BRESS, Circuit Judge

Timothy M. Muscat (argued), Deputy Attorney General; William P. Torngren, Supervising Deputy Attorney General; Sara J. Drake, Senior Assistant Attorney General; Rob Bonta, Attorney General; Office of the Attorney General, Sacramento, California; for Defendants-Appellants.

Lester J. Marston (argued), Rapport and Marston, Ukiah, California; David B. Dehnert, Dehnert Law PC, Marina Del Rey, California; for Plaintiffs-Appellees.

George Forman, Jay B. Shapiro, and Margaret C. Rosenfeld, Forman & Associates, San Rafael, California, for Amici Curiae Bear River Band of Rohnerville Rancheria, Cahuilla Band of Indians, Cachil Dehe Band of Wintun Indians of the Colusa Indian Community, and Soboba Band of Luiseño Indians.

Kristin L. Martin, McCracken Stemerman & Holsberry LLP, San Francisco, California, for Amicus Curiae Unite Here International Union.

Laura E. Hirahara, Associate Counsel, California State Association of Counties, Sacramento, California, for Amicus Curiae California State Association of Counties.

Before: Kim McLane Wardlaw, Daniel A. Bress, and Patrick J. Bumatay, Circuit Judges.

Opinion by Judge Bress ;

Concurrence by Judge Wardlaw ;

Dissent by Judge Bumatay

BRESS, Circuit Judge: Under the Indian Gaming Regulatory Act (IGRA), 25 U.S.C. § 2701 et seq. , Indian tribes must enter a compact with the state in order to conduct high-stakes Las Vegas-style casino gambling, known as Class III gaming. But to prevent states from using their compact approval authority to force regulations on tribes that the states would otherwise be powerless to enact, Congress in IGRA imposed important safeguards on compact negotiations. IGRA strictly limits the topics that states may include in tribal-state Class III compacts to those directly related to the operation of gaming activities. 25 U.S.C. § 2710(d)(3)(C). States are also required to negotiate compact agreements in good faith. Id. § 2710(d)(3)(A). If a state does not negotiate in good faith, the tribe may sue in federal court and obtain remedies designed to force the state to the bargaining table and get the deal done. Id. § 2710(d)(7)(B).

We hold in this case that California failed to act in good faith in its compact negotiations with the plaintiff Tribes. The central problem with California's approach was this: it for years demanded that the Tribes agree to compact provisions relating to family law, environmental regulation, and tort law that were unrelated to the operation of gaming activities and far outside the bounds of permissible negotiation under IGRA. Through its negotiating demands, California effectively sought to use the Class III contracting process as leverage to impose its general policy objectives on the Tribes, which a state may not do. California thereby failed to act in good faith, triggering IGRA's remedial provisions.

We affirm the judgment of the court below, although, importantly, on grounds different than the district court articulated.

I

The plaintiffs are the Chicken Ranch Rancheria of Me-Wuk Indians, Blue Lake Rancheria, Chemehuevi Indian Tribe, Hopland Band of Pomo Indians, and Robinson Rancheria. The history of the Tribes' Class III negotiations with the State of California is extensive, and we recite only those events pertinent to this appeal. Some of this history is wrapped up in the history of IGRA itself, but we will limit ourselves here to the facts giving rise to this case and turn to IGRA next.

After Congress passed IGRA in 1988, California and approximately 60 tribes, including the plaintiff Tribes, entered a 1999 compact that gave Indian tribes in California the exclusive right to host Class III gaming. See In re Indian Gaming Related Cases (Coyote Valley II) , 331 F.3d 1094, 1104 (9th Cir. 2003). In return, the tribes agreed to accept various regulations and duties relating to their gaming activities. Id. at 1104–05. The 1999 compacts were set to expire on December 31, 2020, but provided for an automatic extension through June 30, 2022 for those tribes that were in negotiations to extend or replace their existing compacts, which includes the plaintiff Tribes. Very recently, California and the plaintiff Tribes agreed to extend the 1999 compacts until December 31, 2023.

Negotiations over successor compacts to the 1999 compacts have been ongoing for years. In 2014, the plaintiff Tribes joined various other Indian tribes with existing 1999 compacts to form the Compact Tribes Steering Committee (CTSC). The first formal negotiation session was held in January 2015. Between 2015 and 2019, California and the CTSC held 39 days of in-person negotiation sessions, in addition to numerous smaller sessions focused on discrete issues. Over that time, the State provided at least twelve full draft compacts to the CTSC, and the CTSC offered approximately fourteen drafts of its own.

Although the parties reached consensus on some issues, other aspects of the negotiations were fraught. For example, California sought a provision that would require the Tribes to recognize and enforce state spousal and child support judgments against tribal gaming facility employees. California also requested that the Tribes agree to extensive environmental regulations—devoting nearly 30 pages of detailed draft compact provisions to this topic alone. California also wanted the Tribes to adopt California tort law as tribal law that would apply in various situations disconnected from gaming activities, while insisting the Tribes waive sovereign immunity for tort claims and establish tort claims commissions. The Tribes maintained that these requests were insufficiently related to gaming, and that the State therefore could not negotiate for them under IGRA.

Despite these objections, the CTSC operated on a parallel path and endeavored to negotiate the disputed topics "in anticipation of the State offering meaningful concessions" of significant value. But the Tribes came to believe that California was not offering sufficient additional consideration. And California refused to accept any compact that did not include the challenged topics of negotiation. By the end of 2019, and after nearly five years of formal negotiations with the State, the plaintiff Tribes had seen enough. They withdrew from the CTSC and turned down California's existing offers. The tribes also tried one last time, proposing a "best and final offer." But California did not accept it.

In January 2019, the Tribes sued the State, alleging that California violated IGRA's duty to negotiate in good faith. On cross-motions for summary judgment, the district court agreed with the Tribes. The court concluded that California's demand for tribal enforcement of state domestic support orders "pulled negotiations into a field wholly collateral to the operation of gaming facilities," and thus constituted "per se evidence of bad faith." The court went on to explain that many of the other disputed provisions were "not at the heart of," or only "at the very edge of relevance" to, gaming activities. But the district court believed these other provisions were still "somewhat connected" to gaming and thus not a per se violation of the State's good-faith duty.

Nevertheless, because many of the disputed topics still had tenuous connections to gaming, the district court interpreted our precedents to require that the State provide "meaningful concessions" in exchange for demanding these provisions. It then found that California had failed to offer such concessions, and that California had thus not negotiated with the Tribes in good faith. The district court granted summary judgment for the Tribes and ordered that IGRA's remedial process take hold. See 25 U.S.C. § 2710(d)(7)(B)(iii)–(vii).

California now appeals the district court's decision, which we review de novo. Avery v. First Resolution Mgmt. Corp. , 568 F.3d 1018, 1021 (9th Cir. 2009). We may affirm on any ground supported by the record. Miranda v. City of Casa Grande , 15 F.4th 1219, 1224 (9th Cir. 2021).

II

To understand where California went astray in the compact negotiations, we begin by recognizing the unique and limited powers that IGRA gives states over Indian tribes. "[T]he Constitution grants Congress broad general powers to legislate in respect to Indian tribes, power that [the Supreme Court] ha[s] consistently described as ‘plenary and exclusive.’ " United States v. Lara , 541 U.S. 193, 200, 124 S.Ct. 1628, 158 L.Ed.2d 420 (2004). The corollary to this is that states generally lack the power to regulate tribes: "tribal sovereignty is dependent on, and subordinate to, only the Federal Government, not the States." Washington v. Confederated Tribes of Colville Indian Rsrv. , 447 U.S. 134, 154 (1980). Thus, "State laws generally are not applicable to tribal Indians on an Indian reservation except where Congress has expressly provided that State laws shall apply." McClanahan v. State Tax Comm'n of Ariz. , 411 U.S. 164, 170–71, 93 S.Ct. 1257, 36 L.Ed.2d 129 (1973) (quotations omitted); see Ysleta Del Sur Pueblo v. Texas , ––– U.S. ––––, 142 S.Ct. 1929, 1934, ––– L.Ed.2d –––– (2022) ("From time to time, Congress has exercised its authority to allow state law to apply on tribal lands where it otherwise would not.").

In California v. Cabazon Band of Mission Indians , 480 U.S. 202, 107 S.Ct. 1083, 94 L.Ed.2d 244 (1987), the Supreme Court held that California lacked the federal statutory authority required to regulate bingo halls on tribal lands. The Court started from the well-accepted proposition that "state laws may be applied to tribal Indians on their reservations if Congress has expressly so provided." Id. at 207, 107 S.Ct. 1083. But it found no federal statutory authority for California's attempt to regulate tribal bingo enterprises. Id. at 212–14, 107 S.Ct. 1083.

Congress passed IGRA in response to Cabazon . See Ysleta Del Sur Pueblo , 142 S.Ct. at 1937 ; Coyote Valley II , 331 F.3d at 1095–97. In IGRA, "Congress attempted to strike a delicate balance between the sovereignty of states and federally recognized Native American tribes." Pauma Band of Luiseno Mission Indians v. California (Pauma I) , 813 F.3d 1155, 1160 (9th Cir. 2015). IGRA gave Indian tribes "the exclusive right to regulate gaming activity on Indian lands if the gaming activity is not specifically prohibited by Federal law and is conducted within a State which does not, as a matter of criminal law and public policy, prohibit such gaming." 25 U.S.C. § 2701(5). IGRA thus created "a statutory basis for the operation of gaming by Indian tribes as a means of promoting tribal economic development, self-sufficiency, and strong tribal governments." Id. § 2702(1). IGRA also created a statutory basis for regulating these gaming activities. Id. § 2702(2). The stated objectives of this regulation, however, were generally focused on the integrity of the gaming enterprise itself: "to shield it from organized crime and other corrupting influences, to ensure that the Indian tribe is the primary beneficiary of the gaming operation, and to assure that gaming is conducted fairly and honestly by both the operator and players." Id.

IGRA divides gaming activity into three classes, with each class subject to different degrees of federal and state regulation. See 25 U.S.C. § 2710(d)(1)(C). The most significant aspect of IGRA concerns Class III gaming, "the types of high-stakes games usually associated with Nevada-style gambling." Coyote Valley II , 331 F.3d at 1097. Examples of Class III gaming include blackjack, baccarat, slot machines, and parimutuel horse-wagering. Artichoke Joe's Cal. Grand Casino v. Norton , 353 F.3d 712, 715 (9th Cir. 2003) ; see 25 U.S.C. § 2703(7) – (8).

We have described IGRA as "an example of cooperative federalism in that it seeks to balance the competing sovereign interests of the federal government, state governments, and Indian tribes, by giving each a role in the regulatory scheme." Pauma I , 813 F.3d at 1160 (quotations omitted). That is principally because Class III gaming is permitted on Indian lands only if, inter alia , a tribe and the state enter a tribal-state compact that the Secretary of the Interior then approves. Coyote Valley II , 331 F.3d at 1097 ; see also 25 U.S.C. § 2710(d)(1), (3)(B). A tribal-state compact "prescribes rules for operating gaming, allocates law enforcement authority between the tribe and State, and provides remedies for breach of the agreement's terms." Michigan v. Bay Mills Indian Cmty. , 572 U.S. 782, 785, 134 S.Ct. 2024, 188 L.Ed.2d 1071 (2014).

Although IGRA is an example of cooperative federalism, Congress was clear-eyed that state involvement could turn decidedly uncooperative. Class III gaming is not only "a source of substantial revenue" for tribes, but the lifeblood on "which many tribes ha[ve] come to rely." Coyote Valley II , 331 F.3d at 1097, 1099–1100. The risks inherent in the state compact approval requirement are therefore obvious: Indian tribes, who rely on gaming for economic revenue, are at the potential mercy of the states, which could withhold approval of Class III gaming rights or insist upon onerous compact conditions that would give states greater power to regulate tribes.

Congress was well aware of the danger that states could use their compacting approval powers to encroach on tribal sovereignty. Thus, "Congress enacted IGRA to provide a legal framework within which tribes could engage in gaming—an enterprise that holds out the hope of providing tribes with the economic prosperity that has so long eluded their grasp—while setting boundaries to restrain aggression by powerful states." Rincon Band of Luiseno Mission Indians v. Schwarzenegger , 602 F.3d 1019, 1027 (9th Cir. 2010).

IGRA imposes those boundaries in two critical ways. First, "[b]ecause the compact requirement skews the balance of power over gaming rights in favor of states by making tribes dependent on state cooperation," id. , states have an obligation to engage in compact negotiations in good faith. 25 U.S.C. § 2710(d)(3)(A). That obligation has teeth because a tribe may sue in federal court for a state's violation of its good-faith duty. Id. § 2710(d)(7)(A)(i).

In such an action, the plaintiff tribe must first, "upon the introduction of evidence," demonstrate that the state has not negotiated in good faith. Id. § 2710(d)(7)(B)(ii). Following that initial showing, "the burden of proof shall be upon the State to prove that the State has negotiated with the Indian tribe in good faith." Id. ; see also Pauma Band of Luiseno Mission Indians v. California (Pauma II) , 973 F.3d 953, 958 (9th Cir. 2020) (describing the burden-shifting framework).

If a court finds that a state has failed to act in good faith, this triggers IGRA's remedial provisions. In that event, the district court "shall order the State and Indian Tribe to conclude such a compact within a 60-day period." 25 U.S.C. § 2710(d)(7)(B)(iii). If that process fails to yield an agreement, the court is required to appoint a mediator to select a proposed compact that "best comports" with IGRA. Id. § 2710(d)(7)(B)(iv). The state then has 60 days to accept the mediator's proposal. See id. § 2710(d)(7)(B)(vi). If the state still refuses, the Secretary of the Interior shall promulgate procedures under which the tribe may conduct Class III gaming, consistent with the mediator's proposed compact and IGRA's terms. See id. § 2710(d)(7)(B)(vii).

Second, IGRA polices state overreach by circumscribing the permissible topics of negotiation, setting forth seven allowed areas in which tribes and states may reach agreement. Specifically, IGRA provides that a tribal-state compact "may include provisions relating to—":

(i) the application of the criminal and civil laws and regulations of the Indian tribe or the State that are directly related to, and necessary for, the licensing and regulation of such activity;

(ii) the allocation of criminal and civil jurisdiction between the State and the Indian tribe necessary for the enforcement of such laws and regulations;

(iii) the assessment by the State of such activities in such amounts as are necessary to defray the costs of regulating such activity;

(iv) taxation by the Indian tribe of such activity in amounts comparable to amounts assessed by the State for comparable activities;

(v) remedies for breach of contract;

(vi) standards for the operation of such activity and maintenance of the gaming facility, including licensing; and

(vii) any other subjects that are directly related to the operation of gaming activities.

25 U.S.C. § 2710(d)(3)(C).

In addition, another provision of IGRA emphasizes that states generally lack the authority to tax Indian tribes. Specifically, except for any agreed-upon assessments under § 2710(d)(3)(C)(iii), "nothing in this section shall be interpreted as conferring upon a State or any of its political subdivisions authority to impose any tax, fee, charge, or other assessment upon an Indian tribe or upon any other person or entity authorized by an Indian tribe to engage in a class III activity." Id. § 2710(d)(4).

If a tribe files suit alleging that a state has failed to negotiate in good faith, IGRA provides that the court "may take into account the public interest, public safety, criminality, financial integrity, and adverse economic impacts on existing gaming activities." Id. § 2710(d)(7)(B)(iii)(I). But it "shall consider any demand by the State for direct taxation of the Indian tribe or of any Indian lands as evidence that the State has not negotiated in good faith." Id. § 2710(d)(7)(B)(iii)(II).

III

This statutory background sets the stage for the principal questions in this case, which are (1) did California exceed the permissible topics of negotiation under IGRA, and (2) if so, what is the consequence of this? We hold that through its insistence on family law, environmental law, and tort law provisions, California substantially exceeded IGRA's limitation that any Class III compact provision be directly related to the operation of gaming activities. We further hold that when, as here, a state seeks to negotiate for compact provisions that fall well outside IGRA's permissible topics of negotiation, the state has not acted in good faith.

A

California crossed the line in negotiating far outside IGRA's permitted list of compact negotiation topics. We can begin to see why by examining the statute's list of allowed topics, which, as we will explain, sets forth the only permitted topics of negotiation.

We quoted the list of IGRA's seven permitted topics in full above, but as a reminder, it may be found at 25 U.S.C. § 2710(d)(3)(C). That provision states that a Class III gaming compact "may include provisions relating to" the seven identified topics, which culminate in the catch-all topic of "any other subjects that are directly related to the operation of gaming activities." Id. § 2710(d)(3)(C)(vii).

This list, we hold, is exhaustive. In fact, we have effectively already so held. In Rincon Band , we said that "[t]he language and structure of § 2710(d)(3)(C) suggests it is exhaustive," and then squarely held that "IGRA limits permissible subjects of negotiation in order to ensure that tribal-state compacts cover only those topics that are related to gaming and are consistent with IGRA's stated purposes." 602 F.3d at 1028–29 & n.9 (emphasis added). IGRA, we made clear, "does not permit the State and the tribe to negotiate over any subjects they desire; rather, IGRA anticipates a very specific exchange of rights and obligations." Id. at 1039 ; see also Seminole Tribe , 517 U.S. at 49, 116 S.Ct. 1114 (citing § 2710(d)(3)(C) as setting "the permissible scope of a Tribal-State compact"); Navajo Nation v. Dalley , 896 F.3d 1196, 1205 n.4 (10th Cir. 2018) (explaining that "the negotiated terms of the Compact cannot exceed what is authorized by IGRA" (quotations omitted)).

Although our past precedents did not engage in detail with IGRA's text, the plain language of § 2710(d)(3)(C) confirms that its list of seven topics is exhaustive. That section sets out a list of six permitted specific topics and ends with the catch-all "any other subjects that are directly related to the operation of gaming activities." We could no doubt bring many Latin canons to bear on § 2710(d)(3)(C), but we think it easy enough to say the obvious: that the natural inference from this enumerated list is that it is exclusive. Why else devote such attention to drafting a careful itemized list only to have it impose no limits? Indeed, if the list were not exhaustive there would be little point in including the catch-all provision.

It is true, of course, that § 2710(d)(3)(C) prefaces its list by stating that a tribal-state compact "may include provisions relating to" the seven identified topics. But the word "may" is not necessarily a fully permissive term—it does not always mean one "may" do anything , or that everything following the "may" is merely by way of suggestion. Depending on the context, "may" can be limiting, meaning "may only. " See, e.g., Cooper Indus., Inc. v. Aviall Servs., Inc. , 543 U.S. 157, 166, 125 S.Ct. 577, 160 L.Ed.2d 548 (2004) (explaining that "the natural meaning of ‘may’ in the context of [an] enabling clause is that it authorizes certain ... actions—ones that satisfy the subsequent specified condition—and no others"); Cortez Byrd Chips, Inc. v. Bill Harbert Constr. Co. , 529 U.S. 193, 198, 120 S.Ct. 1331, 146 L.Ed.2d 171 (2000) (explaining that "the mere use of ‘may’ is not necessarily conclusive of congressional intent to provide for a permissive or discretionary authority"); Citizens & S. Nat'l Bank v. Bougas , 434 U.S. 35, 38, 98 S.Ct. 88, 54 L.Ed.2d 218 (1977) (noting it was "settled" that the word "may" in a venue provision meant "may ... only"); Marathon Oil Co. v. United States , 374 F.3d 1123, 1138 (Fed. Cir. 2004) ("[T]he word ‘may’ in some contexts is not permissive but indeed is interpreted as restrictive in nature.").

In the context of § 2710(d)(3)(C)'s list of six specific topics followed by a catch-all seventh, it is more natural to read "may" in its restrictive sense, as "may only." As we explained in Rincon Band , "[a]lthough ‘may’ indicates permissiveness ..., to grant permission is not necessarily to grant carte blanche. What is ‘permitted’ is limited." 602 F.3d at 1028 n.9. Thus, a tribal-state compact may include provisions relating to the seven identified topics (though it is not necessarily required to), but it may not include provisions that do not relate to the topics listed.

That interpretation makes a great deal of sense when one steps back and appreciates the critical role of § 2710(d)(3)(C) in IGRA's overall structure and design. Congress "limit[ed] the proper topics for compact negotiations to those that bear a direct relationship to the operation of gaming activities." Coyote Valley II , 331 F.3d at 1111. The reason this was so essential, we have explained, is because "Congress intended to prevent compacts from being used as subterfuge for imposing State jurisdiction on tribes concerning issues unrelated to gaming." Id. ; see also Rincon Band , 602 F.3d at 1028 n.9. If the list of seven topics in § 2710(d)(3)(C) were not exhaustive, states could try to impose on the tribes a potentially infinite range of provisions reflecting general state policy objectives, even ones that strike deep at the heart of tribal sovereignty. That would be directly contrary to the background principle that—for the operation of gaming activities only—IGRA creates a limited exception to states' general lack of power to regulate Indian tribes on Indian lands. See Confederated Tribes of Colville , 447 U.S. at 154, 100 S.Ct. 2069 ; McClanahan , 411 U.S. at 170–71, 93 S.Ct. 1257 ; Coyote Valley II , 331 F.3d at 1095–96.

B

With the exhaustive nature of § 2710(d)(3)(C) established, we can begin to focus more intently on the three most heavily disputed parts of the parties' compact negotiations: those relating to family, environmental, and tort law. These items do not fit with any of the first six listed topics in § 2710(d)(3)(C), and California does not attempt to argue otherwise. So if these topics are permissible subjects of compact negotiation, they must fit within the residual provision of "any other subjects that are directly related to the operation of gaming activities." 25 U.S.C. § 2710(d)(3)(C)(vii). Some parsing of this phrase is therefore required.

Contrary to California's apparent suggestion, the phrase "directly related to the operation of gaming activities" imposes meaningful limits on compact negotiations. The word "directly" is significant. "Directly" connotes a more linear connection between the subject that is to be negotiated and the "operation of gaming activities." "Directly" means "[i]n a straightforward manner" or "[i]n a straight line or course." Black's Law Dictionary (11th ed. 2019). In some sense, everything is "related" to everything else; the word "directly" ensures that we cannot give § 2710(d)(3)(C)'s catch-all that sort of expansive interpretation. Thus, topics of negotiation that have attenuated relationships to the operation of gaming activities, or merely tangential, incidental, or collateral relationships, are not permitted. Chemehuevi Indian Tribe v. Newsom , 919 F.3d 1148, 1153 (9th Cir. 2019) ("[T]he proper inquiry is whether a compact's [topic] is so attenuated from gameplay that it falls outside of paragraph 3(C)(vii).").

The broader structure of § 2710(d)(3)(C) confirms the focused nature of the phrase "directly related to" in § 2710(d)(3)(C)(vii). As a residual clause, § 2710(d)(3)(C)(vii) takes its meaning from, and is limited by, the rest of § 2710(d)(3)(C). See, e.g., Yates v. United States , 574 U.S. 528, 545, 135 S.Ct. 1074, 191 L.Ed.2d 64 (2015) ; Washington State Dept. of Soc. and Health Servs. v. Guardianship Estate of Keffeler , 537 U.S. 371, 384, 123 S.Ct. 1017, 154 L.Ed.2d 972 (2003). It is of course true that as a residual clause, § 2710(d)(3)(C)(vii) is inevitably "broader than the more specific topics enumerated in [the] paragraphs" that precede it. Chemehuevi Indian Tribe , 919 F.3d at 1152. But its scope can only be understood in the context of § 2710(d)(3)(C) as a whole.

The residual clause allows states and tribes to agree on any "other subjects that are directly related to the operation of gaming activities," confirming that the preceding six topics are themselves "directly related to the operation of gaming activities." And indeed they are: they pertain to the licensing and regulation of gaming activities and the enforcement of the same, id. § 2710(d)(3)(C)(i)–(ii); state assessment and tribe taxation of gaming activities, id. § 2710(d)(3)(C)(iii)–(iv); remedies for breach of contract, id. § 2710(d)(3)(C)(v); and standards for the operation of gaming activities and facilities, id. § 2710(d)(3)(C)(vi). Section 2710(d)(3)'s overall focus on the actual "operation of gaming activities" is apparent. In interpreting the residual clause, we must thus take heed of the provisions that precede it, which contribute to its substantive content.

We therefore agree with the Department of Interior that:

In the context of applying the "catch-all" category, we do not simply ask "but for the existence of the Tribe's class III gaming operation, would the particular subject regulated under a compact provision exist?" If this question were used to provide the standard for determining whether a particular object of regulation was "directly related to the operation of gaming activities," it would permit states to use tribal-state compacts as a means to regulate tribal activities far beyond that which Congress intended when it originally enacted IGRA.

We do not have occasion to decide whether this agency interpretation requires any deference; we simply find it persuasive in its alignment with our own independent conclusion. What is required, as the Department of Interior has correctly recognized, is a "direct connection" to the operation of gaming activities.

That is consistent with our analysis in Rincon Band concerning a proposed general revenue sharing provision. There, we rejected as "circular" California's argument that general revenue sharing was "directly related to the operation of gaming activities" simply because "the money [wa]s paid out of income from gaming activities." 602 F.3d at 1032. As we explained, "[w]hether revenue sharing is an authorized negotiation topic under § 2710(d)(3)(C)(vii) ... depends on the use to which the revenue will be put, not on the mere fact that the revenue derives from gaming activities." Id. at 1033. That was, in essence, a rejection of a pure "but for" test. The phrase "directly related" is also not the only limiting feature of § 2710(d)(3)(C)(vii). The referent phrase—"the operation of gaming activities"—is further limiting. It cannot mean "anything that takes place on tribal lands." Nor can it mean "anything relating to a casino property." The phrase " ‘[d]irectly related to the operation of gaming activity’ is narrower than ‘directly related to the operation of the Casino.’ " Flandreau Santee Sioux Tribe v. Noem , 938 F.3d 928, 935 (8th Cir. 2019). We can see this in, among other sources, IGRA's own declaration of policy, which states that a purpose of IGRA is to

provide a statutory basis for the regulation of gaming by an Indian tribe adequate to shield it from organized crime and other corrupting influences, to ensure that the Indian tribe is the primary beneficiary of the gaming operation, and to assure that gaming is conducted fairly and honestly by both the operator and players.

25 U.S.C. § 2702(2). These objectives "for the regulation of gaming" are tied to the operations of the gaming activities themselves, not to anything that may happen on tribal lands simply because the tribe has endeavored to build a casino there.

C

We now turn to the disputed compact provisions and analyze whether they fall within the catch-all. This inquiry is an objective one that is not based on the state's subjective belief that it was acting reasonably. Rincon Band , 602 F.3d at 1041. We have little difficulty concluding that the disputed topics well exceed IGRA's bounds. While there may be close cases in which states slightly overstep the "directly related" to the operation of gaming activities line, this is not one of them.

Family Law Provisions. California demanded that the tribes enact ordinances granting their tribal courts jurisdiction over state spousal and child support orders for gaming facility employees. Under this proposal, state tribal courts or tribal hearing officers would be required to recognize and enforce these family law orders. When the gaming facility was presented with such an order, it would then be required to withhold required amounts from an employee's paycheck and remit them to the party in whose favor the judgment was entered.

Environmental Law Provisions. California insisted upon nearly 30 pages of highly detailed environmental law provisions. These provisions are sprawling, but some highlights will show the breadth of California's ask:

• Tribes would be required to adopt an ordinance incorporating as tribal law significant aspects of California's Environmental Quality Act (CEQA) and the National Environmental Policy Act (NEPA).

• Tribes could not commence any construction on any "Project" until the required environmental processes and associated disputed resolution procedures were completed. The

term "Project" was defined expansively to include "construction of a new Gaming Facility," the "renovation, expansion or modification of an existing Gaming Facility," or any "other activity involving a physical change to the reservation environment, provided the principal purpose of which is directly related to the activities of the Gaming Operation, and any one of which may cause a Significant Effect on the Off-Reservation Environment." "Gaming Facility" was itself defined broadly to "include parking lots, walkways, rooms, buildings, and areas that provide amenities to Gaming Activity patrons, if and only if, the principal purpose of which is to service the activities of the Gaming Operation."

• Absent an exemption—exemptions are themselves a whole further set of provisions—the tribes would be required for any qualifying "Project" to prepare "Tribal Environmental Impact Documents" or "Tribal Environmental Impact Reports," whose requirements varied and depended upon the size of the gaming operation. The Tribal Environmental Impact Documents, for example, would address the impacts of a Project on "(i) air quality; (ii) water resources; (iii) traffic; (iv) public services; (v) hazardous materials; and (vi) noise."

• Tribes would consent to elaborate reporting requirements, as well as dispute resolution mechanisms for any disputes that arose between the tribes and State and local governments.

• Tribes would commit to entering intergovernmental agreements with local governments before commencing any Project. These agreements would include provisions for environmental mitigation and mitigation of traffic impacts, among other things. Tribes were further required to agree to binding arbitration with local governments concerning intergovernmental agreements.

Tort Law Provisions. California insisted that the tribes broadly adopt California tort law as part of tribal law. Tribes would be required to follow California tort law for

all claims of bodily injury, personal injury, or property damage directly arising out of, connected with, or relating to the operation of the Gaming Operation, Gaming Facility, or the Gaming Activities, including but not limited to injuries resulting from entry onto the Tribe's land for purposes of patronizing the Gaming Facility or providing goods or services to the Gaming Facility, provided that such injury occurs at the Gaming Facility or on a road accessing the Facility exclusively.

Tribes would also be required to waive their sovereign immunity for tort claims in tribal court and, if the tribe lacks a tribal court system, to create a tribal claims commission to resolve covered claims. California further demanded that tribes agree to employ in tribal courts and claims commissions discovery procedures analogous to those found in the California Code of Civil Procedure. Tribes could require exhaustion of a tribe's administrative remedies, but the tribes were required to agree to procedures associated with those remedies.

We hold that through these various provisions, California overstepped its proper role under IGRA. Whether considered separately or as a collective whole, these family, environmental, and tort law provisions are not "directly related to the operation of gaming activities." 25 U.S.C. § 2710(d)(3)(C)(vii). Indeed, these disputed provisions have minimal connection to the operation of gaming activities, much less the required "direct " relationship. While the disputed provisions may all reflect worthy policy objectives, that does not allow California to insist upon them in the course of negotiating a Class III gaming compact.

It takes very little to see why California has far exceeded IGRA. Child and spousal support orders have no direct relationship whatsoever to the operation of gaming activities. The environmental provisions would impose extensive environmental review and reporting obligations on tribes for a broadly defined set of "Projects" that includes any physical change "the purpose of which is directly related to the activities of the Gaming Operation." That standard is not found in IGRA, and by California's requested language it would include parking lots and walkways, among various other locations that are at best adjacent to gaming areas. And then there are the intergovernmental agreements that give state and local government agencies an apparent veto (or at least significant control) over tribal projects. All of this is far afield of the actual operation of gaming activities and the mitigation of organized crime and unfair gaming practices that were at the heart of IGRA's limited extension of regulatory authority to the states. 25 U.S.C. §§ 2702(2), 2710(d)(3)(C)(vii).

The disputed tort provisions encounter analogous problems. They would similarly require tribes to commit to adopting and applying an entire body of state law as tribal law, waive sovereign immunity, and create claims commissions for injuries that are merely "connected with" or "relating to" a casino gaming facility, including injuries sustained while entering the facility. It is not hard to imagine the degree to which these provisions would sweep in claims that have no direct connection to the actual operation of gaming activities.

And all the disputed provisions, we hasten to add, strike at core aspects of tribal sovereignty concerning the tribes' governance over their land and people and their decisions about how to structure entire areas of tribal law. This cannot be what Congress had in mind when it enacted statutory text that gave states modest authority to regulate tribal gaming operations through "a very specific exchange of rights and obligations." Rincon Band , 602 F.3d at 1039.

Much of California's defense of the disputed provisions boils down to the same argument: without Indian gaming activities, there would be no wages of employees that could be garnished for spousal and child support orders, no construction projects that would need to be built to support gaming, and no relevant personal injuries that would have occurred on tribal lands. But this is just a reprise of the same "circular" argument we rejected in Rincon . 602 F.3d at 1033. The catch-all language in § 2710(d)(3)(C)(vii) requires an affirmative showing that the state is seeking to negotiate over a subject that has a direct relationship to the operation of gaming activities. That showing is not made simply because gaming activities through some chain of causation produced a situation or event that the state now believes it imperative to regulate. The logic of California's argument is essentially limitless, and it would enable states to force tribes to agree to all manner of state regulations, contrary to IGRA's text, structure, and objectives. See Rincon Band , 602 F.3d at 1028–29 & n.9 ; Coyote Valley II , 331 F.3d at 1111.

Finally, we note that our conclusion that California exceeded the limits of permissible negotiation under § 2710(d)(3)(C) is consistent with the Department of Interior's recent guidance. As noted above, the Department of Interior is responsible for approving or rejecting proposed tribal-state compacts. See 25 U.S.C. § 2710(d)(8). The Tribes have provided us with recent letters from the Assistant Secretary of Indian Affairs in the Office of the Secretary of Interior, issued while this case was on appeal, in which Interior rejected proposed California compacts containing provisions essentially identical to those here. In fact, the Assistant Secretary refused to approve those compacts even though there the Tribes had assented to them.

In these letters, Interior rejected the proposed compacts "as a violation of IGRA because [they] contain[ ] terms that are outside of the narrow scope of IGRA approved topics and are not ‘directly related to the operation of [Class III] gaming activities.’ " Indeed, the letters explained, "we have found certain provisions blatantly in violation of IGRA," and many of them "seek to impose state control where it does not belong."

Evaluating a similar compact provision regarding child and spousal support orders, the Department concluded that it "violates IGRA because it falls outside the permissible scope of subjects that may be included in a compact." Analyzing similar environmental and intergovernmental agreement provisions to those here, the Department concluded that they "fall outside the narrow range of topics IGRA permits" and thus "must be disapproved." Indeed, the Department went on, "requiring a Tribe to adopt state law or its equivalent and permitting for the State to review and object to the Tribe's environmental review is effectively one step removed from the direct application of State law on the Tribe's reservation." And reviewing a tort provision similar to the one here, the Department was again "highly concerned with the State requiring [a] Tribe to adopt a tort claim ordinance that could be interpreted to apply to more than just activity directly related to gaming."

We do not cite the Department of Interior's letters as authoritative, and we acknowledge California's submission indicating that it is seeking reconsideration of the decisions rejecting the compacts at issue. But we find the reasoning in the Department's letters persuasive, and it coincides with our own.

IV

A

Having concluded that California substantially exceeded IGRA's permissible topics of negotiation, we turn to the next question: can a state negotiate well outside the enumerated topics while simultaneously acting in good faith? IGRA does not supply a direct answer to this question. But its text, structure, and our precedents confirm that the much better answer is "no." When a state, as here, seeks to negotiate for compact provisions that fall well outside IGRA's permissible topics of negotiation, the state does not act in good faith.

We reach this conclusion by returning to the core textual and structural features of the statute—and the core principles underlying IGRA—that we set forth above. The defining feature of IGRA, as the statutory text demonstrates, is that it sharply limits the permissible topics of negotiation to prevent states from misusing their compact approval powers to unduly infringe on tribal sovereignty. 25 U.S.C. § 2710(d)(3)(C) ; Rincon Band , 602 F.3d at 1027–29 & n.9 ; Coyote Valley II , 331 F.3d at 1111. The exhaustive nature of IGRA's enumerated list of permissible topics of negotiation means nothing if states can ultimately go beyond that list.

Accordingly, when a state has demanded that a tribe negotiate on topics well outside IGRA's scope, it follows that the state has not negotiated in good faith, end of story. "IGRA limits permissible subjects of negotiation in order to ensure that tribal-state compacts cover only those topics that are related to gaming." Rincon Band , 602 F.3d at 1028–29 (emphasis added). For that statement to be true—and it assuredly is based on IGRA's text—a state that far exceeds the permissible topics of negotiation cannot be acting in good faith. Any contrary conclusion would mean that the seven permitted topics of negotiation are not exhaustive after all, contrary to the statutory text, our precedents, and IGRA's core objectives.

B

Our fine dissenting colleague parts ways with us at this step of the analysis. After agreeing that IGRA's seven topics of permissible negotiation are exhaustive and that California exceeded those topics through its family, environmental, and tort law proposals, the dissent concludes that California can still show it was negotiating in good faith. According to the dissent, California can "establish its good faith despite negotiating for off-list topics." Dissent at 1057.

In the dissent's view, "IGRA's good-faith analysis works like this":

(1) If a tribe fails to show that a State negotiated for topics outside of § 2710(d)(3)(C), absent more, it has failed to meet its initial showing of bad faith and the inquiry ends.

(2) But if a tribe shows that a State negotiated on a topic outside the list, it has satisfied its initial burden of proving bad faith; and

(3) The burden then shifts to the State to show its good faith under the § 2710(d)(7)(B)(iii)(I) factors.

Dissent at 1059. The latter is a reference to an IGRA provision that we quoted above, which states that "in determining whether a State has negotiated in good faith, the court" "may take into account the public interest, public safety, criminality, financial integrity and adverse economic impacts on existing gaming activities." 25 U.S.C. § 2710(d)(7)(B)(iii)(I). The dissent believes these good-faith factors apply to the evaluation of off-list topics of negotiation, and it would thus remand for the district court to evaluate the State's actions under these factors. The dissent's position reflects a misunderstanding of the statutory text that is also at odds with the statute's core objectives, as reflected in both the text itself and our precedents. "IGRA limits permissible subjects of negotiation in order to ensure that tribal-state compacts cover only those topics that are related to gaming." Rincon Band , 602 F.3d at 1028–29. The dissent squarely contradicts this core statement of law by maintaining that states may insist on negotiating topics that are not related to gaming—and that states that do this can somehow still meet IGRA's requirement of good-faith negotiation. The dissent's approach would torpedo the statutory scheme, is inconsistent with our cases, and would inject dramatic uncertainty into compact negotiations.

1

We begin with the first step in the dissent's reasoning, which sets up its entire analytical structure: that if a state only negotiates within IGRA's exhaustive list of negotiating topics, there cannot be a lack of good faith. Dissent at 1059–60. The dissent qualifies this with an "absent more," but the "more" it refers to is "procedural bad faith." Dissent at 1059–60, 1063. So the dissent is thus clear that "if a tribe fails to show that a State has put off-list topics on the negotiation table, then the tribe hasn't met its initial burden and the State may prevail" (again, absent evidence of procedural bad faith). Dissent at 1059 & n.2. This means, according to the dissent, that "Congress gives a State free rein to haggle for permissible gaming-related topics," and that "greater scrutiny" is required only "if the State overreaches by negotiating for topics outside of § 2710(d)(3)(C)." Dissent at 1059 (emphasis added).

The dissent errs at the outset because apart from procedural failings, a state can still fail to act in good faith even when it is negotiating within IGRA's list of exhaustive topics. Nothing in the statute says, much less suggests, that a state has "free rein" to insist upon any compact requirement so long as it fits within the list of seven permitted topics. For example, § 2710(d)(3)(C)(iii) allows a state to seek provisions relating to "the assessment by the State of such [gaming] activities in such amounts as are necessary to defray the costs of regulating such activity." If, for example, a state insisted on highly burdensome and unnecessary audit procedures, that might provide a basis for arguing under § 2710(d)(7)(B)(iii)(I)'s good-faith factors that the state is not negotiating in good faith, even though the state's request falls within a substantively permissible topic of negotiation under IGRA.

The dissent's position that negotiating for topics within IGRA's list cannot be evidence of bad faith is also inconsistent with Coyote Valley II. There, we held that a labor provision was within the list of seven permitted topics and thus an allowed subject of negotiation, but we then proceeded to evaluate whether the provision was nonetheless consistent with § 2710(d)(7)(B)(iii)(I)'s good-faith factors. Coyote Valley II , 331 F.3d at 1115–16. That approach accords not only with the statutory text, but with common sense: just because a given compact requirement is within the list of seven permitted topics based on subject matter does not mean the state is acting in good faith by demanding it. The dissent's contrary analysis leaves the § 2710(d)(7)(B)(iii)(I) good-faith factors with no other work to do besides applying to the negotiation of impermissible topics outside the exhaustive list. But because a state does not necessarily act in good faith by sticking to the list of permitted topics, § 2710(d)(7)(B)(iii)(I) applies in that context. This then brings us to the crux of our disagreement with the dissent: should the good-faith factors in § 2710(d)(7)(B)(iii)(I) also apply when, as here, the state has far exceeded the list of IGRA's permissible topics of negotiation? Or is the state's negotiation of off-list topics itself sufficient evidence of a lack of good faith, as we have reasoned above? The dissent chooses the former because, it tells us, the list of seven topics would otherwise be rendered "meaningless." Dissent at 1058–59, 1062–63. But the dissent has it backwards.

This is one of those classic situations in which a particular textual provision, here § 2710(d)(7)(B)(iii)(I), does not tell us when it should apply, but where the rest of the statute and our precedents overwhelmingly demonstrate the better answer. See Dolan v. U.S. Postal Serv. , 546 U.S. 481, 486, 126 S.Ct. 1252, 163 L.Ed.2d 1079 (2006) ("Interpretation of a word or phrase depends upon reading the whole statutory text, considering the purpose and context of the statute, and consulting any precedents or authorities that inform the analysis."). IGRA is not always a model of clarity, but it provides more than enough guidance for us to resolve this appeal in favor of the tribes.

By concluding that the State may still demonstrate its good faith under § 2710(d)(7)(B)(iii)(I) despite having negotiated for off-list topics, the dissent renders the exhaustive list of negotiating topics in § 2710(d)(3)(C) non-exhaustive. That is directly contrary to the statutory text, IGRA's core objectives, and binding circuit precedent, which all make clear that the list of negotiating topics is exhaustive, and indeed must be exhaustive, in order to prevent states from unduly intruding on tribal governance. See 25 U.S.C. § 2710(d)(3)(C) ; Rincon Band , 602 F.3d at 1028–29 & n.10 ; Coyote Valley II , 331 F.3d at 1111. Treating a violation of § 2710(d)(3)(C) as definitive proof that a state did not fulfill its good-faith duty is, therefore, the only way to give proper meaning to § 2710(d)(3)(C).

For its part, the dissent fully agrees that § 2710(d)(3)(C)'s list is "exhaustive," describing it as "creat[ing] a binary of on-list/off-list subjects with only on-list topics permissible for negotiation." Dissent at 1057. But the dissent then refuses to acknowledge the full implications of this reasoning. Under the dissent's framework, negotiating off-list topics is at once strictly "impermissible," Dissent at 1058, but then ultimately permitted if a court finds that the state was nonetheless acting in good faith under § 2710(d)(7)(B)(iii)(I). We should not interpret the statute to be at war with itself.

Of course, under our interpretation § 2710(d)(7)(B)(iii)(I)'s good-faith factors still have plenty of work to do; we certainly have not "erase[d] these factors from the statute," as the dissent grandiosely charges. Dissent at 1061. The § 2710(d)(7)(B)(iii)(I) factors require analysis when a state seeks to negotiate within the seven topics, but when the specific nature of the state's request is such that an inference of bad faith still arises (remember the oppressive audit procedures hypothetical). But extending § 2710(d)(7)(B)(iii)(I) outside that context to off-list negotiation topics, as the dissent proposes, would render the exhaustive list of negotiating topics non-exhaustive, which is unacceptable as a matter of both statutory interpretation and governing precedent. Rincon Band also repeatedly invoked the canon of construction construing ambiguous laws in favor of tribal interests, see 602 F.3d at 1028 n.9, 1031 n.14, 1032, and that canon cuts firmly against the dissent's interpretation as well.

In addition, our interpretation of § 2710(d)(7)(B)(iii)(I) is one this particular provision readily permits. That section directs that the court "may take into account the public interest, public safety, criminality, financial integrity and adverse economic impacts on existing gaming activities." 25 U.S.C. § 2710(d)(7)(B)(iii)(I) (emphasis added). The very next provision says that the court "shall consider any demand by the State for direct taxation of the Indian tribe or of any Indian lands as evidence that the State has not negotiated in good faith." Id. § 2710(d)(7)(B)(iii)(II) (emphasis added).

Although "may" must mean "may only" in the context of the list of permitted negotiation topics in § 2710(d)(3)(C), we think § 2710(d)(7)(B)(iii)(I) is quite different. The structure of § 2710(d)(7)(B)(iii)(I) is dissimilar from that of § 2710(d)(3)(C) and does not consist of an itemized list punctuated by a catch-all. In addition, the juxtaposition of the "may" and "shall" directives in the two adjoining sub-provisions of § 2710(d)(7)(B)(iii) strongly suggests that in § 2710(d)(7)(B)(iii)(I), "may" should really just mean the optional "may." The word "may" thus gives us ample textual room to choose an interpretation that does not undermine the rest of the statute.

Also wrong is the dissent's attempt to claim that Rincon Band supports its interpretation of § 2710(d)(7)(B)(iii)(I). Purporting to quote Rincon Band , the dissent says that " ‘the State may attempt to rebut bad faith’ by showing that any off-list topic was justified by ‘the public interest, public safety, criminality, financial integrity, and adverse economic impacts on existing gaming activities.’ " Dissent at 1057 (quoting Rincon Band , 602 F.3d at 1032 ). The problem is that the dissent through selective quotation has changed the meaning of the quote, and thus the case as a whole.

The actual line from Rincon Band is: "the State may attempt to rebut bad faith by demonstrating that the revenue demanded was to be used for the public interest, public safety, criminality, financial integrity, and adverse economic impacts on existing gaming activities." Rincon Band , 602 F.3d at 1032 (quotations omitted; emphasis added). The dissent errs because revenue demands are not categorically equivalent to "any off-list topic" under IGRA.

To the contrary, a wholly separate statutory provision governs how demands for revenue inform the good faith analysis. See 25 U.S.C. § 2710(d)(7)(B)(iii)(II). That provision, as we noted above, states that courts "shall consider any demand by the State for direct taxation of the Indian tribe or of any Indian lands as evidence that the State has not negotiated in good faith." Id. (emphasis added). Some assessments are within the permitted topics of negotiation, see id. § 2710(d)(3)(C)(iii), but another provision states that "nothing in this section shall be interpreted as conferring upon a State or any of its political subdivisions authority to impose any tax, fee, charge, or other assessment upon an Indian tribe." Id. § 2710(d)(4).

Bound by Coyote Valley II and its treatment of revenue-sharing provisions, Rincon Band explained that "IGRA requires courts to consider a state's demand for taxation as evidence of bad faith, not conclusive proof." 602 F.3d at 1032 (citing Coyote Valley II , 331 F.3d at 1112–13 ). Rincon Band thus analyzed at length whether the State had provided "meaningful concessions" for a revenue-sharing requirement. See id. at 1036–40. As we will explain in detail below, a "meaningful concessions" analysis only applies within the context of § 2710(d)(4), to revenue-sharing demands (a point the dissent does not dispute). But that only confirms the dissent's error in selectively quoting Rincon Band. Rincon Band did not hold that off-list negotiations were subject to a further good-faith analysis under § 2710(d)(7)(B)(iii)(I). It instead held that "IGRA limits permissible subjects of negotiation in order to ensure that tribal-state compacts cover only those topics that are related to gaming." Rincon Band , 602 F.3d at 1028–29 (emphasis added). If the dissent had its way, that would no longer be true.

The dissent's heavy reliance on IGRA's burden-shifting framework, see 25 U.S.C. § 2510(d)(7)(B)(ii), is thus a non sequitur. Dissent at 1058–59. That the burden may shift to the state to demonstrate its good faith says nothing about whether and when the state will be able to meet its burden. And here we find it inconceivable that the Congress that ringfenced the permissible topics of negotiation to prevent states from encroaching on tribal sovereignty, see Rincon Band , 602 F.3d at 1028–29 & n.9 ; Coyote Valley II , 331 F.3d at 1111, would at the same time give states the opportunity to maintain, in both compact negotiations and before courts, that their off-list topics are nonetheless reasonable. See Quarles v. United States , ––– U.S. ––––, 139 S. Ct. 1872, 1879, 204 L.Ed.2d 200 (2019) ("We should not lightly conclude that Congress enacted a self-defeating statute.").

2

The dissent's approach would also destabilize the compact negotiation process, creating significant uncertainty for vital rights of tribal self-governance and dragging out compact negotiations. Section 2710(d)(7)(B)(iii)(I) allows courts to "take into account the public interest, public safety, criminality, financial integrity, and adverse economic impacts on existing gaming activities." Any compact provision that a state could demand may plausibly be said to be in the "public interest" or have a sufficient relationship to one or more of these good-faith factors. Environmental regulation can be said to be in the public interest. The same is true of spousal support payments.

Contrary to the dissent's suggestion, we are not "wonder[ing] why Congress inserted the good-faith factors into IGRA." Dissent at 1061. We do not question that choice at all. But that choice does not answer the question of when those factors apply. And there is no reason to believe that a Congress that set up an exhaustive list of topics of negotiation to preserve tribal sovereignty, 25 U.S.C. § 2710(d)(3)(C), would choose to undermine that crystalline textual objective through the particular choice of these good-faith factors.

In an effort to bring clarity to its approach, the dissent attempts to assure us that the good faith factors in § 2710(d)(7)(B)(iii)(I) must be still interpreted in a way that sufficiently relates to gaming and the purposes of IGRA. Dissent at 1059. But the dissent provides no guidance beyond that. And it would set the district court on the seemingly hopeless mission of determining whether negotiation topics that are not even "directly related to the operation of gaming activities," 25 U.S.C. § 2710(d)(3)(C)(vii), are nonetheless sufficiently tied to the State's interest in regulating gaming and the purposes of IGRA, so as to somehow demonstrate the State's good faith. Dissent at 1059–60.

To make matters more confounding, under the dissent's view the district court would need to undertake this enigmatic good-faith analysis even though many of the State's off-list topics, such as extensive family, environmental, and tort regulation, strike at the core of the Tribes' governance over their land and people. If the dissent were the law, it is entirely unclear what evidence the parties would be required to marshal on remand, much less how the district court is supposed to go about resolving the further good-faith question.

The substantial uncertainty that the dissent's approach would create is itself directly contrary to IGRA and our precedents. We said in Rincon Band that "the function of the good faith requirement and judicial remedy is to permit the tribe to process gaming arrangements on an expedited basis, not to embroil the parties in litigation over their subjective motivations." 602 F.3d at 1041. The dissent would produce exactly what Rincon Band sought to avoid. The result of the dissent's approach would be that every topic would be in play in tribal-state compact negotiations until things could eventually get sorted out in court. The compacting parties would thus not learn the rules of engagement until litigation over their negotiation process was complete. This would give states enormous leverage at the negotiating table—leverage Congress did not want states to have. And it means that it would take years for compact disputes to get resolved.

Of course, the irony in the dissent's approach is that if a state is ultimately able to persuade a court that it acted in good faith and its disputed provisions remain in the compact (or the tribes cave to them), there is no present likelihood that the Secretary of the Interior would even approve such an arrangement. As we discussed above, the Department of Interior has recently refused to approve compacts with provisions like California's, even though there the State and the Tribes had agreed to them. It is hard to see the wisdom (or basis in law) for forcing the parties and the district court into further proceedings on a compact that the Department of Interior has telegraphed it will reject.

The dissent's main rejoinder to everything we have said in this opinion is that we are relying on our own "sense of IGRA's core ‘principles’ and ‘objectives.’ " Dissent at 1060. But the dissent's bromides about the role of judges do little to advance the dialogue here. The disagreement between us is over how to interpret statutory text, within the confines of a methodology that treats statutory language and structure as the only true indicators of legislative intent. The dissent's interpretation narrowly focuses on the good-faith factors in § 2710(d)(7)(B)(iii)(I), with no appreciation for how that interpretation fundamentally destroys the core textual and structural feature of the statute it purports to interpret. Our interpretation considers the language of § 2710(d)(7)(B)(iii)(I) within the context of the statute as a whole. And in that sense, our interpretation—which is based on statutory language and precedent, not unenacted statutory purpose—is more faithful to the text and structure of the statute that Congress enacted. The dissent's assertion that we are "divin[ing] some broad legislative purpose" to "override the statute's plain meaning" is therefore simply unfounded. Dissent at 1060. And the dissent repeatedly invoking the jargon of textualist interpretation does not make its interpretation textually sound.

V

The district court agreed that California had not negotiated in good faith and that IGRA's remedial provisions were triggered. But it reached this same result through a different path. Because it may affect the parties' future negotiations and the IGRA remedial process, we take this opportunity to explain why the district court's approach rested on a mistaken understanding of our precedents.

As we recounted above, the district court concluded that because various of the State's negotiating demands were at the outer edge of relevance to gaming, to demonstrate its good faith the State was required to provide "meaningful concessions" in return. The State argued in the district court that it had offered to provide the Tribes with particularly valuable consideration in exchange for the disputed provisions. But the district court found that these concessions were insufficiently specific to the State's particular compact demands. The district court acknowledged that "[t]he Ninth Circuit has only discussed ‘meaningful concessions’ in the context of fee demands," but believed that framework could be expanded to other topics of negotiation.

The district court erred in relying on the "meaningful concessions" framework. We first introduced the analytical concept of "meaningful concessions" in Coyote Valley II. There we considered the State's insistence on a provision in the 1999 tribal-state compact that would create a Revenue Sharing Trust Fund (RSTF). 331 F.3d at 1110. The RSTF would have required that gaming tribes share their gaming revenue with tribes that did not have gaming operations. Id. The Tribes maintained that the State had not negotiated in good faith, claiming that the requested RSTF provision was a prohibited tax under § 2710(d)(4). See id. That provision, as we noted above, states that "nothing in this section shall be interpreted as conferring upon a State or any of its political subdivisions authority to impose any tax, fee, charge, or other assessment upon an Indian tribe or upon any other person or entity authorized by an Indian tribe to engage in a class III activity." 25 U.S.C. § 2710(d)(4).

We held that California had acted in good faith in requesting the RSTF provision. We first made clear that this was a permissible topic of negotiation because " § 2710(d)(3)(C)(vii) authorizes the RSTF provision." Coyote Valley II , 331 F.3d at 1111 ; see also id. ("It is clear that the RSTF provision falls within the scope of paragraph(3)(C)(vii)."). We then explained that the RSTF provision did not run afoul of § 2710(d)(4)'s prohibition on taxes in light of the State's meaningful concessions: "Given that the State offered meaningful concessions in return for its demands, it did not ‘impose’ the RSTF within the meaning of § 2710(d)(4)." Id. Through these meaningful concessions, we held, "the State ha[d] successfully rebutted any inference of bad faith." Id.

We addressed "meaningful concessions" again in Rincon Band. There, we considered California's request during renegotiations of the 1999 compacts that tribes pay a significant portion of net gaming revenues into the State's general fund. 602 F.3d at 1022. We first held, as recounted above, that the general revenue sharing provision was not "directly related to the operation of gaming activities," rejecting the State's argument as "circular." Id. at 1033 ; see also id. at 1034. We then reasoned that even if general revenue sharing were a permissible topic of negotiation, the State had still failed to act in good faith because it had not provided meaningful concessions. Id. at 1036–40.

Properly considered, a "meaningful concessions" analysis does not apply in this case, for two reasons. First , we have never held that the "meaningful concessions" doctrine applies to requested topics of negotiation that are well outside the seven permitted topics in § 2710(d)(3)(C). That is, a state cannot negotiate well outside the bounds of § 2710(d)(3)(C) and then attempt to excuse this by arguing that it provided extra goodies in the negotiation package. The reason is the same reason that we do not further analyze the good faith factors in § 2710(d)(7)(B)(iii)(I) when it comes to off-list topics: if states were permitted to offer meaningful concessions in exchange for negotiating for topics well beyond IGRA's permitted list, Congress's entire scheme of limiting the parties to topics directly related to gaming operations would be fatally undermined. Allowing a meaningful concessions analysis to supersede § 2710(d)(3)(C) would mean that compacts could be "used as a subterfuge for imposing State jurisdiction on tribes concerning issues unrelated to gaming." Coyote Valley II , 331 F.3d at 1111.

This conclusion coheres with how we approached the "meaningful concessions" issue in Coyote Valley II. There, we were careful to address meaningful concessions only after concluding that the disputed provisions fell within the scope of permissible negotiation subjects under § 2710(d)(3)(C). See id. at 1111–15. In Rincon Band , we thus explained that " Coyote Valley II thus stands for the proposition" that a state may request revenue sharing provisions if backed by meaningful concessions, but only "if the revenue sharing provision is ... for uses ‘directly related to the operation of gaming activities’ in § 2710(d)(3)(C)(vii)." 602 F.3d at 1033. We have never suggested that a state could point to material concessions that it offered in return for bargaining well outside the allowed seven topics. We specifically reject that notion here.

Second , even when a state seeks to negotiate on a topic within § 2710(d)(3)(C), a meaningful concessions requirement still only applies to demands for taxes, fees, or other revenue-sharing provisions. This is due to the specific prohibitory language in IGRA relating to states seeking to "impose" taxes or other assessments on tribes. Section 2710(d)(4) provides that except for assessments agreed to under § 2710(d)(3)(C)(iii), "nothing in this section shall be interpreted as conferring upon a state or any of its political subdivisions authority to impose any tax, fee, charge, or other assessment upon an Indian tribe." 25 U.S.C. § 2710(d)(4) (emphasis added).

The material concessions requirement is based on states' need to avoid "imposing" such improper taxes. As we explained in Rincon Band , "[t]he relevance of ‘meaningful concessions’ arises from § 2710(d)(4)." 602 F.3d at 1036. That is, offering material concessions allows a state to negotiate for certain revenue-sharing provisions without running afoul of § 2710(d)(4). As we explained in Coyote Valley II , when "a State offers meaningful concessions in return for fee demands, it does not exercise ‘authority to impose’ anything." 331 F.3d at 1112 ; see also at 1111 ("Given that the State offered meaningful concessions in return for its demands, it did not ‘impose’ the RSFT within the meaning of § 2710(d)(4).").

Provisions that do not concern taxes, fees, or revenue-sharing, by contrast, are not subject to the § 2710(d)(4) prohibition that the state not "impose" such requirements. Thus, to the extent a state negotiates within the enumerated topics in § 2710(d)(3)(C), for a state to act in good faith there is no further affirmative requirement to demonstrate material concessions, unless the provision sought concerns taxes or fees. Because the State here negotiated well outside § 2710(d)(3)(C) and the disputed family, environmental, and tort law provisions did not involve revenue sharing, the district court erred in considering the State's good faith through a meaningful concessions lens, even as the district court ultimately reached the correct result in this case.

* * *

"[T]he good faith requirement exists" because "Congress anticipated that states might abuse their authority over compact negotiations to force tribes to accept burdens on their sovereignty in order to obtain gaming opportunities." Rincon Band , 602 F.3d at 1042. We hold that by negotiating for topics well outside § 2710(d)(3)(C)'s permitted list, California did not bargain in good faith. We therefore direct the parties to proceed to IGRA's remedial framework under the district court's continued supervision.

AFFIRMED.

WARDLAW, Circuit Judge, concurring:

The Indian Gaming Regulatory Act (IGRA) is ambiguous on the central question this case presents: whether a state conducts Tribal-State Compact negotiations in bad faith when it insists on negotiating topics beyond the exclusive topics set forth in IGRA § 2710(d)(3)(C). Congress did not clearly explain how the exhaustive list of negotiating topics interacts with the good faith burden-shifting provisions that apply once a tribe files an enforcement action; nor did it define "good faith" to include or exclude the State's introduction of unauthorized topics.

In answering the question presented, my fine colleagues confine themselves to the text of the statute and our precedents construing IGRA—and reach equally plausible, but diametrically opposed, conclusions. Judge Bumatay would layer the burden-shifting provision on top of both on-topic and off-topic negotiations, while Judge Bress concludes that if the State injects an off-topic term into the negotiations, that is sufficient for a finding of bad faith, which triggers IGRA's remedial provisions.

I agree with Judge Bress's analysis of the text and structure of IGRA. My agreement with his conclusions, however, is further supported by IGRA's stated purpose and its legislative history. Moreover, given the inherently ambiguous nature of IGRA's statutory text and structure, it is helpful to bear in mind the words of Justice Scalia thirty years ago:

When we are faced with these two possible constructions [of a federal statute implicating Tribal interests], our choice between them must be dictated by a principle deeply rooted in this Court's Indian jurisprudence: statutes are to be construed liberally in favor of the Indians, with ambiguous provisions interpreted to their benefit.

Cnty. of Yakima v. Confederated Tribes , 502 U.S. 251, 269, 112 S.Ct. 683, 116 L.Ed.2d 687 (1992) (cleaned up). These additional considerations lead me to conclude that when a state insists on negotiating topics that are not even arguably related to "the operation of gaming activities," as California did here, the State has not conducted such negotiations in good faith under IGRA. I therefore concur in Judge Bress's majority opinion.

I.

Although the federal government has the power to grant states jurisdiction over tribal affairs, see McClanahan v. State Tax Comm'n of Ariz. , 411 U.S. 164, 170–71, 93 S.Ct. 1257, 36 L.Ed.2d 129 (1973), it has rarely exercised that power in light of the historically fraught relationship between states and tribes.

"The policy of leaving Indians free from state jurisdiction and control is deeply rooted in the Nation's history." Rice v. Olson , 324 U.S. 786, 789, 65 S.Ct. 989, 89 L.Ed. 1367 (1945). For centuries, states have been the "deadliest enemies" of tribes, United States v. Kagama , 118 U.S. 375, 384, 6 S.Ct. 1109, 30 L.Ed. 228 (1886), and the entities "least inclined to respect" tribal sovereignty, McGirt v. Oklahoma , ––– U.S. ––––, 140 S. Ct. 2452, 2462, 207 L.Ed.2d 985 (2020). Accordingly, delegations of jurisdiction over tribal affairs to states have been perceived as an abandonment of the federal government's duty to safeguard tribal sovereignty. As the Supreme Court put it in 1886:

[Tribes] owe no allegiance to the states, and receive from them no protection. Because of the local ill feeling, the people of the states where they are found are often their deadliest enemies. From their very weakness and helplessness, so largely due to the course of dealing of the federal government with them, and the treaties in which it has been promised, there arises the duty of protection, and with it the power. This has always been recognized by the executive, and

by congress, and by this court, whenever the question has arisen.

Kagama , 118 U.S. at 384, 6 S.Ct. 1109.

Congress enacted IGRA following the Supreme Court's ruling that California lacked Pub. L. 280 authority to regulate bingo (or other gaming) on tribal lands in California v. Cabazon Band of Mission Indians , 480 U.S. 202, 107 S.Ct. 1083, 94 L.Ed.2d 244 (1987). The Cabazon holding was itself rooted in "traditional notions of Indian sovereignty and the congressional goal of Indian self-government, including its overriding goal of encouraging tribal self-sufficiency and economic development." Id. at 216, 107 S.Ct. 1083 (cleaned up).

Congress's choice to involve states in regulating gaming was controversial. When Congress began debating IGRA, there was widespread agreement about its twin goals. IGRA's first purpose in legalizing gaming on Indian lands was "as a means of promoting tribal economic development, self-sufficiency, and strong tribal governments." 25 U.S.C. § 2702 ; see also S. Rep. No. 100-446, at *1–2 (1988). Its second purpose was to regulate gaming on Indian lands in a manner that would "shield it from organized crime and other corrupting influences." 25 U.S.C. § 2702. The difficulty Congress faced was how to structure a regulatory scheme that would meet both purposes, i.e., a scheme that would effectively deter corruption, but did not undermine tribal sovereignty.

In an effort to ensure the proposed regulation did not infringe on tribal sovereignty more than was essential, Congress assigned regulatory authority according to three "classes" of gaming. It divided the classes based on how lucrative they were, reasoning that more lucrative games were more likely to be the target of "corrupting influences." Class I games, including traditional tribal games associated with tribal ceremonies and celebrations, which posed "little risk of corruption," were regulated exclusively by the tribes. Class II games, which included Bingo and other games which tribes had regulated on their own for decades with "relatively few problems," were regulated by the tribes with some oversight by the federal government.

The regulation of Class III gaming proved trickier. On the one hand, there was relatively widespread agreement that mere federal oversight would not be sufficient. As one commentator explained:

Congress saw casino-style gambling as carrying greater risks and raising different issues than bingo. The states' interests in preventing the infiltration of organized crime and controlling gambling generally appeared most persuasive in the context of the "cash business" of

casino-style gaming, or "Class III" gaming.

However, it was not obvious what form of regulatory authority should take its place. "Recognizing that the extension of State jurisdiction on Indian lands has traditionally been inimical to Indian interests," one early version of the bill proposed creating a new federal body, housed within either the Department of Justice (DOJ) or the Department of the Interior (DOI), which would be responsible for regulating Class III gaming. However, both Departments vigorously opposed that proposal. DOJ resisted the role, asserting that it wasn't a "regulatory agency," and was therefore ill suited to the role. DOI representatives testified that the Department "did not have the capacity to undertake such a mission and that, furthermore, it would be in conflict with the Department's effort to reduce its past paternalistic role in favor of support for tribal independence and self-determination."

An alternate proposal, drafted by DOJ, suggested that Congress delegate responsibility for regulating Class III gaming to the states. Proponents of this plan, including DOI, pointed out that states were well suited to the role because they already regulated gaming within their borders, and therefore had the regulatory infrastructure in place. Unsurprisingly, tribes as well as a number of members objected to that plan, expressing concern that "state regulation of tribal gaming would violate tribal sovereignty and, on a more practical level, undermine tribal gaming as an economic development strategy," as states might try to protect their own competing gaming operations, like lotteries. For example, Senator McCain testified to his belief that the states, if given jurisdiction over gaming on Indian lands, would not give the "Indians a fair shake." Those concerns were particularly salient based on the states' recent attempts to eradicate tribal gaming entirely, including California's failed effort to regulate tribal gaming to the point of extinction. See Cabazon , 480 U.S. at 205–06, 107 S.Ct. 1083.

The Tribal-State Compact provision finally enacted was a grand compromise. States were given a role in regulating Class III gaming because federal agencies did not want responsibility for it, and states already had relevant expertise. However, that role was strictly limited to ensure that states could not demand inappropriate concessions from tribes or use Compacts "as a subterfuge for imposing State jurisdiction on tribal lands." Therefore, Tribal-State Compacts could only include provisions that related to "the issues" listed in § 2710(d)(3)(C) —those "directly related to gaming." The Senate Report does not specifically discuss how Congress intended the exclusive negotiating topic provision to relate to the good faith negotiation requirement. However, it is clear that the limit on negotiation topics was the barrier erected against state overreach, while the good faith requirement was the "incentive" to ensure that States "deal[t] fairly with tribes as sovereign governments."

Several members concerned with tribal rights only reluctantly agreed to grant states regulatory authority over even Class III gaming. They stated that they did so only because of the strong protections the legislation afforded tribes should states exceed their bounds. For example, Senator Evans issued a statement that he supported the bill "with great reluctance," even accounting for the limited scope of state intervention allowed:

We intend that the two sovereigns—the tribes and the States—will sit down together in negotiations on equal terms and come up with a recommended methodology for regulating class III gaming on Indian lands. Permitting the States even this limited say in matters that are usually in the exclusive domain of tribal government has been permitted only with extreme reluctance.

Senator McCain similarly stated that, "[i]t is with great reluctance that I am supporting [IGRA]," and emphasized that he had only done so due to the limited nature of the state's involvement:

The Committee Report is clear as to the purpose of Tribal/State compacts as called for in Section [2710(d)]. I understand

Senator Evans' concerns regarding the potential overextension of the intended scope of the Tribal/State compact approach. Toward this end, I believe it is important to again underscore the statement that appears on page 10 of the Report: "The Committee does not intend to authorize any wholesale transfer of jurisdiction from a tribe to a state." From time immemorial, Tribes have been and will continue to be permanent governmental bodies exercising those basic powers of government, as do Federal and State governments, to fulfill the needs of their members. Under our constitutional system of government, the right of Tribes to be self-governing and to share in our federal system must not be diminished.

II.

In light of IGRA's purpose to facilitate tribes' operation of gaming activities, and Congress's reluctance to allow states to have any involvement in oversight of even Class III gaming, I agree with Judge Bress that when a state insists upon negotiating topics plainly beyond the scope of § 2710(d)(3)(C)(vii), that is per se bad faith.

This case is a textbook example of how a contrary interpretation would undermine IGRA's purposes. California began negotiations with the Compact Tribe Steering Committee (CTSC) nearly eight years ago, in 2014. From the start of those negotiations, the State demanded that the final Compact include a number of provisions not even tangentially related to gaming. For example, the State demanded that the Tribes adopt state tort law, subject themselves to state environmental regulations, and enforce state child and spousal support orders against their members. California's insistence on those provisions eventually led the Plaintiff tribes to pull out of the CTSC in 2019, reject the State's offer, and initiate this suit, which has now dragged on for more than three years.

If we were to adopt Judge Bumatay's position, hold that the State could justify its insistence on those terms by showing that they were in the "public interest," and remand this case to the district court for further proceedings, Congress's objectives would be frustrated in at least two key ways. First, it would subject the Tribes to even more costly delays, and potentially jeopardize their ability to engage in gaming at all, as due to these failed negotiations, the existing Compact authorizing gaming was on the brink of expiring until the parties extended its end date from June 2022 to December 2023. Second, it would fatally undermine the key barrier Congress erected against states' overreach in regulating tribal gaming. If a state could insist on topics as far afield as spousal and child support so long as they were in the "public interest," the "exhaustive" list of topics wouldn't be exhaustive at all. And in that event, the worst fears of Congressional members—that states would place their interests above the tribes'—would be realized.

III.

This conclusion is bolstered by the application of traditional Indian canons of construction. As my colleagues' competing views demonstrate, the statutory language is ambiguous. When a statute is ambiguous, we apply the Indian canons of construction, one of which the Supreme Court codified in County of Yakima. As noted at the outset, where, as here, we are faced with two possible constructions of a federal statute implicating tribal interests, "our choice between them must be dictated by a principle deeply rooted in this Court's Indian jurisprudence: statutes are to be construed liberally in favor of the Indians, with ambiguous provisions interpreted to their benefit." 502 U.S. at 269, 112 S.Ct. 683 (cleaned up).

Congress in fact contemplated that we would apply this very canon to IGRA when interpreting it. The Senate Committee Report states that Congress "trusts that courts will interpret any ambiguities [in IGRA] in a manner that will be most favorable to tribal interests consistent with the legal standard used by courts for over 150 years in deciding cases involving Indian tribes." S. Rep. No. 100-446, at *15.

We are faced with not only two "possible" constructions of IGRA, but two equally plausible ones. Applying the canon, we must choose the interpretation most favorable to the tribes, which, in this case, is the construction proposed by Judge Bress. Holding that it is a per se breach of "good faith" for a state to insist on negotiating a compact provision not even arguably within the "operation of gaming activities" protects tribal sovereignty, while ensuring that the twin purposes of IGRA are fulfilled.

IV.

Contrary to Judge Bumatay's assertions, this approach does not conflict with § 2710(d)(7)(B)(iii) by "bar[ring] courts from even considering the [good faith] factors before forcing parties into IGRA's remedial procedures" in all cases. Courts would still be able to consider the good faith factors as part of the burden-shifting framework in a number of instances including, for example, when evaluating whether a state's actions constitute procedural bad faith, and whether a state proposed on-topic terms in good faith. This approach does not read the burden-shifting good faith inquiry out of the statute, it simply applies it in a manner consistent with the limited role of states in regulating Indian gaming.

BUMATAY, Circuit Judge, dissenting:

This case poses a difficult question of statutory interpretation. It requires us to parse several provisions of the Indian Gaming Regulatory Act ("IGRA") to determine the meaning of "good faith." But as with any case, we are duty-bound to follow the text of the law wherever it leads. And because the majority creates an automatic bad-faith rule not found in IGRA's text, I respectfully dissent.

IGRA establishes a comprehensive framework governing gaming on Indian lands. See 25 U.S.C. § 2701, et seq. Under IGRA, tribal casinos may only run Class III gaming (Las-Vegas style games such as blackjack) under a valid Tribal-State gaming compact—an agreement that authorizes and regulates gaming activity on tribal land within a State. § 2710(d)(1)(C). ¹ IGRA enumerates permissible gaming-related topics that may be included within these Tribal-State compacts. § 2710(d)(3)(C). IGRA also requires States to negotiate in "good faith" when a tribe requests a compact for Class III gaming. § 2710(d)(3)(A). While "good faith" is not specifically defined, IGRA provides multiple factors that courts may consider in determining whether a State has met its duty. Finally, to enforce this good-faith duty, IGRA creates a cause of action for tribes to bring against recalcitrant States, § 2710(d)(7)(A)(i), and a mandatory remedial process that is triggered when a State violates its duty. § 2710(d)(7)(B)(iii)–(vii).

The Chicken Ranch Rancheria of Me-Wuk Indians, along with several other tribes (collectively, the "Tribes") sued the State of California and Governor Gavin Newsom (collectively, "California") for violating their "good faith" duty by negotiating for topics outside IGRA's scope. The district court agreed with the Tribes and found that California violated its good-faith duty of negotiation. Chicken Ranch Rancheria of Me-Wuk Indians v. Newsom , 530 F. Supp. 3d 970, 988 (E.D. Cal. 2021). And while I agree with the majority that California negotiated for topics outside IGRA's scope, I disagree with the majority's purpose-based analysis of "good faith." Instead, by looking to IGRA's text and structure, I would vacate and remand the district court's judgment for a proper analysis of whether California satisfied its good-faith duty.

I.

IGRA's text and structure command a nuanced approach to the good-faith analysis. Under IGRA, a State must negotiate to enter a Tribal-State compact in "good faith." § 2710(d)(3)(A). And IGRA creates a burden-shifting framework for analyzing whether a State has negotiated in good faith. Pauma Band of Luiseno Mission Indians v. California , 973 F.3d 953, 958 (9th Cir. 2020). A plaintiff tribe must first "introduc[e] ... evidence" that a compact was not entered and that the State either failed to respond or "did not respond to [the tribe's] request in good faith." § 2710(d)(7)(B)(ii)(I)–(II). Once that initial showing is made, the State bears the "burden of proof" in establishing that it negotiated with the tribe in good faith. § 2710(d)(7)(B)(ii). If the State fails to make such a showing, a court may order the parties into IGRA's "multi-step judicial remedy." Pauma Band , 973 F.3d at 958.

IGRA does not define "good faith," id. at 957, but it provides several textual and structural clues on how courts are to enforce it. Two provisions, in particular, guide that inquiry:

First, IGRA's list of permissible topics for Tribal-State compacts. Section 2710(d)(3)(C) provides that a Tribal-State compact "may include provisions relating to—":

(i) the application of the criminal and civil laws and regulations of the Indian tribe or the State that are directly related to, and necessary for, the licensing and regulation of such activity;

(ii) the allocation of criminal and civil jurisdiction between the State and the Indian tribe necessary for the enforcement of such laws and regulations;

(iii) the assessment by the State of such activities in such amounts as are necessary to defray the costs of regulating such activity;

(iv) taxation by the Indian tribe of such activity in amounts comparable to amounts assessed by the State for comparable activities;

(v) remedies for breach of contract;

(vi) standards for the operation of such activity and maintenance of the gaming facility, including licensing; and

(vii) any other subjects that are directly related to the operation of gaming activities.

Through this provision, "IGRA limits permissible subjects of negotiation in order to ensure that tribal-state compacts cover only those topics that are related to gaming." Rincon Band of Luiseno Mission Indians v. Schwarzenegger , 602 F.3d 1019, 1028–29 (9th Cir. 2010).

And the Supreme Court has explained that "gaming activities" in the § 2710(d)(3)(C) context is narrowly construed: it "means just what it sounds like—the stuff involved in playing class III games." Michigan v. Bay Mills Indian Cmty. , 572 U.S. 782, 792, 134 S.Ct. 2024, 188 L.Ed.2d 1071 (2014). It refers only to "what goes on in a casino—each roll of the dice and spin of the wheel"—not to "off-site licensing or operation of the games." Id. ; see also id. ("[G]aming activity is the gambling in the poker hall, not the proceedings of the off-site administrative authority."); id. (An agency's power to " ‘close a gaming activity’ " means "to shut down crooked blackjack tables, not the tribal regulatory body meant to oversee them." (simplified)).

The list of permissible negotiation topics is also circumscribed by "one key limitation on state negotiating authority" found in § 2710(d)(4). Rincon , 602 F.3d at 1028. Under that provision, IGRA generally prevents a State from "impos[ing] any tax, fee, charge, or other assessment upon an Indian tribe ... to engage in a class III activity." § 2710(d)(4).

Second, IGRA tells courts how to analyze good faith. It establishes that, in determining "whether a State has negotiated in good faith, the court—"

(I) may take into account the public interest, public safety, criminality, financial integrity, and adverse economic impacts on existing gaming activities, and

(II) shall consider any demand by the State for direct taxation of the Indian tribe or of any Indian lands as evidence that the State has not negotiated in good faith.

§ 2710(d)(7)(B)(iii).

These provisions work in tandem to determine whether a State has met its good-faith duty.

To start, § 2710(d)(3)(C) enumerates permissible topics that may be negotiated as part of a Tribal-State compact. If a tribe shows that a State violated § 2710(d)(3)(C) by negotiating for an off-list topic, that serves as "evidence" of a State's bad faith under IGRA's burden-shifting framework. See § 2710(d)(7)(B)(ii). At that point, the burden shifts to the State to prove its good faith. Id.

Courts then assess a State's claim of good faith under § 2710(d)(7)(B)(iii). Section 2710(d)(7)(B)(iii) provides two ways to analyze good faith. First, "a court must consider a ‘demand’ for a tax to be made in bad faith." Rincon , 602 F.3d at 1029 ; see also § 2710(d)(7)(B)(iii)(II). In other words, such a demand is "evidence of the State's bad faith." Rincon , 602 F.3d at 1030. Second, "the State may attempt to rebut bad faith" by showing that any off-list topic was justified by "the public interest, public safety, criminality, financial integrity, and adverse economic impacts on existing gaming activities." Id. at 1032 ; see also § 2710(d)(7)(B)(iii)(I). In other words, the State can use these statutory factors to establish its good faith despite negotiating for off-list topics.

I explain IGRA's meaning in greater detail below.

A.

The first step in the good-faith inquiry is understanding that § 2710(d)(3)(C) creates an exhaustive list of permissible, gaming-related topics that a State may bargain for as part of a compact with a tribe. In other words, § 2710(d)(3)(C) creates a binary of on-list/off-list subjects with only on-list topics permissible for negotiation. I would reach this conclusion for three reasons.

One reason is the text. Section 2710(d)(3)(C)'s language shows that it limits the field of permissible negotiation topics. It states that a gaming compact "may include" seven gaming-related topics. § 2710(d)(3)(C)(i)–(vi). While the word "may" is ordinarily considered permissive, its meaning is judged based on context. Cf. Black's Law Dictionary (11th ed. 2019) ("In dozens of cases, courts have held may to be synonymous with shall or must [.]"). Based on the context here, "may" should be interpreted as mandatory or as meaning "may only." That's because the provision identifies only gaming-relating topics as permissible subjects. By detailing a list of six specific gaming issues, along with a catchall provision that permits other topics "directly related to the operation of gaming activities," § 2710(d)(3)(C)(vii), Congress limited the permissible gaming compact topics to activities that are gaming related. Put simply, there would be no need for Congress to include six specific topics and a limited catchall if States were free to insert extraneous topics into gaming compacts.

The second reason is our interpretative canons. Two well-established canons confirm § 2710(d)(3)(C)'s exhaustive nature. First, the expressio unius canon says that the identification of related topics within an associated group generally excludes unrelated topics. See NLRB v. SW Gen., Inc. , 580 U.S. 288, 137 S. Ct. 929, 940, 197 L.Ed.2d 263 (2017) ; see also Wheeler v. City of Santa Clara , 894 F.3d 1046, 1054 (9th Cir. 2018). Second, the ejusdem generis principle provides that a "general term following more specific terms means that the things embraced in the general term are of the same kind as those denoted by the specific terms." United States v. Lacy , 119 F.3d 742, 748 (9th Cir. 1997) (simplified). Both canons work here to show § 2710(d)(3)(C)'s exclusive nature. For example, a "sign outside a veterinary clinic saying ‘Open for treatment of dogs, cats, horses, and all other farm and domestic animals’ does suggest (by its detail) that the circus lion with a health problem is out of luck." Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 107–08 (2012). IGRA permits six detailed subjects and a seventh catchall related to gaming activities—the negative implication of § 2710(d)(3)(C) is that unrelated subjects are impermissible.

The third reason—our precedent. In Rincon , we explained that "[t]he language and structure of § 2710(d)(3)(C) suggest[ ] it is exhaustive." 602 F.3d at 1028 n.9. We observed that "[a]lthough ‘may’ indicates permissiveness ... to grant permission is not necessarily to grant carte blanche. What is ‘permitted’ is limited." Id. I agree with Rincon 's suggestion and would hold that the § 2710(d)(3)(C) list is exhaustive.

B.

Given § 2710(d)(3)(C)'s limits, a State's violation of its list of permissible topics constitutes evidence of a State's lack of good faith. When a State limits its negotiation to on-list, gaming-related subjects, that is evidence that the State has operated in good faith. Off-list negotiations, however, show the opposite and require further scrutiny. There are several textual and structural reasons why this is true.

First, IGRA's burden-shifting framework makes clear that evidence of off-list negotiation impacts the good-faith analysis. As a reminder, plaintiff tribes have the initial burden of "introduc[ing] ... evidence" of a State's refusal to negotiate a compact in "good faith." § 2710(d)(7)(B)(ii). Once the tribe meets that evidentiary showing, the "burden of proof" shifts to the State to prove that it has acted in good faith. Id. By enumerating an exhaustive list of negotiable subjects, Congress provided tribes with a way to meet their initial burden of proof—by showing that a State violated § 2710(d)(3)(C). In other words, going beyond the scope of § 2710(d)(3)(C) is evidence that a State has violated its good-faith duty under the burden-shifting framework.

Second, the § 2710(d)(3)(C) list notes that it applies to "[a]ny Tribal-State compact negotiated under subparagraph (A). " § 2710(d)(3)(C) (emphasis added). Subparagraph A of § 2710(d)(3)(A) establishes the good-faith requirement. Thus, the list is textually connected to a State's good-faith duty.

Third, a contrary reading would render § 2710(d)(3)(C) nearly meaningless. If a violation of § 2710(d)(3)(C) did not constitute evidence of bad faith, then the provision would have little significance in IGRA's overall structure. Instead, by linking the list to evidence of bad faith, this interpretation "give[s] effect ... to every clause and word" of IGRA. United States v. Menasche , 348 U.S. 528, 538–39, 75 S.Ct. 513, 99 L.Ed. 615 (1955) (simplified).

So if a tribe has introduced evidence that a State sought provisions outside the confines of the § 2710(d)(3)(C) list, then the tribe has satisfied its burden to show the State's bad faith and the burden shifts to the State to prove its good faith. But if a tribe fails to show that a State has put off-list topics on the negotiation table, then the tribe hasn't met its initial burden and the State may prevail. ² In effect, Congress gives a State free rein to haggle for permissible gaming-related topics—but if the State overreaches by negotiating for topics outside of § 2710(d)(3)(C), then it is subject to greater scrutiny.

C.

Greater scrutiny comes in the form of § 2710(d)(7)(B)(iii)(I)'s good-faith factors. As we have said, the "good faith inquiry is nuanced and fact-specific." Coyote Valley II , 331 F.3d at 1113. So a State's negotiation for an off-list subject doesn't automatically lead to a finding of bad faith. Congress instead provided several factors that courts "may take into account" when determining a State's good faith. § 2710(d)(7)(B)(iii)(I). These factors include the "public interest, public safety, criminality, financial integrity and adverse economic impacts on existing gaming activities." § 2710(d)(7)(B)(iii)(I). Based on these factors, a State may prove that its off-list negotiations still amounted to good faith.

While the five good-faith factors may appear freewheeling, they aren't to be construed "broadly in favor of the State's interests." Rincon , 602 F.3d at 1032. Instead, "those terms clearly apply to protecting the State against the adverse consequences of gaming activities." Id. So the good-faith factors are to be read considering "legitimate state interests regarding gaming and the purposes of IGRA." Id. at 1039. The factors must also be read in line with IGRA's statutory command to "promot[e] tribal economic development, self-sufficiency, and strong tribal governments" and "ensure that the Indian tribe is the primary beneficiary of the gaming operation." § 2702(1), (2).

In short, IGRA's good-faith analysis works like this:

(1) If a tribe fails to show that a State negotiated for topics outside of § 2710(d)(3)(C), absent more, it has failed to meet its initial showing of bad faith and the inquiry ends.

(2) But if a tribe shows that a State negotiated on a topic outside the list, it has satisfied its initial burden of proving bad faith; and

(3) The burden then shifts to the State to show its good faith under the § 2710(d)(7)(B)(iii)(I) factors.

While not the most straightforward statute, this reading interprets IGRA "as a symmetrical and coherent regulatory scheme ... and fit[s] ... all [its] parts into an harmonious whole." FDA v. Brown & Williamson Tobacco Corp. , 529 U.S. 120, 133, 120 S.Ct. 1291, 146 L.Ed.2d 121 (2000) (simplified). D.

With the proper legal framework in mind, I would hold that California negotiated on four topics outside the scope of the § 2710(d)(3)(C) list—spousal and child support orders, environmental mitigation requirements, local government agreements, and state tort law remedies. The Tribes thus met their burden to show that California exceeded the scope of IGRA's permissible-topics list. And I would hold that the burden shifted to California to prove that it complied with its good-faith duty under the factors provided in § 2710(d)(7)(B)(iii)(I). I would therefore vacate the district court's grant of summary judgment for the Tribes and remand so the district court may conduct the good-faith analysis under the statutory factors in the first instance.

II.

My main disagreement with the majority results from violating § 2710(d)(3)(C) —IGRA's list of permissible topics. The majority asserts that when a State negotiates for a topic "well outside" IGRA's list of permitted topics, it constitutes per se bad faith—automatically triggering IGRA's remedial procedures. See Maj. Op. 1041 ("Accordingly, when a state has demanded that a tribe negotiate on topics well outside IGRA's scope, it follows that the state has not negotiated in good faith, end of story."). But there are several problems with the majority's interpretation.

A.

First, the majority relies heavily on its sense of IGRA's core "principles" and "objectives." Maj. Op. 1040–41. Instead of relying on text and structure, the majority tells us "what Congress had in mind" when it enacted IGRA, Maj. Op. 1039–41, and construes the statute from there. For example, we are told to intuit Congress's "aware[ness] of the danger that states could use their compacting approval powers to encroach on tribal sovereignty," Maj. Op. 1032, and to read the provisions at issue here based on "core aspects of tribal sovereignty," Maj. Op. 1039. This is not a text-based approach; it's a purpose-driven one. And what the majority hints at, the concurrence makes clear. The concurrence says that we should not be confined by "the text of the statute and our precedents construing IGRA," but instead we should glean meaning from IGRA's "purpose," "legislative history," and atextual canons of construction. Concurrence at 1049–50.

But of course, the analysis must start and end with IGRA's text and structure—regardless of whether we believe it best achieves the statute's purpose. As judges, we must reject "surmise about legislative purpose" and instead look to a statute's text and structure for guidance. HollyFrontier Cheyenne Ref., LLC v. Renewable Fuels Assn. , ––– U.S. ––––, 141 S. Ct. 2172, 2181–83, 210 L.Ed.2d 547 (2021). The majority tries to cloak its reliance on IGRA's purposes by calling it the "context of the statute as a whole." Maj. Op. 1047. But it is one thing to look to context to interpret the words of a statute; it is another to use context to divine some broad legislative purpose to override the statute's plain meaning. See John F. Manning, What Divides Textualists from Purposivists? , 106 Colum. L. Rev. 70, 92–93 (2006) (explaining the difference between "contextual evidence of semantic usage" and "contextual evidence of the policy considerations that apparently justified the statute").

Contrary to the view of my colleagues, "we can never let perceived legislative purpose eclipse the ordinary meaning of statutory text." Rojas v. FAA , 989 F.3d 666, 695 (9th Cir. 2021) (en banc) (Bumatay, J., dissenting in part). Indeed, "[l]awmaking is not a tidy affair;" "[i]t can be a clumsy, inefficient, even unworkable process." Id. (simplified). So, in my view, it is a fruitless exercise to try to divine why Congress chose to write IGRA the way it did. For example, the majority wonders why Congress inserted the good-faith factors into IGRA when doing so could "undermine" the statute's "crystalline" objective. Maj. Op. 1045–46. But such arguments have little to do with uncovering meaning and ignore what are often "hard-fought compromises" in legislation. Bd. of Governors of Fed. Reserve Sys. v. Dimension Fin. Corp. , 474 U.S. 361, 374, 106 S.Ct. 681, 88 L.Ed.2d 691 (1986). The Supreme Court has warned us that "[c]ourts are not authorized to rewrite a statute because they might deem its effects susceptible of improvement." Badaracco v. Comm'r , 464 U.S. 386, 398, 104 S.Ct. 756, 78 L.Ed.2d 549 (1984). Rather than question the "wisdom" of Congress, see Maj. Op. 1046–47, we must enforce the statute Congress wrote—not offer judicial amendments for the sake of sound policy. ³

B.

Most fundamentally, the majority's reading of IGRA conflicts with its plain text. The majority concludes that a State's negotiation for topics "well outside" § 2710(d)(3)(C) automatically constitutes bad faith. See Maj. Op. 1049. But that bears little resemblance to the law Congress enacted. IGRA establishes that courts "may take into account" the good-faith factors—public interest, public safety, criminality, financial integrity and adverse economic impacts—before "determining ... whether a State has negotiated in good faith." § 2710(d)(7)(B)(iii)(I). The majority's approach effectively erases these factors from the statute.

Instead, the majority's automatic bad-faith rule bars courts from even considering the factors before forcing parties into IGRA's remedial procedures. So while the majority sees a violation of § 2710(d)(3)(C) as the "end of [the] story," see Maj. Op. 1040–41, IGRA tells us that courts may continue to analyze good faith based on the statutory good-faith factors. Essentially, the majority substitutes "may not" where the statute clearly says "may"—something we are not authorized to do.

Such a reading is also inconsistent with IGRA's structure. IGRA authorizes States to impose assessments on a tribe's gaming activities, but precludes them from imposing any tax or assessment "upon an Indian tribe" itself. § 2710(d)(3)(C), 2710(d)(4). In the same subsection as the good-faith factors, IGRA establishes that courts "shall consider any demand by the State for direct taxation of the Indian tribe ... as evidence that that the State has not negotiated in good faith." § 2710(d)(7)(B)(iii)(II) (emphasis added). So even when a State unquestionably violates IGRA's list of permissible topics, such negotiation still doesn't trigger automatic bad faith because IGRA tells courts to consider such a demand "as evidence" of bad faith. Id. If any off-list negotiation constituted automatic bad faith (as the majority contends), there would be no point in making "direct taxation of the Indian tribe" evidence of bad faith since that topic is already outside the list of permissible subjects. ⁴ The majority thus neglects IGRA's nuanced scheme for analyzing good faith by fashioning a per se rule divorced from the text.

At the same time, the majority also creates a new atextual test for complying with § 2710(d)(3)(C). Now, § 2710(d)(3)(C) is seemingly violated only when a State's proposed topic is "well outside" or "far exceeds" the list of permissible topics. Maj. Op. 1040–41 Where does this test come from? And how exactly are district courts supposed to administer it? The majority doesn't say.

Now judges will be tasked with resolving a multi-step test whenever they confront an IGRA challenge. First, courts will have to draw a line between on-list and off-list topics. Second, courts will need to define when an off-list topic is "well outside" the list. How will they know when a topic "far exceeds" or is just "slightly" outside the list? See Maj. Op. 1041 n.5. Presumably, courts will have to make it up. And what happens when a topic is slightly outside the list? Does the State get a pass? If so, that conflicts with the majority's view that the "exhaustive nature of IGRA's enumerated list ... means nothing if states can ultimately go beyond that list." Maj. Op. 1041. In my view, these tough questions are all unnecessary given that the words "far exceed" or "well outside" appear nowhere in IGRA's text.

Perhaps this confusing framework is because the majority is unwilling to live with the consequences of its automatic bad-faith rule—that any deviation from the list is per se bad faith and triggers IGRA's remedial procedures. Take this case as an example. The negotiations between California and the Tribes took place over four years, with 39 days of in-person negotiations, and 26 draft compacts exchanged between the parties. Under the majority's framework, if California proposes any topic or proposal minimally outside the list, the Tribe could go to court, petition for the remedial procedures, and wipe out all those hard days of negotiation. Seemingly too much to stomach, the majority gives itself breathing room by fashioning the "well outside" requirement out of whole cloth. So now, instead of relying on the statutory factors of good faith, parties will now have to litigate the inches between what's "permissible" and what's "well outside" the line.

C.

The majority tries to preserve some use for the good-faith factors by suggesting that they can be used to evaluate a State's on-list negotiations. Maj. Op. 1043–44. But such a reading makes complying with § 2710(d)(3)(C) nearly meaningless. Even when a State follows § 2710(d)(3)(C)'s list of permissible topics to a tee, the majority would still require the State to defend its negotiations under the good-faith factors. This approach dramatically expands IGRA's scope, abandons the statute's burden-shifting framework, and makes litigation more likely.

While the majority agrees that § 2710(d)(3)(C) constitutes a list of permissible negotiation topics, those topics may not be so permissible after the majority's good-faith analysis. In other words, if a State does exactly what IGRA tells it to do by negotiating for on-list topics, the majority would still have courts review every detail of the negotiations to sniff out evidence of bad faith. But IGRA's text does not give courts a license to put every negotiation detail under the microscope—especially when a State negotiates for exactly what IGRA allows.

As a supposed justification for its approach, the majority imagines a scenario where a State bargains for a permissible, on-list topic, but the proposal would be "highly burdensome" and "unnecessary." Maj. Op. 1042. I agree that negotiating for "unnecessary" topics may constitute bad faith in some cases, but we need not adopt the majority's atextual framework to get there. Under a proper reading of IGRA, a tribe may still allege that a State engaged in procedural bad faith by disguising an unnecessary poison pill as a substantively permissible topic. Cf. Pauma Band , 973 F.3d at 963, 966 (analyzing a claim of procedural bad faith for including "unduly harsh" language in the draft compact).

III.

In the end, my disagreement with the majority comes down to principles of statutory interpretation. Rather than interpreting text and structure, the majority instead focuses on reading IGRA based on its "core principles," Maj. Op. 1040–41 "core objectives," id. at 1040–41; "legislative intent," id. at 1046–47; "legislative history," Concurrence at 1049–50; and "purpose," id. at 1054. The unfortunate result is the expansion of IGRA beyond what its text allows.

For these reasons, I respectfully dissent.