving that portion that did not); Warner Bros. Pictures v. Gittone, *1013110 F.2d 292, 293 (3d Cir.1940) (per curiam) (“Irreparable loss resulting from refusal to accord the plaintiff a new status, as distinguished from interference with rights previously enjoyed by him, does not furnish the basis for interlocutory relief.”); Levy v. Rosen, 258 Ill.App. 262 (Ill.App.Ct.1930) (“An interlocutory order is usually granted to preserve the status quo, but the order in this appeal did not do that, but changed the status quo. The entry of such order was clearly erroneous.”); Gill v. Hudspeth County Conservation & Reclamation Dist. No. 1, 88 S.W.2d 517, 519 (Tex.Civ.App.1935) (“[T]he court’s discretion should be exercised against the writ if its issuance would change the status quo.”); Bowling v. Nat’l Convoy & Trucking Co., 101 Fla. 634, 135 So. 541 (1931) (“Since the object of a preliminary injunction is to preserve the status quo, the court will not grant such an order where its effect would be to change the status.”); Gates v. Detroit & Mackinac Ry. Co., 151 Mich. 548, 115 N.W. 420, 421 (1908) (dissolving that portion of a preliminary injunction that went beyond the status quo); Jones v. Dimes, 130 F. 638, 639 (D.Del.1904) (relaxing the burden on the moving party when the requested injunction merely maintained the status quo);1 James L. High, A Treatise on the Law of Injunctions (Chicago: Callaghan & Co. 1890, 3d ed.) § 4 at 5 (“The sole object of an interlocutory injunction is to preserve the subject in controversy in its then condition, and, without determining any questions of right”).
To be sure, it is sometimes necessary to require a party who has recently disturbed the status quo to reverse its actions. Such an injunction restores, rather than disturbs, the status quo ante, and is thus not an exception to the rule. “Status quo” does not mean the situation existing at the moment the law suit is filed, but the “last peaceable uncontested status existing between the parties before the dispute developed.” 11A Charles Alan Wright, Arthur R. Miller & Mary Kay Kane, Federal Practice and Procedure § 2948 (2d ed.1995).2 Thus, courts of equity have long issued preliminary injunctions requiring parties to restore the status quo ante. Shanaman v. Yellow Cab Co., 491 Pa. 516, 421 A.2d 664, 667 (1980) (reversing a preliminary injunction because “the purpose of a mandatory preliminary injunction is to restore the status quo” and the injunction actually disrupted that status); Morgan v. Smart, 88 S.W.2d 769, 772 (Tex.Civ.App.1935) (“[Tjhere are no real exceptions to the rule that the status quo will not be disturbed by a preliminary injunction, and when by such an injunction the possession of property is properly ordered to be restored it is not to disturb the status quo, but to avoid mistaking the true status and to avoid preserving a false one.”).
In recent decades, most courts — and all federal courts of appeal — have come to recognize that there are cases in which preservation of the status quo may so clearly inflict irreparable harm on the movant, with so little probability of being upheld on the merits, that a preliminary injunction may be appropriate even though it requires a departure from the status quo. See, e.g., Canal Authority v. Callaway, 489 F.2d 567, 576 (5th Cir.1974).3 *1014But preliminary injunctions that disturb the status quo, while no longer categorically forbidden, remain disfavored. Only one federal court of appeals has concluded that courts should simply strive to minimize irreparable harm, with no special attention to the status quo, as our dissenters suggest. United Food & Commercial Workers Union, Local 1099 v. Southwest Ohio Reg’l Transit Auth., 163 F.3d 341, 348 (6th Cir.1998); see Opinion of Seymour, J., at 1004.4
There are sound reasons of jurisprudence in support of the traditional view that preliminary injunctions that disturb the status quo require heightened justification. A preliminary injunction of any sort is an “extraordinary” and “drastic” remedy. See United States ex rel. Potawatomi Indian Tribe v. Enter. Mgmt. Consultants, Inc., 883 F.2d 886, 888-89 (10th Cir.1989). Judicial power is inseparably connected with the judicial duty to decide cases and controversies by determining the parties’ legal rights and obligations. See Marbury v. Madison, 5 U.S. (1 Cranch) 137, 177, 2 L.Ed. 60 (1803). A preliminary injunction is remarkable because it imposes a constraint on the enjoined party’s actions in advance of any such determination. That is, a preliminary injunction forces a party to act or *1015desist from acting, not because the law requires it, but because the law might require it. This is all the more striking because, given that many preliminary injunctions must be granted hurriedly and on the basis of very limited evidence, deciding whether to grant a preliminary injunction is normally to make a choice under conditions of grave uncertainty. See Heideman v. South Salt Lake City, 348 F.3d 1182, 1188 (10th Cir.2003).
It is one thing for a court to preserve its power to grant effectual relief by preventing parties from making unilateral and irremediable changes during the course of litigation, and quite another for a court to force the parties to make significant alterations in their practices before there has been time for a trial on the merits. See, e.g., Gittone, 110 F.2d at 293 (“[T]he effect of the preliminary injunction which the court granted was not to preserve the status quo but rather to alter the prior status of the parties fundamentally. Such an alteration may be directed only after final hearing.”); In re Marriage of Schwartz, 131 Ill.App.3d 351, 86 Ill.Dec. 698, 475 N.E.2d 1077, 1079 (1985) (“It is not the purpose of the preliminary injunction to determine controverted rights or decide the merits of the case.... A preliminary injunction is merely provisional in nature, its office being merely to preserve the status quo until a final hearing on the merits.”).
Moreover, preserving the status quo enables the court to stay relatively neutral in the underlying legal dispute. The restrictions placed on the parties can be understood as requiring only that they act in a manner consistent with the existence of a good-faith dispute about the relevant legal entitlements. The moving party is not given any rights, even temporarily, that would normally be his only if the legal dispute were resolved in his favor. For example, ownership disputes often raise concerns that the defendant in possession would overuse or waste the property before a complainant could regain possession through legal proceedings. Under those circumstances, equitable courts regularly enjoin the waste, ordering the defendant to preserve the property in statu quo. The general rule, however, is that except in the most exceptional cases, a court of equity cannot go beyond the status quo by putting the moving party into possession of the disputed property, even though, presumably, being deprived of the interim ability to enjoy the property would often constitute irreparable harm. See, e.g., Farmers’ R.R. Co., supra; Morgan v. Smart, 88 S.W.2d 769, 771 (Tex.Civ.App.1935) (“It is not the function of a preliminary injunction to transfer the possession of land from one person to another pending an adjudication of the title, except in cases in which the possession has been forcibly or fraudulently obtained ... [and the injunction is necessary so that] the original status of the property [may] be preserved pending the decision of the issue.”), quoting Simms v. Reisner, 134 S.W. 278, 280 (Tex.Civ.App.1911). See generally Mandatory injunction prior to hearing of case, 15 A.L.R.2d 213, §§ 22-23 (collecting dozens of cases on this issue).
Fundamentally, the reluctance to disturb the status quo prior to trial on the merits is an expression of judicial humility. As Judge Murphy points out, a court bears more direct moral responsibility for harms that result from its intervention than from its nonintervention, and more direct responsibility when it intervenes to change the status quo than when it intervenes to preserve it. See Opinion of Murphy, J., at 978. Moreover, like the doctrine of stare decisis, preserving the status quo serves to protect the settled expectations of the parties. Disrupting the status quo may provide a benefit to one party, but only by *1016depriving the other party of some right he previously enjoyed. Although the harm and the benefit may be of equivalent magnitude on paper, in reality, deprivation of a thing already possessed is felt more acutely than lack of a benefit only hoped for. As the Supreme Court observed in Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 282-83, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986), “[djenial of a future employment opportunity is not as intrusive as loss of an existing job.” Percipient students of human nature have often made similar observations. David Hume, for example, wrote:
Such is the effect of custom, that it not only reconciles us to any thing we have long enjoy’d, but even gives us an affection for it, and makes us prefer it to other objects, which may be more valuable, but are less known to us. What has long lain under our eye, and has often been employ’d to our advantage, that we are always the most unwilling to part with; but can easily live without possessions, which we never have enjoy’d, and are not accustom’d to.
David Hume, A Treatise of Human Nature, bk. 3, pt. 2, § 3, para. 4 (1739). See also, e.g., Aristotle, Nichomachean Ethics, bk. IX, ch. 1, at 1164M7-19 (W.D. Ross trans.), in The Basic Works of Aristotle (Richard McKeon ed., 1941) (“For most things are not assessed at the same value by those who have them and those who want them; each class values highly what is its own.... ”). Justice Holmes has justified the doctrine of adverse possession on these grounds:
[Tjhe foundation of the acquisition of rights by lapse of time is to be looked for in the position of the person who gains them, not in that of the loser.... A thing which you have enjoyed and used as your own for a long time, whether property or an opinion, takes root in your being and cannot be torn away without your resenting the act and trying to defend yourself, however you came by it. The law can ask no better justification than the deepest instincts of man.
Oliver Wendell Holmes, Jr., The Path of the Law, 10 Harv. L.Rev. 457, 477 (1897).
Notwithstanding the tendency of those trained in economics to view opportunity costs as equivalent to actual expenditures, modern social science research has confirmed the reality of “loss aversion” (the tendency to attach greater value to losses than to foregone gains of equal amount) and the closely related “endowment effect” (the tendency to value already possessed goods more than prospective acquisitions). See, e.g., Daniel Kahneman, Jack L. Knetsch & Richard H. Thaler, The Endowment Effect, Loss Aversion, and Status Quo Bias, 5 J. Econ. Persp. 193 (1991); Amos Tversky & Daniel Kahneman, Loss Aversion in Riskless Choice: A Reference-Dependent Model, 106 Q.J. Econ. 1039 (1991); Daniel Kahneman et al., Experimental Tests of the Endowment Effect and the Coase Theorem, 98 J. Pol. Econ. 1352 (1990); Jack L. Knetsch & J.A. Sin-den, Willingness to Pay and Compensation Demanded: Experimental Evidence of an Unexpected Disparity in Measures of Value, 99 Q.J. Econ. 507, 512-13 (1984). To take one of many illustrations, one study found that duck hunters would pay, on average, $247 to obtain the privilege of keeping a particular wetland undeveloped, but if they already had the right to block development, they would demand an average of $1,044 to give it up. Judd Ham-mack & Gardner M. Brown, Jr., Waterfowl and Wetlands: Toward Bioeconomic Analysis 26 (1974).
Moreover, adverse disruptions in the status quo carry along with them the cost and difficulty associated with adjusting to change. These involve not only direct *1017transition costs but also the costs associated with uncertainty, which manifest themselves in a reluctance to invest human or other capital in an enterprise where the returns could disappear at the drop of a judicial hat. Disruption is expensive. When a court requires a change in the status quo only to find that its grant of preliminary relief was mistaken and must be undone, the process is twice as disruptive as when the court preserves the status quo on a preliminary basis and later issues a final judgment requiring the change.
The status quo is also relevant to the credibility of the parties’ claims of irreparable harm. It is difficult to measure irreparable harm, and either party’s willingness to put up with a situation in the past can serve as an indication that the party’s injury is not as serious as alleged, or that the party has implicitly consented to the supposed injury. See Heideman, 348 F.3d at 1191 (“[T]he City has tolerated nude dancing establishments for many years.... This invites skepticism regarding the imperative for immediate implementation [of a new ordinance].”); Majorica, S.A v. R.H. Macy & Co., 762 F.2d 7, 8 (2d Cir.1985) (noting that while delay alone is not enough to constitute laches, it is ground for doubting a claim of irreparable harm). Plaintiffs, especially, have the burden of complaining of injuries promptly, before defendants come to rely on the status quo. “[E]quity aids the vigilant, not those who slumber on their rights.” Allred v. Chynoweth, 990 F.2d 527, 536 n. 6 (10th Cir.1993); Standard Oil Co. of N.M. v. Standard Oil Co. of Cal., 56 F.2d 973, 975 (10th Cir.1932); Natural Res. Defendant Council v. Pena, 147 F.3d 1012, 1026 (D.C.Cir.1998). Thus, when a plaintiff is complaining of irreparable injury from a long-established state of affairs, a court may naturally ask why, if the injury is so pressing as to warrant preliminary relief, the plaintiff waited so long before bringing a claim. 11A Charles Alan Wright, Arthur R. Miller & Mary Kay Kane, Federal Practice and Procedure § 2946, at 113-16 (2d ed.1995); Edward & John Burke v. Bishop, 144 F. 838, 839 (2d Cir.1906); Savage v. Port Reading R.R. Co., 73 N.J. Eq. 308, 67 A. 436, 438 (N.J.Ch.1907).
The status quo is also a useful reference point because litigants often have incentives to engage in counterproductive strategic behavior. A defendant facing the loss of property, for example, has a natural incentive to extract as much of the value of the land as possible before losing possession, even in ways that limit the land’s productivity for years to come. And even when doing so produces no advantages to the defendant, it is an unfortunate reality of human nature that many defendants would prefer to destroy the property in question than to let their adversary have the use of it, both out of spite and as a way of making the resort to the courts. less attractive in the first place.
Likewise, plaintiffs have incentives to seek injunctions not only to avert irreparable harm to themselves, but also to impose costs on the other party. This, too, may be done out of spite, or because the higher the costs to the defendant in complying, the more pressure he will feel to “bargain desperately to buy his way out of the injunction.” Am. Hosp. Supply Corp. v. Hosp. Prods., Ltd., 780 F.2d 589, 594 (7th Cir.1986). A preliminary injunction aims in part at achieving temporary peace between the parties. However, if it substantially shifts the lines of conflict, it is more likely to function as a weapon in the plaintiffs arsenal than as a cease-fire. Preserving the last peaceable uncontested status of the parties maintains a position to which both parties at least tacitly consented before their dispute, and its concomitant perverse incentives, arose.
*1018Without a heightened standard, these concerns will likely not be given due weight. In the context of the balance of harms analysis, it is all too easy to stop at comparing the absolute magnitudes of the parties’ irreparable harms, without distinguishing between foregone gains and actual losses, and without considering whether granting an injunction implicates other institutional concerns about the proper role of the courts. Unless the district court self-consciously takes the nature of the injunction into account by applying a heightened standard, the four factors likely will lead to an overconfident approach to preliminary relief, increasing the cost and disruption from improvidently granted preliminary injunctions.
A particularly important category of cases where the status quo will often be determinative of whether a court should provide preliminary relief is challenges to the constitutionality of statutes. When a statute is newly enacted, and its enforcement will restrict rights citizens previously had exercised and enjoyed, it is not uncommon for district courts to enjoin enforcement pending a determination of the merits of the constitutional issue. See, e.g., Eagle Books, Inc. v. Ritchie, 455 F.Supp. 73, 77-78 (D.Utah 1978); Reproductive Services v. Keating, 35 F.Supp.2d 1332, 1337 (N.D.Okla.1998); ACLU v. Johnson, 194 F.3d 1149, 1152 (10th Cir.1999); Utah Licensed Beverage Ass’n v. Leavitt, 256 F.3d 1061, 1076-77 (10th Cir.2001); Elam Constr., Inc. v. Reg’l Transp. Dist., 129 F.3d 1343, 1347-48 (10th Cir.1997) (per curiam). When a statute has long been on the books and enforced, however, it is exceedingly unusual for a litigant who challenges its constitutionality to obtain (or even to seek) a preliminary injunction against its continued enforcement. See, e.g., Walters v. Nat’l Ass’n of Radiation Survivors, 468 U.S. 1323, 1324, 105 S.Ct. 11, 82 L.Ed.2d 908 (Rehnquist, Circuit Justice 1984) (“It would take more than the respondents have presented in their response ... to persuade me that the action of a single District Judge declaring unconstitutional an Act of Congress that has been on the books for more than 120 years should not be stayed.... ”). This is not because the balance of harms to the litigants is different. Presumably, the loss of constitutional rights from enforcement of an old statute is no less harmful or irreparable than from enforcement of a new. The dissent’s suggested approach of considering the status quo only insofar as it bears on “the process of balancing the various interests and harms among the parties and the public,” (Opinion of Seymour, J., at 1002), without a heightened standard, is thus likely to yield the conclusion that it does not matter whether the statute is old or new. That would be a dramatic change in our practice. The reason for weighing the status quo is not to be found in the four preliminary injunction factors. It is rooted, instead, in the institutional concerns we have canvassed above.
I thus join in the en banc court’s decision to continue to require litigants seeking a preliminary injunction, that would alter the status quo, to meet a heightened burden of justification.
II. Does this Preliminary Injunction Satisfy the Heightened Standard?
This case satisfies even the heightened standard for preliminary injunctions. The applicable statute, the Religious Freedom Restoration Act, 42 U.S.C. § 2000bb-l(b), sets a most demanding burden of proof for the government: the compelling interest test. The factual findings of the district court, which are not challenged on appeal, make it clear that the government has not and cannot meet that burden on this record, and that the balance of equities is *1019overwhelmingly in favor of the movant. The en banc majority is therefore right, in my opinion, to affirm the district court’s grant of a preliminary injunction.
Plaintiffs establish, and the government does not dispute, that enforcement of the CSA in this context would impose a substantial burden on a sincere exercise of religion. It is common ground that such a burden constitutes irreparable injury. The plaintiffs have thus established a pri-ma facie case (relevant to the probability of success on the merits) and an irreparable injury (relevant to the balance of harms). It is also common ground that the evidence at the hearing regarding the government’s assertions of an interest in the health of hoasca users and the prevention of diversion to recreational drug users was in “equipoise” and “virtually balanced.” What is not common ground is the effect of evenly-balanced evidence regarding possible harms from hoasca use on UDVs ultimate likelihood of success on the merits, and on the balancing of the equities required for the grant of a preliminary injunction.
A
The dissent insists that the government is more likely to prevail on the merits than is UDV. In Judge Murphy’s formulation, the government’s interest in the uniform enforcement of drug laws and its interest in full compliance with the obligations imposed by international treaties are sufficient to meet the compelling interest standard. He is silent on whether, even if the government’s interests in enforcement and compliance were adjudged compelling, the government has employed the least restrictive means at its disposal, as RFRA requires. 42 U.S.C.2000bb-l(b)(2).
The dissent is premised on the view that “RFRA was never intended to result in [a] case-by-case evaluation of the controlled substances laws, and the scheduling decisions made pursuant to those laws .... [i]t is particularly improper for the court to assume such a function in this case.” Opinion of Murphy, J., at 18. On the contrary, that is precisely what RFRA instructs courts to do. The dissent does not make clear whether it interprets RFRA as precluding “case-by-ease evaluation” in all contexts, or whether this is a special rule for controlled substance cases. Neither interpretation is tenable.
In cases where federal law “substantially burdens” the exercise of religion, RFRA requires courts to determine whether “application of the burden” to a specific “person” is in furtherance of a compelling governmental interest and is the least restrictive means of furthering that interest. 42 U.S.C. § 2000bb-l(b) (emphasis added). That cannot be done without a case-by-case evaluation. “Thus, under RFRA, a court does not consider the ... regulation in its general application, but rather considers whether there is a compelling government reason, advanced in the least restrictive means, to apply the ... regulation to the individual claimant.” Kikumura v. Hurley, 242 F.3d 950, 962 (10th Cir.2001) (Murphy, J.). Accordingly, contrary to the dissent, Congress’s general conclusion that DMT is dangerous in the abstract does not establish that the government has a compelling interest in prohibiting the consumption of hoasca under the conditions presented in this case.
Nor is there an implied exemption from RFRA in cases involving the controlled substances laws. By its terms, RFRA applies to “all Federal law, and the implementation of that law, whether statutory or otherwise, and whether adopted before or after [enactment of RFRA],” unless the law “explicitly excludes such application by reference to this chapter.” 42 U.S.C. *1020§ 2000bb — 3(a), (b). The CSA contains no such explicit exception.
Judge Murphy argues that “courts simply lack the institutional competence to craft a set of religious exemptions to the uniform enforcement” of the drug laws. Opinion of Murphy, J., at 983. But the same may be said for application of RFRA to virtually any field of regulation that may conflict with religious exercise. Whatever our justifiably low opinion of our own competence, we are not free to decline to enforce the statute, which necessarily puts courts in the position of crafting religious exemptions to federal laws that burden religious exercise without sufficient justification.
The dissent’s notion that the drug laws are impliedly exempt from RFRA scrutiny is especially surprising in light of the fact that the impetus for enactment of RFRA was the Supreme Court’s decision in a case involving the sacramental use of a controlled substance. See Congressional Findings and Declaration of Purposes, 42 U.S.C. § 2000bb(a)(4) (criticizing Employment Division v. Smith, 494 U.S. 872, 110 S.Ct. 1595, 108 L.Ed.2d 876 (1990)). It may well be that most examples of enforcement of the drug laws will satisfy strict scrutiny under RFRA, see id. at 903-07, 110 S.Ct. 1595 (O’Connor, J., concurring) (applying strict scrutiny to, and upholding, the application of Oregon drug laws to the Native American Church’s sacramental use of peyote), but it can scarcely be clearer that Congress intended such scrutiny to occur.
The dissent asserts that courts applying the compelling interest test both before and after RFRA have “routinely rejected religious exemptions from laws regulating controlled substances,” and that “the same result should obtain in this case.” Opinion of Murphy., J., at 984-985 (citing cases). There is no support in the cases cited, however, for the proposition that any religious use of any drug is outside the scope of RFRA (or, before Smith, free exercise) protection. Four of the five pre-RFRA cases cited involve the same group, the Ethiopian Zion Coptic Church, which advocated the use of marijuana “continually all day, through church services, through everything [they] do.” Olsen v. Drug Enforcement Admin., 878 F.2d 1458, 1459 (D.C.1989). The constant and uncircum-scribed use of a drug presents different health risks and risks of diversion than the use of hoasca suggested by UDV. The significance of these differences is underscored by the conviction of the Ethiopian Zion Coptic Church for the importation of twenty tons of marijuana. United States v. Rush, 738 F.2d 497, 501 (1st Cir.1984). The post-RFRA cases cited offer no more support for the proposition that the findings of the CSA will always outweigh the interest in a particular religious use. In U.S. v. Brown, 1995 WL 732803, *2, for example, the Eighth Circuit found that the “broad use” of marijuana advocated by the church in question, which included supplying the drug to the sick and distributing it to anyone who wished it, including children with parental permission, made accommodation impossible. Both the unconstrained character of the proposed use and the popularity of marijuana affected the outcome in these cases: “the vast difference in demand for marijuana on the one hand and peyote on the other warranted the DEA’s response [in declining to grant an exception.]” Olsen v. DEA at 1463-64. These cases accordingly provide very little insight into the appropriate result when the standard required by RFRA is applied to a case involving a tightly circumscribed use of a drug not in widespread use.
Even assuming RFRA’s compelling interest test applies, the dissent takes the position that “the government need turn *1021only to express congressional findings concerning Schedule I drugs” to satisfy RFRA scrutiny. Opinion of Murphy, J., at 983. The dissent cites no authority for such an approach, and there is none. Congressional findings are entitled to respect, but they cannot be conclusive. RFRA requires the government to “demonstrate[ ]” that application of a challenged federal law to religious exercise satisfies strict scrutiny under RFRA. 42 U.S.C. § 2000bb-l(b). The term “demonstrates” is defined as “meeting] the burdens of going forward with the evidence and of persuasion.” Id., § 2000bb-2(3). Obviously, Congress contemplated the introduction of “evidence” pertaining to the justification of “application” of the law in the particular instance. If such a burden of proof could be satisfied by citing congressional finding in the preambles to statutes, without additional evidence, RFRA challenges would rarely succeed; congressional findings invariably tout the importance of the laws to which they are appended.
The dissent points to two such congressional findings. First, Congress has made a general finding that the “illegal importation ... and possession and improper use of controlled substances have a substantial and detrimental effect on the health and general welfare of the American people.” Opinion of Murphy, J., at 983-984. Second, Congress has placed DMT on the list of Schedule I controlled substances, which implies that it “has high potential for abuse and is not safe to consume even under the supervision of medical personnel.” Id. These generalized expressions of the government’s interest in prohibiting hoasca are very similar to the sweeping statements of interest that the Supreme Court found wanting in Wisconsin v. Yoder, 406 U.S. 205, 92 S.Ct. 1526, 32 L.Ed.2d 15 (1972)-one of the cases to which Congress referred as illustrating the compelling interest test it wished to “restore” by means of RFRA. See § 2000bb(b)(l). In that case, the Supreme Court rejected the State of Wisconsin’s “contention that its interest in its system of compulsory education is so compelling that even the established religious practices of the Amish must give way”:
Where fundamental claims of religious freedom are at stake, however, we cannot accept such a sweeping claim; despite its admitted validity in the generality of cases, we must searchingly examine the interests that the State seeks to promote by its requirement for compulsory education to age 16, and the impediment to those objectives that would flow from recognizing the claimed Amish exemption.
406 U.S. at 221, 92 S.Ct. 1526 (emphasis added). A similarly “searching examination” is required here, and can no more be satisfied by quotation of “sweeping claim[s]” in statutory preambles than it could in Yoder.
If Congress or the executive branch had investigated the religious use of hoasca and had come to an informed conclusion that the health risks or possibility of diversion are sufficient to outweigh free exercise concerns in this case, that conclusion would be entitled to great weight. But neither branch has done that. The two findings on which the dissent relies address the broad question of the dangers of all controlled substances, or all Schedule I substances, in the general run of cases. Such generalized statements are of very limited utility in evaluating the specific dangers of this substance under these circumstances, because the dangers associated with a substance may vary considerably from context to context.
Congress itself recognized this and gave the Attorney General authority to make exemptions, from many of the CSA’s requirements:
*1022The Attorney General may, by regulation, waive the requirement for registration of certain manufacturers, distributors, or dispensers if he finds it consistent with the public health and safety.
21 U.S.C. § 822(d) (emphasis added). Thus, the CSA itself recognizes that, despite Congress’s general findings about Schedule I substances, it may sometimes be “consistent with the public health and safety” to exempt certain people from its requirements. Indeed, the government evidently believed this to be true with respect to the Native American Church’s peyote use, since it relied primarily on § 822(d) to authorize its regulation exempting the Native American Church from the CSA. See 21 C.F.R. § 1307.31 (“The listing of peyote as a controlled substance in Schedule I does not apply to the non-drug use of peyote in bona fide religious ceremonies of the Native American Church, and members of the Native American Church so using peyote are exempt from registration.” (emphasis added)).
Judge Murphy responds that 21 U.S.C. § 822(d) should not be construed as giving the Attorney General authority to exempt religious groups other than the Native American Church from registration without specific authorization from Congress, because the “government’s regulatory exemption for peyote ... was at all times a product of congressional will.” Opinion of Murphy, J., at 24. I think he is wrong about the scope of the Attorney General’s authority under § 822(d),5 but that is not the point. Even if in practice the only religious exemption authorized by § 822(d) were for the Native American Church, the plain text of that provision indicates Congress’s belief that at least some use of substances controlled by the Act are “consistent with the public health and safety,” despite the generalized congressional finding that any Schedule I substance is not safe to consume even under the supervision of medical personnel. 21 U.S.C. § 812(b)(1)(C). More recently, Congress has passed legislation requiring the states to allow the Native American Church to use peyote, a Schedule I substance, in religious ceremonies. See American Indian Religious Freedom Act Amendments of 1994, 42 U.S.C. § 1996a. Congress’s consistent position has been that concerns for religious freedom can sometimes outweigh risks that otherwise justify prohibiting Schedule I substances. Neither Congress nor the Executive has treated the CSA’s general findings about Schedule I substances as precluding a particularized assessment of the risks involved in a specific sacramental use. Neither should we.
Several factors make hoasca atypical in its likely health consequences. For instance, although DMT is typically taken intravenously or inhaled in the nonreligious settings that Congress presumably had in mind when it proscribed the substance, UDV members ingest it orally. There was some evidence at the hearing that the resulting doses are considerably smaller than typical intravenous or inhaled *1023doses, and there has been very little study of the effects of orally ingested DMT. Furthermore, the fact that hoasca is a relatively uncommon substance used almost exclusively as part of a well-defined religious service makes an exemption for bona fide religious purposes less subject to abuse than if the religion required its constant consumption, or if the drug were a more widely used substance like marijuana or methamphetamine. Cf. Employment Div. v. Smith, 494 U.S. 872, 913-14, 110 S.Ct. 1595, 108 L.Ed.2d 876 (1990) (Blackmun, J., dissenting). These and other differences undermine any claim that, in placing DMT on Schedule I, Congress made a factual finding that should control our assessment of the relative dangerousness of hoasca.
Judge Murphy expresses disbelief that a claimant’s rights under RFRA could “turn on whether the adherent has a religious affinity for street drugs or more esoteric ones.” Opinion of Murphy, J., at 987. Of course it is true that in theory, at least, it is possible to have the same religious interest in shooting heroin as in drinking hoasca. But one’s rights under RFRA depend not only on the nature of the religious interest but also on the strength of the government’s opposed interest. Here, the government’s professed interests include avoiding diversion to nonreligious use and ensuring that a multitude of spurious free exercise claims do not hamstring its enforcement efforts. Given those concerns, I do not see why Judge Murphy finds it surprising that the extent of nonreligious use is relevant to the analysis. Indeed, it would be far more surprising if the differences between street drugs and more “esoteric” ones were irrelevant. See Olsen v. DEA, 878 F.2d 1458, 1464 (D.C.Cir.1989) (R. Ginsburg, J.) (“[W]e rest our decision [not to grant an exemption for religious marijuana consumption] on the immensity of the marijuana control problem in the United States.... ”).
Finally, the dissent also urges that the government’s interest in strict compliance with the 1971 United Nations Convention on Psychotropic Substances, Feb. 21, 1971, 32 U.S.T. 543 (the “Convention”) is sufficiently compelling to outweigh the burden imposed on UDV. The district court held that the Convention does not apply to the hoasca tea used by UDV. O Centro Espirita Beneficiente Uniao Do Vegetal v. Ashcroft, 282 F.Supp.2d 1236, 1269 (D. New Mexico 2002). Judge Murphy categorically asserts the opposite, based on the “plain language of the Convention.” Opinion of Murphy, J., at 989.
To reverse on the basis of the Convention would require us to go far beyond what the record can support. After reviewing the initial briefs filed by the parties, the district court determined that the government’s strongest grounds for prohibiting UDV from using hoasca were based on concerns about the safety of drinking the tea and the risk of diversion to non-religious uses. 282 F.Supp.2d at 1266. The court therefore limited evidence to those issues. Plaintiffs attempted to present evidence regarding the interpretation of the Convention by the International Narcotics Control Board, the international enforcing agency, including a letter by the Secretary of the Board stating that hoasca is not controlled under the Convention. The government objected on the ground that “We are now introducing testimony about whether or not aya-huasca is controlled under the International Convention. That is not one of the issues in this hearing.” Supp.App. 1634. After discussion, the district court forbade questioning on the subject, and plaintiffs were unable to introduce evidence on the interpretation of the Con*1024vention by the Board. For this Court to attempt to interpret a complex treaty on the basis of its “plain language,” without the benefit of its interpretive history, would be premature.
More to the point, the government utterly failed to carry its statutory burden (42 U.S.C. § 2000bb-l(b)(2)) of demonstrating that complete prohibition of hoasca is the “least restrictive means” of furthering its interest in compliance with the Convention, even assuming the Convention applies. Contrary to the dissent, neither the Convention’s terms nor the practice of its interpretation is without flexibility when religious and other constitutional countervailing interests are implicated. For example, the CSA provides a mechanism by which the government may protest a scheduling decision made under Article 2 of the Convention. When the government receives notice of a scheduling decision pursuant to Article 2 of the Convention, if the requirements demanded are not met by existing controls, the Secretary of State may “ask for a review by the Economic and Social Council of the United Nations” or “take appropriate action under the Convention to initiate proceedings to remove the drug or substance from the schedules under the Convention or to transfer the drug or substance to a schedule under the Convention different from the one specified in the schedule notice.” 21 U.S.C. § 811 (c)(3)(C)(iii) & (iv). Article 2 of the Convention creates a process for a signatory state to request a reconsideration of a scheduling decision already made, and in considering that request, the Commission is permitted to take into account “economic, social, legal, administrative and other factors it may consider relevant.” Article 2(1), (5), (6). The availability of these procedures suggests that compliance with the Convention is not wholly inconsistent with the needs of signatory states to tailor some scheduling decisions to local requirements.
The Convention allows signatory states at the time of signature, ratification, or accession to make a reservation for indigenous plants traditionally used by “small, clearly determined groups in magical or religious rites.” Article 32(4). To interpret the Convention rigidly, as having no possibility of accommodation for new religious groups (or groups newly arriving in the United States), for which no reservation was sought at the time, raises troubling constitutional concerns of denominational discrimination. See Olsen, 878 F.2d at 1461. We should not lightly assume this is the correct interpretation of the Convention.
In the case of peyote, as the district court pointed out, 282 F.Supp.2d at 1268, the United States permits the exportation of the substance to Native American Church groups in Canada, despite the fact that exportation of a Schedule I substance for other than scientific or medical purposes would appear to violate the Convention.6 This suggests that, in practice, there is room for accommodation of the legitimate needs of religious minority groups.
RFRA places the burden on the government to demonstrate that application of the law to the particular religious exercise is the least restrictive means of furthering its interest. As far as the government’s argument and the record reveal, the government has undertaken no steps to inquire regarding the status of hoasca or to *1025work with the Economic and Social Council or the International Narcotics Control Board to find an acceptable accommodation. Rather, it has posited an unrealistically rigid interpretation of the Convention, attributed that interpretation to the United Nations, and then pointed to the United Nations as its excuse for not even making an effort to find a less restrictive approach.
To be sure, treaty compliance might well implicate governmental interests beyond the health and safety interests considered above. For example, if it could be shown that if the United States failed to proscribe hoasca, another country would seize upon that as an excuse to refuse to proscribe another controlled substance of great importance to our national well-being, that might well constitute a compelling interest. But there is no way to know whether that is so without asking.
The government submitted the affidavit of one State Department lawyer stating in general terms that noncompliance with the treaty would interfere with the ability of the United States to demand cooperation from other nations. But while some level of deference to Congressional and Executive findings is appropriate in the context of foreign relations, this affidavit does not provide any information specific enough to be relevant in assessing the damage that would flow from an exemption for the UDV. Presumably that lawyer did not mean to say that all violations, from the smallest infraction to blatant disregard for the treaty as a whole, are equally damaging to the diplomatic interests of the United States. He made no mention of whether the International Narcotics Control Board deems hoasca to be within the Convention or whether there may be ways to comply with the Convention without a total ban. Had the government presented an affidavit about the particular harms that this particular infraction would cause, it might be a different matter. See Ashcroft, — U.S. at -, 124 S.Ct. at 2794; Sable Communications v. FCC, 492 U.S. 115, 130, 109 S.Ct. 2829, 106 L.Ed.2d 93 (1989) (dismissing conclusory statements that a complete ban on dial-a-porn messages was necessary to protect children because “the congressional record ... contained] no evidence as to how effective or ineffective” less restrictive alternatives would be).
B
Even if UDV were likely to prevail on the merits, the dissent believes this to be one of those rare cases in which the balancing of the equities would dictate that the injunction not issue. See Opinion of Murphy, J., at 996. The disagreement rests, I think, on whether the statutory policies and burdens of proof set forth in RFRA should guide our consideration of each of the four preliminary injunction factors — or are relevant only to the first, the probability of success on the merits. I believe Judge Murphy’s dissent is wrong to disregard RFRA in balancing the equities. That is not because RFRA implicitly modifies the standards that apply to preliminary injunctions; I agree the normal standards remain in place unless Congress clearly manifests an intent to modify them. See Weinberger v. Romero-Barcelo, 456 U.S. 305, 320, 102 S.Ct. 1798, 72 L.Ed.2d 91 (1982). Rather, the point is that the normal standards for injunctive relief require courts to weigh the private and public interests in free exercise on the one hand against the government’s interests in regulation on the other, and RFRA is relevant to that weighing. When Congress has expressed its view of the proper balance between conflicting statutory policies, it is incumbent upon the courts to give effect to that view:
*1026‘Balancing the equities’ when considering whether an injunction should issue, is lawyers’ jargon for choosing between conflicting public interests. When Congress itself has struck the balance, has defined the weight to be given the competing interests, a court of equity is not justified in ignoring that pronouncement under the guise of exercising equitable discretion.
Youngstown Sheet & Tube Co. v. Sawyer, 343 U.S. 579, 609-610, 72 S.Ct. 863, 96 L.Ed. 1153 (1952) (Frankfurter, J., concurring).
By “restor[ing]” the compelling interest test of Sherbert v. Verner, 374 U.S. 398, 83 S.Ct. 1790, 10 L.Ed.2d 965 (1963), and Wisconsin v. Yoder, 406 U.S. 205, 92 S.Ct. 1526, 32 L.Ed.2d 15 (1972), 42 U.S.C. § 2000bb(b)(l), RFRA expressed Congress’s judgment that the free exercise of religion outweighs all but the most compelling governmental interests. See 42 U.S.C. § 2000bb-l; Yoder, 406 U.S. at 215, 92 S.Ct. 1526 (“The essence of all that has been said and written on the subject is that only those interests of the highest order and those not otherwise served can overbalance legitimate claims to the free exercise of religion.”). Once the plaintiff has established a prima facie case, RFRA places on the government the burden of demonstrating that application of the law is the least restrictive means of furthering its interest. 42 U.S.C. § 2000bb-l(b).7 It is not that RFRA “legislatively overrule[s]” the traditional principle that the moving party bears the burden of establishing the four preliminary injunction factors. See Opinion of Murphy, J., at 996. Rather, RFRA speaks to the quality of evidence and nature of the interest that the government must put forward. RFRA makes it clear that only demonstrated interests of a compelling nature are sufficient to justify substantial burdens on religious exercise. Mere “equipoise” with respect to not-necessarily-compelling governmental interests is not enough.
Thus, the dissent is wrong to assume that, with the evidence of the government’s interest in “equipoise,” the plaintiff “has not carried its burden of demonstrating that the third and fourth preliminary injunction factors ... weigh in its favor.” See Opinion of Murphy, J., at 983. The government’s evidence, on this record, demonstrates only that there might be some adverse health consequences or risks of diversion associated with UDV’s hoasca consumption. See Gov’t Br. 45 (describing the government’s interest as an interest in prohibiting substances that are “just as likely to be dangerous as ... safe”). But under RFRA, mere possibilities, based on limited evidence supplemented by speculation, are insufficient to counterbalance the certain burden on religious practice caused by a flat prohibition on hoasca. See United States v. Hardman, 297 F.3d 1116, 1130 *1027(10th Cir.2002) (en banc); Sherbert, 374 U.S. at 407, 83 S.Ct. 1790.
In effect, the dissent attempts to make an end run around RFRA’s reinstatement of strict scrutiny by repackaging all of the arguments that would be relevant to the merits (where the presumption of invalidity would clearly apply) as arguments about the equities (where it is disregarded). That approach is unprecedented. When the government fails to demonstrate its compelling interest in burdening a constitutional right, courts routinely find that, in the absence of a compelling justification for interference, the balance of harms and public interest also favor protecting the moving party’s burdened rights. See, e.g., Fifth Ave. Presbyterian Church v. City of New York, 293 F.3d 570, 576 (2d Cir.2002) (affirming the grant of a preliminary injunction because the City “ha[d] not sufficiently shoim the existence of a relevant law or policy ... that would ... justify its actions in dispersing the homeless from the Church’s landings and steps” (emphasis added)); Jolly v. Coughlin, 76 F.3d 468, 482-83 (2d Cir.1996) (applying a heightened standard but nevertheless upholding a RFRA-based preliminary injunction because the plaintiff had established a prima facie case and the government had not established that its policy was the least restrictive means of furthering a compelling interest); Eisenberg ex rel. Eisenberg v. Montgomery County Public Schools, 197 F.3d 123, 127 n. 11, 133 (4th Cir.1999) (reversing the denial of a preliminary injunction because the school district had not presented evidence sufficient to rebut the strict-scrutiny presumption that race-based decisions are invalid). See also Stuart Circle Parish v. Bd. of Zoning Appeals, 946 F.Supp. 1225, 1235-36, 1240 (E.D.Va.1996) (finding that because plaintiffs demonstrated a substantial likelihood of success on their RFRA claim, their interest in religious freedom tipped the balance of harms and the public interest in their favor); Luckette v. Lewis, 883 F.Supp. 471, 483 (D.Ariz.1995) (balance of harms weighed sharply in favor of prisoner given that his religious exercise was burdened and defendants had not demonstrated a countervailing public interest); Howard v. United States, 864 F.Supp. 1019, 1029 (D.Colo.1994) (in light of likelihood of success, public interest in protecting First Amendment rights outweighed any possible harm to the government); McCormick v. Hirsch, 460 F.Supp. 1337, 1350 (M.D.Pa.1978), abrogated on other grounds, see Bakery, Confectionery and Tobacco Workers’ Int’l Union, Local 6 v. NLRB, 799 F.Supp. 507, 511 (E.D.Pa.1992) (“When the protection of First Amendment liberties are [sic] involved, little else need be said of balancing the public interest, as protection of these rights is the most fundamental.”).8
If there was any doubt before, the Supreme Court’s recent opinion in Ashcroft v. ACLU, — U.S. -, 124 S.Ct. 2783, 159 L.Ed.2d 690 (2004), forecloses the dissent’s approach. Like this case, Ashcroft involved a preliminary injunction in which the merits were governed by the compelling interest/least restrictive means test. The issue there was the constitutionality of the Children’s Online Protection Act, (“COPA”), which requires businesses posting certain sexually explicit content on the web to require viewers to submit information verifying their age before they could access the materials. See id. at 2789-90, 124 S.Ct. 2783. The main question was whether that means of keeping the content *1028away from children was the least restrictive means, as compared with other methods (prominently, making internet filtering programs more readily available to parents). As in our case, there was evidence on both sides, and substantial factual questions remained about the relative effectiveness of the two alternatives. See id. at 2794. On that record, the Court found that the plaintiffs were likely to succeed primarily because the burden of proof was allocated to the government. See id. at 2791, 2793 (noting that movants had no burden to demonstrate the effectiveness of alternative means of serving the government’s interest; the government bore the burden of proving that other alternatives were less effective than COPA).
By the dissent’s logic, the Court should have gone on to reverse the district court’s preliminary injunction on the theory that with respect to the balance of harms and public interest prongs, it was not the government but the plaintiffs who bore the burden of proving that the COPA regime was not the least restrictive means of serving the government’s interests. In fact the Court did quite the opposite. In affirming the preliminary injunction, the Court had this to say about the equities supporting the injunction:
As mentioned above, there is a serious gap in the evidence as to the effectiveness of filtering software.... For us to assume, without proof, that filters are less effective than COPA would usurp the District Court’s factfinding role. By allowing the preliminary injunction to stand and remanding for trial, we require the Government to shoulder its full constitutional burden of proof respecting the less restrictive alternative argument, rather than excuse it from doing so.
See id. at 2794. The Court thus held that even with regard to the balance of harms, the government must “shoulder its full constitutional burden of proof respecting the less restrictive alternative argument.”9 Id. Under controlling Supreme Court precedent, therefore, we cannot “excuse” the government from meeting its burden simply by shifting the analysis from the likelihood of success to the equities.
C
Even putting aside any special features of RFRA or strict scrutiny more generally, there is a more basic problem with the dissenters’ approach. While Judge Murphy is correct to insist that UDV carry its burden with regard to each of the four factors of the preliminary injunction test, he underestimates the significance of the likelihood of success on the remaining factors, thereby misconceiving the relationship between the four preliminary injunction factors. A primary purpose of the balance-of-harms inquiry is to determine the relative cost of an error favoring one side as compared with an error favoring the other. See, e.g., Ashcroft, — U.S. -at -, 124 S.Ct. 2783, 159 L.Ed.2d 690, 2004 WL 1439998 at *9 (noting that “the potential harms from reversing the injunction outweigh those of leaving it in *1029place by mistake”); Tri-State Generation & Transmission Ass’n, Inc. v. Shoshone River Power, Inc., 805 F.2d 351, 358 (10th Cir.1986) (“In essence, it would be easier to correct a mistake in favor of Tri-State in issuing an injunction than it would be to correct a mistake in favor of Shoshone and Pacific by not issuing it.”). It follows that the balance-of-harms inquiry depends in part on the merits inquiry, since the only way of assessing whose harms are likely to be erroneously imposed is to judge them in light of the likelihood of success on the merits. Thus, no matter how great the interim harm to UDV if it is prevented from using hoasca until the final resolution of this case, that harm must be discounted to the extent that it is likely that UDV will not ultimately prove entitled to use hoasca; by the same token, no matter how great the interim harm to the government if it is wrongfully forced to allow the UDV to use and import hoasca, that harm must be discounted by the likelihood that UDV will ultimately prevail. Cf. Opinion of Seymour, J., at 1002-1003.
Although not always explicitly, courts commonly evaluate the balance of harms in light of the likelihood of success. See, e.g., Abbott Labs. v. Mead Johnson & Co., 971 F.2d 6, 12 & n. 2 (7th Cir.1992); Star Fuel Marts, LLC v. Sam’s East, Inc., 362 F.3d 639, 652 (10th Cir.2004) (downplaying the harm to the defendant because the defendant had not rebutted the plaintiffs prima facie case on the- merits, and therefore the preliminary injunction required the defendant to do no more than it was legally obligated to do). It may be possible, as the dissent suggests, for the harm and public interest factors to favor the party likely to lose on the merits so strongly that the (likely) losing party should succeed at the preliminary injunction stage. Such an outcome is highly unlikely, however, when the merits determination hinges on the strength of the governmental interest. In such cases, it is to be expected that the merits and the balance of equities would overlap. If the government’s interest is not strong enough to outweigh the plaintiffs interest in religious exercise for purposes of the merits, it is highly unlikely to do so for purposes of the balance of harms.
D
Besides insisting that UDV has not met its supposed burden of disproving the government’s interest, Judge Murphy’s dissent also suggests several substantive reasons for finding that the balance of harms favors the government. First, he relies on the government’s general interest in enforcing the law. See Opinion of Murphy, J., at 993-994, quoting New Motor Vehicle Bd. v. Orrin W. Fox Co., 434 U.S. 1345, 1351, 98 S.Ct. 359, 54 L.Ed.2d 439 (Rehnquist, Circuit Justice 1977). However, we must not forget that this case involves the intersection of two Acts of Congress of equal dignity: RFRA and the CSA. As a result, the government’s interest in complying with the law cuts both ways: the government has no less interest in obeying RFRA than it has in enforcing the CSA. Whether the public interest in enforcing the law favors accommodation under RFRA or strict application of the CSA depends on whether there is a compelling interest that requires strict enforcement of the latter. It would be circular to rely on that interest to establish the government’s compelling interest in the first place.
The government also stresses its interest in uniform enforcement of the law and avoiding the burdens of case-by-case management of religious exemptions, raising concerns that if UDV is allowed an exemption in this case, it will make enforcement of the CSA (and the Convention) unworkable by encouraging a host of spurious *1030claims for religious drug use. I find the panel opinion’s reasons for skepticism on this front convincing. See O Centro Espirita Beneficiente Uniao do Vegetal v. Ashcroft, 342 F.3d 1170, 1186 (10th Cir.2003). In any event, it is most unlikely that those fears will materialize during the pendency of a preliminary injunction. Assuming the government is entitled to enforce the CSA, a final judgment in its favor will serve as adequate discouragement for future claims similar to UDV’s. If the government is serious about the dangers, it can always seek expedited treatment of this case on the merits, and prove its case to the district court.
Finally, even when the government is able to demonstrate a compelling interest under RFRA, it remains necessary to establish that there is no other way of furthering that interest that would have less impact on the religious exercise. See Yoder, 406 U.S. at 215, 92 S.Ct. 1526 (requiring the government’s interest to be one “not otherwise served”); 42 U.S.C. § 2000bb — 1(b)(2). Thus, although the parties spend the bulk of their efforts arguing about whether the government has a compelling interest in prohibiting UDV’s use of hoasca, that is only part of the analysis. In United States v. Hardman, when this Court applied RFRA to a statutory scheme that allowed Native American tribe members to possess eagle parts but denied access to other practitioners of Native American religion, the Court en bane held that the government could not prevail without presenting evidence about the effects of alternative, less restrictive approaches on the compelling government interests in question. 297 F.3d at 1132. “[W]e must first determine where along [the continuum of policy alternatives] the government’s present solution lies, and where other, less restrictive means would lie.” Id. at 1135.
This case, like Hardman, raises the question of why an accommodation analogous to that extended to the Native American Church cannot be provided to other religious believers with similar needs. As the panel majority noted, the apparent workability of the accommodation for Native American Church peyote use strongly suggests that a similar exception would adequately protect the government’s interests here. See O Centro, 342 F.3d at 1186. The preliminary injunction issued in this case allows the government some degree of control over UDV’s importation, storage, and use of hoasca. At least to some extent, then, the preliminary injunction works a compromise, attempting to respond to the government’s legitimate concerns while still allowing UDV to continue its religious activity. It is incumbent on the government to show why no such compromise regime could adequately serve its interests.
E
All told, this is the unusual case in which the plaintiff not only prevails on each of the four preliminary injunction factors, but does so with sufficient clarity that a preliminary injunction is warranted even though it would disturb the status quo. The dissent does not challenge that the plaintiff would suffer serious and irreparable injury from continued prohibition of its religious sacrament. With the burden on the government to prove that its interest in enforcing the CSA against religious hoasca use is compelling but the evidence in support of that interest no better than “in equipoise,” the plaintiff has also demonstrated a likelihood of success on the merits. The same state of the record demonstrates conclusively that the plaintiff prevails on the other two factors. With a proven interest of high order on one side, and mere uncertainty, or “equipoise,” on *1031the other, the balance of equities is plainly in the plaintiffs favor. And in light of Congress’s determination that the public interest is served by accommodating the free exercise of religion except in cases of a proven compelling governmental interest, the plaintiff prevails on the “public interest” prong as well.
In conclusion, courts should issue preliminary injunctions that disturb the status quo only when the traditional balance is strongly in the plaintiffs favor, but on this record, plaintiff UDV has satisfied that demanding test.
. Judges Seymour and Murphy have each written opinions that concur in part of the holding of the en banc court and dissent from the other part. For convenience, I will refer to those portions of these opinions that dissent from the en banc holding as a "dissent,” and to those portions that concur in the holding as a "concurrence.” I join the per cu-riam opinion in its entirety. I join Part I of Judge Murphy’s separate opinion, and Part II of Judge Seymour’s separate opinion, on the understanding that the analysis holds "even under the heightened standard affirmed by a majority of this court.” Opinion of Seymour, J., at 1011.
. This, too, is a traditional principle of equity practice. See, e.g., Fredericks v. Huber, 180 Pa. 572, 37 A. 90, 91 (1897); Bowling v. Nat’l Convoy & Trucking Co., 101 Fla. 634, 135 So. 541, 544 (1931); Bellows v. Ericson, 233 Minn. 320, 46 N.W.2d 654, 659 n. 9 (1951); State ex rel. McKinley Automotive, Inc. v. Oldham, 283 Or. 511, 584 P.2d 741, 743 n. 3 (1978); Weis v. Renbarger, 670 P.2d 609, 611 (Okla.Ct.App.1983).
. Some states continue to make preservation of the status quo a necessary requirement for all preliminary injunctions. See, e.g., Postma *1014v. Jack Brown Buick, Inc., 157 Ill.2d 391, 193 Ill.Dec. 166, 626 N.E.2d 199, 203 (1993) (stating categorically that "preliminary injunctions are improper where they tend to change the status quo of the parties rather than preserve it”); County of Richland v. Simpkins, 348 S.C. 664, 560 S.E.2d 902, 906 (S.C.Ct.App.2002) (noting that the sole purpose of a preliminary injunction is to preserve the status quo, and affirming the denial of an injunction that would change that status).
. I am puzzled by the dissent’s suggestion that abandoning heightened scrutiny for preliminary injunctions that disturb the status quo would "bring our jurisprudence in closer accord” with "other circuits.” Opinion of Seymour, J., at 1000-1001 citing In re Microsoft Corp. Antitrust Litig., 333 F.3d 517, 526 (4th Cir.2003); Tom Doherty Assocs. v. Saban Entm’t, 60 F.3d 27, 34-35 (2d Cir.1995); Acierno v. New Castle County, 40 F.3d 645, 653 (3d Cir.1994); Wetzel v. Edwards, 635 F.2d 283, 286 (4th Cir.1980); Anderson v. United States, 612 F.2d 1112, 1114-15 (9th Cir.1979). Certainly that is not true of the Second Circuit. In the very opinion cited by the dissent, Tom Doherty, the Second Circuit states:
[W]e have required the movant to meet a higher standard where: (i) an injunction will alter, rather than maintain, the status quo, or (ii) an injunction will provide the movant with substantially all the relief sought and that relief cannot be undone even if the defendant prevails at a trial on the merits.
60 F.3d at 33-34 (emphasis added). The other cited circuits blend the disfavored categories of mandatory injunctions and those that disturb the status quo, but continue to treat the latter as requiring a heightened showing. For example, the Third Circuit decision cited by the dissenters holds as follows:
A primary purpose of a preliminary injunction is maintenance of the status quo until a decision on the merits of a case is rendered. A mandatory preliminary injunction compelling issuance of a building permit fundamentally alters the status quo.... A party seeking a mandatory preliminary injunction that will alter the status quo bears a particularly heavy burden in demonstrating its necessity."
Acierno v. New Castle County, 40 F.3d at 647, 653 (emphasis added; citation omitted). The other cited cases are to similar effect. See Anderson v. United States, 612 F.2d at 1114-15 ("Mandatory preliminary relief, which goes well beyond simply maintaining the status quo pendente lite, is particularly disfavored, and should not be issued unless the facts and law clearly favor the moving party.”); In re Microsoft Corporation Antitrust Litigation, 333 F.3d at 526 ("Mandatory preliminary injunctions [generally] do not preserve the status quo and normally should be granted only in those circumstances when the exigencies of the situation demand such relief.”), quoting Wetzel v. Edwards, 635 F.2d at 286.
. The text and legislative history of the CSA suggest that Congress meant to give the Attorney General authority to make other religious exemptions. See generally Native American Church v. United States, 468 F.Supp. 1247, 1249-51 (S.D.N.Y.1979) (recounting the legislative history of the exemption for the Native American Church). As Judge Murphy notes, Opinion of Murphy, J., at 986, this precise question was presented in Olsen v. DEA, 878 F.2d 1458 (D.C.Cir.1989). In that case, now-Justice Ginsburg refused to accept the DEA’s position that it had the authority to exempt the Native American Church but no other churches, noting that the DEA's interpretation preferred one church above others in a way that would raise serious questions concerning the statute’s constitutionality. See id. at 1461.
. Peyote use by Native American Church groups within the United States is permitted by an express reservation to the Convention.
. In the free exercise/RFRA context, it is important to note that evidence of a compelling government interest rebuts the plaintiffs pri-ma facie case not by disputing the plaintiff's interest in the religious practice but by outweighing it. Not all burden-shifting regimes share this feature. For instance, in the Title VII context, once a plaintiff is able to show disparate treatment of a similarly situated employee of another race, the burden shifts to the employer to show a nondiscriminatory motive for the differing treatment. To the extent that an employer makes such a showing, it does not present considerations that outweigh the plaintiffs interest in a nondiscriminatory workplace; rather, it undercuts the plaintiff's claim of discrimination. Thus, if an employer’s case for a nondiscriminatoiy motive is in equipoise, then it follows that the plaintiff's case for discrimination is also in equipoise. In that context, the dissent’s view of the consequences of equipoise as to the government’s showing is well-founded; in the RFRA context, it seems mistaken.
. The dissent argues that the right at issue in this case is statutory, rather than constitutional, making several of the cases cited above inapposite. Opinion of Murphy, J., at 995 n. 17. But RFRA dictates that the government must meet the same exacting standard as when it seeks to justify a burden on a constitutional right.
. The dissent complains that this passage "does not relate in any fashion” to the balance of the harms or public interest factors. Opinion of Murphy, J., at 997. This is not correct. The Court referred to "important practical reasons to let the injunction stand pending a full trial on the merits.” The first of these was that "the potential harms from reversing the injunction outweigh those of leaving it in place by mistake.” Ashcroft, - U.S. at -, 124 S.Ct. at 2794. But the principal point is what the Court did not do— it did not, as Judge Murphy says we should— treat the plaintiffs as not having met their burden of proof on the balances of equities where the same evidence had been held sufficient to establish that they were likely to prevail on the merits under a compelling interest test.