dissenting:
This case requires us to balance two important governmental interests: the ability of states to control the use of their own funds, and the federal government’s national labor policy. The majority’s critical error is in failing to recognize our responsibility to honor both of these interests to the extent possible — instead, it gives short shrift to California’s sovereign*995ty interests through an overbroad application of federal labor preemption doctrine that conflicts with both our past precedent and that of the Supreme Court. The majority’s preemption analysis subverts important federalism principles, and therefore I must respectfully dissent.
I agree with many — although not all — of the majority’s characterizations of the complex labor preemption doctrine upon which this case turns. My main difference is in how I answer the question upon which the fate of AB 1889 turns: if a state allows employers to spend their own funds, in whatever manner they please, to advocate for or against unionization, does national labor preemption doctrine require the full preemption of a restriction on the use by employers of state funds alone for such purposes? I think not. As the Supreme Court has recognized, the State of California has the right to restrict the use of its own funds; and a more nuanced application of labor preemption law reveals that although certain parts of the statute may be preempted, its core restriction should survive.
“We are reluctant to infer preemption,” Bldg. & Constr. Trades Council of the Metro. Dist. v. Associated Builders & Contractors of Mass./R.I., Inc., 507 U.S. 218, 224, 113 S.Ct. 1190, 122 L.Ed.2d 565 (1993) (“Boston Harbor”), and any analysis of preemption begins with the “basic assumption that Congress did not intend to displace state law.” Maryland v. Louisiana, 451 U.S. 725, 746, 101 S.Ct. 2114, 68 L.Ed.2d 576 (1981).The majority concludes that AB 1889 is wholly preempted under both the Garmon and Machinists strands of labor preemption doctrine, so I will separately explain why neither conclusion is correct. See San Diego Bldg. Trades Council v. Garmon, 359 U.S. 236, 79 S.Ct. 773, 3 L.Ed.2d 775 (1959); Lodge 76, International Association of Machinists & Aerospace Workers v. Wisconsin Employment Relations Commission, 427 U.S. 132, 96 S.Ct. 2548, 49 L.Ed.2d 396 (1976) (“Machinists”). However, although the two doctrines require different analyses, the majority misses a fundamental truth that explains why neither doctrine requires the preemption of AB 1889. As the Supreme Court has explained in a related context, and as logic dictates, when a state restricts merely the use of state funds for a given activity — without regulating an employer’s ability to pursue that activity with its own funds — the state is not directly regulating that activity. Because both the Garmon and Machinists preemption doctrines share a predicate inquiry into what activity is being regulated, the fact — ignored by the majority — that California is not actually regulating the speech at issue precludes full preemption under either doctrine.
I. AB 1889 Is Not Preempted Under Garmon
Garmon preemption arises when there is an actual or potential conflict between state regulation and federal labor law— when a state regulates activity that is actually or arguably protected or prohibited by the NLRA. Garmon, 359 U.S. at 244, 79 S.Ct. 773. The majority concludes that AB 1889 is preempted because it regulates activity that is actually protected by the NLRA.
First, employer union-related speech is not actually protected by the NLRA, but rather protected by the First Amendment, and therefore simply left unregulated by the NLRA. AB 1889 is therefore properly analyzed under the Machinists strand of labor preemption doctrine. Second, AB 1889 does not actually regulate such activity in any case. Third, were I wrong about both of these conclusions, AB 1889 would still not be preempted under Garmon.
*996A. The NLRA Does Not “Actually Protect” Employer Speech
The majority concludes that the NLRA actually protects employer free speech rights — that it affirmatively grants protection to employers’ speech on unionization. An examination of the structure and text of the NLRA demonstrates that the majority has misread it.
Section 7 of the Act is entitled “Right of employees as to organization, collective bargaining, etc.” 29 U.S.C. § 157. It identifies areas of protected employee conduct, and can fairly be characterized as setting forth those employee practices that are actually protected by the NLRA. Section 8, conversely, is entitled “Unfair Labor Practices.” 29 U.S.C. § 158. By its plain terms, it sets forth activities that are actually prohibited by the NLRA. It is easy to see how both Sections 7 and 8, then, can be implicated in a Garmon preemption analysis: if a state regulates employee activities that are actually protected under Section 7, or activities by either employers or labor unions that are actually prohibited under Section 8, that regulation will be preempted.
The majority, however, has a novel view of Section 8’s third subsection, which is entitled “Expression of views without threat of reprisal or force or promise of benefit.” 29 U.S.C. § 158(c). This subsection has been termed the “free speech” exemption to Section 8’s definition of unfair labor practices, because it carves out noncoercive speech from the category of actually prohibited activity. From this statutory structure, the majority concludes, “[w]e hold that Section 8(c) of the [NLRA] ... explicitly protects the right of employers to express their views about unions and union organizing efforts.” Maj. Op. at 12185. This is simply wrong, explaining why this circuit has never so held, requiring the majority to do it now, so late in the day. See, e.g., Hotel Employees, Restaurant Employees Union, Local 2 v. Marriott Corporation, 961 F.2d 1464, 1470 n. 9 (9th Cir.1992) (“[Sjection 8(c) merely states an employer does not commit an unfair labor practice by expressing its views regarding unionization.”); UAW-Labor & Employment & Training Corp. v. Chao, 325 F.3d 360, 364-65 (D.C.Cir.2003) (“Fitting a Garmon claim under the language of § 8(c) is awkward.... [T]he activities described in § 8(c) ... are not ‘protected by ’ the NLRA, except from the NLRA itself.”) (emphasis in original).
The majority sets forth no textually grounded explanation for its conclusion that one section of the Act, entitled “Unfair Labor Practices,” actually grants an affirmative right, and instead offers only misleading quotations from legislative history and caselaw. Again and again, the majority cites references to employers’ “right of free speech.” See, e.g., Maj. Op. at 982. But there is no mystery about the actual source of this right. It is the First Amendment. When Congress passed the NLRA and set forth a list of prohibited actions, it saw fit to clarify that noncoer-cive speech- — the sort protected by the First Amendment — did not constitute an unfair labor practice. Indeed, without such a qualifier, the constitutionality of the NLRA might well have been at issue.1 The majority’s authorities never go any further than explaining that employer speech is a necessary part of healthy labor *997relations. The majority seeks to use these broad platitudes to support its illogical reading of the NLRA, in order to identify an “actually protected right” in the Act itself.
This is not to say that there is no preemption issue here — it is just that it is best understood as a Machinists issue, not a Garmon issue. As I shall illustrate below, Machinists preemption is the more appropriate doctrine under which to evaluate AB 1889, because the arguable impermissible effects of AB 1889 are on activities meant to be left unregulated by the NLRA, not because of any direct conflict with the protections and prohibitions of the Act.2
B. Even If AB 1889 Arguably Protects Employer Speech Rights, Garmon Preemption is Inappropriate
It should be clear that Section 8(c) does not definitely protect employer speech rights; even if one were to disagree with my reading of the NLRA, one would have to grant that these two alternative readings present an unresolved conflict in our circuit. Let us assume, then, that AB 1889 arguably protects employer speech rights. Nonetheless, Supreme Court precedent shows that Garmon preemption is not appropriate here.
1. The Local Interest Exception
The Supreme Court has cautioned that “inflexible application of [the Garmon ] doctrine is to be avoided, especially where the State has a substantial interest in regulation of the conduct at issue and the State’s interest is one that does not threaten undue interference with the federal regulatory scheme.” Farmer v. United Bhd. of Carpenters & Joiners, 430 U.S. 290, 302, 97 S.Ct. 1056, 51 L.Ed.2d 338 (1977). When conducting a Garmon preemption inquiry into whether regulation overlaps with NLRB jurisdiction, we must therefore conduct “a balanced inquiry into such factors as the nature of the federal and state interests in regulation and the potential for interference with federal regulation.” Id. at 300, 97 S.Ct. 1056. This is to avoid preempting state regulation of conduct that involves “interests so deeply rooted in local feeling and responsibility that, in the absence of compelling congressional direction, we [cannot] infer that Congress ha[s] deprived the States of the power to act,” or where “the activity regulated [is] a merely peripheral concern” of the NLRA. Garmon, 359 U.S. at 243-44, 79 S.Ct. 773; see Belknap, Inc. v. Hale, 463 U.S. 491, 498, 103 S.Ct. 3172, 77 L.Ed.2d 798 (1983).
The previously recognized exceptions to Garmon preemption have involved exercises of state court jurisdiction over universally recognized common law torts, rather than, as here, the exercise of a state’s spending power. See, e.g., Linn v. United Plant Guard Workers of America, 383 U.S. 53, 62, 86 S.Ct. 657, 15 L.Ed.2d 582 (1966) (“[A] State’s concern with redressing malicious libel is ‘so deeply rooted in local feeling and responsibility’ that it is not preempted despite arguable overlapping NLRB jurisdiction over the speech at issue.”) (quoting Garmon, 359 U.S. at 244, 79 S.Ct. 773). But the logic that compelled the Supreme Court to recognize the exception in the former category applies just as powerfully — if not more so— to local spending decisions. A state’s con*998trol of its own purse strings is of at least as great concern to it as its power to regulate defamatory speech or trespassing. Just as the state has a responsibility to protect its citizens from such torts, so it has a responsibility and a right to spend its treasury — largely generated from the pockets of its citizens — based on principles and guidelines that the democratically elected legislature of that state deems to be appropriate. Such spending decisions are, of course, subject to federal supremacy concerns. But the Supreme Court has commanded us to be extremely cautious before concluding that a federal regulatory scheme is meant to intrude upon such a fundamental state prerogative.
The majority gets the denominator wrong when it concludes that “regulating employer speech in the labor relations context ... is not a traditional state interest, nor is it ‘so deeply rooted in local feeling and responsibility’ as to invoke the exception.” Maj. Op. at 992. The majority ignores the broader state interest at issue — control over its own fisc — and inappropriately second-guesses the California legislature’s motivations in exercising this prerogative. In an era of tightening budgets, where many important competing interests vie for every dollar of a state’s treasury, it is all the more important that states retain their right to determine the best way to allocate their scarce resources. Today, the majority effectively forces California to fund employers’ union-related expression with those scarce dollars.
2. The Supreme Court’s Refinement of Garmon in Sears
In Sears, Roebuck & Co. v. San Diego County Dist. Council of Carpenters, 436 U.S. 180, 98 S.Ct. 1745, 56 L.Ed.2d 209 (1978), the Supreme Court refined its Gar-mon preemption doctrine in the context of an employer’s common law trespassing suit against picketing union members where the picketing was “arguably — but not definitely' — prohibited or protected by federal law.” Sears, 436 U.S. at 182, 98 S.Ct. 1745. Sears divided the inquiry into two related but distinct questions: whether the state court’s jurisdiction over the trespassing claim was preempted (1) by the arguably prohibited nature of the picketing, or (2) by its arguably protected nature.
As to whether the union’s picketing was arguably prohibited by the NLRA, Sears articulated a relatively straightforward “primary jurisdiction” test: if the claim considered by the state tribunal is identical to one that could be presented to the NLRB, the state’s jurisdiction is preempted. Id. at 197, 201, 98 S.Ct. 1745; see also Belknap, 463 U.S. at 511, 103 S.Ct. 3172 (applying primary jurisdiction test to state regulation of arguably prohibited conduct). As to whether the picketing was arguably protected by the NLRA, the Court went beyond the primary jurisdiction test to address additional federal supremacy concerns — whether, despite the lack of identi-cality between issues the state court and NLRB might consider, preemption was warranted to protect against the risk of “misinterpretation of [the NLRA] and the consequent prohibition of protected conduct.” Sears, 436 U.S. at 203, 98 S.Ct. 1745. We employed this “primary jurisdiction plus” approach in Radcliffe v. Rainbow Const. Co., 254 F.3d 772, 786 (9th Cir.2001) (holding that state jurisdiction over claims by union members against employer for false arrest, false imprisonment and malicious prosecution were not preempted under Garmon ).3
*999a. Primary Jurisdiction
The parties do not dispute that the NLRB has no interest in resolving the central controversy that a state court would have to resolve in enforcing AB 1889 — whether state funds were used to “assist, promote, or deter union organizing.” The focus of a lawsuit under AB 1889 would not be on whether employer speech was proper or improper, but on whether state funds were used to support that speech. Far from being identical to the kind of question the NLRB might consider, a suit under the California statute would be an accounting for the employer’s possible misuse of funds, with no attention at all to the nature of the advocacy itself.
The Chamber of Commerce makes two arguments in response. It first argues that a state court enforcing AB 1889 would in some instances have to determine whether a union was a “labor organization” under § 16647, an area that it claims is reserved to the NLRB under Marine Engineers Beneficial Ass’n v. Interlake Steamship Co., 370 U.S. 173, 82 S.Ct. 1237, 8 L.Ed.2d 418 (1962). But even if the state court had to make such a determination, it would be relevant only to the ultimate question of whether an employer spent state funds on advocacy, not whether the advocacy violated the NLRA. We have previously rejected an argument that the incidental determination by a state court of whether persons were engaged “in lawful union activity” was sufficient to create Garmon preemption, where the focus of a state proceeding was on “state concerns of accommodating such union activity with the state-law rights of private property.” Radcliffe, 254 F.3d at 786. Moreover, AB 1889 is not comparable to the Minnesota statute at issue in Marine Engineers, under which a state law determination that certain groups were not labor organizations permitted the state court to regulate picketing and other activities identical to those that could have been raised before the NLRB. Marine Engineers, 370 U.S. at 176, 82 S.Ct. 1237.
The Chamber of Commerce also argues that because AB 1889 restricts an employer’s ability to “influence” its employees using state funds, see § 16645(a), California courts would effectively be deciding whether employers had improperly acted under § 8(a) of the NLRA “to restrain or coerce employees in the exercise of the rights contained in [§ 7 of the NLRA].” 29 U.S.C. § 158(a) (NLRA § 8(a)). However, were the NLRB to consider an unfair labor practice charge arising from the employer’s conduct, it would focus on whether the employer had interfered with the employees’ § 7 rights, regardless of whether the employer used state funds in the process. Under AB 1889, the California court would be determining only whether an employer used state funds for any attempt whatsoever to influence employees, not whether that attempt violated the NLRA. Thus, because the statute is focused only on control over the use of state funding, there is no identicality of claims, and the primary jurisdiction test is not met.4
b. Additional Supremacy Concerns
Because of the arguably protected character of the speech at issue, we would next assess whether the state statute is preempted because Congress would “prefer[ ] the costs inherent in a jurisdictional *1000hiatus to the frustration of national labor policy which might accompany the exercise of state jurisdiction.” Sears, 436 U.S. at 203, 98 S.Ct. 1745.
For essentially the same reasons detailed above in the primary jurisdiction analysis, the risk that a state court applying AB 1889 could “misinterpret[ ] ... federal law,” id., is nonexistent in this case. Not only is there no identicality of claims, but the subject matter of an AB 1889 suit is so far removed from the NLRA’s primary focus — the determination of what constitutes an unfair labor practice — that any rationale for Garmon preemption is absent.
Sears itself provides a strong analogy. The Court, after noting the state interest in hearing trespass claims, identified a clear potential overlap between the NLRB’s jurisdiction and that of the state court: “[T]he state court was obligated to decide [whether] the trespass was not actually protected by federal law, a determination which might entail an accommodation of Sears’ property rights and the Union’s § 7 rights. In an unfair labor practice proceeding initiated by the Union, the Board might have been required to make the same accommodation.” Id. at 201, 98 S.Ct. 1745. The Court further noted that the trespass at issue was arguably protected by the NLRA, whereas previously recognized exceptions to Gar-mon preemption did not “involve[] protected conduct.” Id. at 204, 98 S.Ct. 1745. Nevertheless, the Court concluded that the risk that regulation of trespass might impermissibly trench upon the NLRB’s jurisdiction over the speech rights of union members was too unlikely to justify usurpation of the state’s prerogative. There was no “significant risk of prohibition of protected conduct,” so the Court was “unwilling to presume that Congress intended the arguably protected character of the [regulated] conduct to deprive the California courts of jurisdiction to entertain Sears’ trespass action.” Id. at 207, 98 S.Ct. 1745.
Here, I have noted the important state interest in determining how its funds are spent. Even if the exercise of this state prerogative could have some effect on the arguably protected advocacy rights of employers, given the nature and functioning of the statute — which has no concern for whether the speech at issue was protected by the NLRA, but only for where the money to fund it came from- — there is no “significant risk of prohibition of protected conduct.” Id. Indeed, the state interest is so strong that we cannot “presume that Congress intended the arguably protected character of the [regulated] conduct to deprive” California of the ability to control the use of its own funds in this manner. Id.
II. AB 1889 Is Not Wholly Preempted Under Machinists
Machinists preemption operates as a form of “labor field preemption” — it requires the preemption of any state regulation of activity that although not directly regulated by the NLRA, was intended by Congress “to be controlled by the free play of economic forces.” Machinists, 427 U.S. at 140, 96 S.Ct. 2548 (internal quotation and citation omitted). AB 1889 is best viewed through this lens, because the employer expressive activity at issue is not protected or prohibited by the NLRA, but rather is an essential part of the NLRA’s overall labor relations structure.
In enacting a restriction on the use of state funds with the purpose of remaining neutral in labor disputes, California has not intruded on labor relations meant to be left unregulated by the states. The Supreme Court has specifically emphasized the legitimate governmental interest in re*1001maining neutral in labor disputes. See Lyng v. UAW, 485 U.S. 360, 371-72, 108 S.Ct. 1184, 99 L.Ed.2d 380 (1988) (holding that a Congress’ decision to avoid “subsidization” of striking workers through a food stamp program was legitimately based in Congress’s “considered efforts to avoid favoritism in labor disputes”; noting that “we have little trouble in concluding that [the food stamp restriction] is rationally related to the legitimate government objective of avoiding undue favoritism to one side or the other in private labor disputes”). Indeed, state neutrality is not only consistent with the NLRA — it is in many cases mandated by the NLRA. Were a state to opt against neutrality by intervening on the side of employer or employee in a labor dispute, such an intervention could itself raise preemption problems. See Golden State Transit Corp. v. City of Los Angeles, 475 U.S. 608, 619, 106 S.Ct. 1395, 89 L.Ed.2d 616 (1986) (“Even though agreement is sometimes impossible, government may not step in and be a party to the negotiations.”). Thus, California’s decision to avoid having its own funds used on either side of a labor conflict is clearly legitimate.5
Nor does the restriction on the use of state funds interfere with an employer’s ability to engage in “self-help” in union organizing in the sense protected by Machinists. Restricting the use of the state’s funds does not imply an attempt to alter an employer’s private decision to support or oppose unionization. An employer who wishes to oppose unionization using its own funds will not suffer any penalty in doing so. It is ironic that a doctrine meant to ensure that employers can engage in self-help is applied today to force government to help employers oppose unionization.
To be sure, under AB 1889 those employers in California who are currently receiving money from the state will be prevented from using those funds to speak about unionization. Because, according to the Chamber of Commérce, some employers are entirely funded by the state, under the statute some California employers may be prevented from spending any portion of their budgets to advocate during a labor dispute. However, such employers will be prevented from labor advocacy only because they have already chosen to fund their budgets entirely from the state— itself a business decision, and part of the free interaction of market forces that the NLRA is designed to protect.
It is implausible that Congress intended the use of state funds to be an area “unregulated because left to be controlled by the free play of economic forces”; state funds are by definition not controlled by the free play of economic forces. Machinists, 427 U.S. at 140, 96 S.Ct. 2548 (internal quotation marks omitted); see also Boston Harbor, 507 U.S. at 225-26, 113 S.Ct. 1190. States are not “employers” under the NLRA, and the congressional goal of balancing relations between private employers and employees does not apply to state governments. 29 U.S.C. § 152(2). Addressing only the use of state funds does not address employer behavior in the private marketplace, which is what the NLRA is designed to regulate. See Archibald Cox, Labor Law Preemption Revisited, 85 Harv. L.Rev. 1337, 1352 (1972) (cited with approval in Machinists, 427 U.S. at 140 n. 4, 96 S.Ct. 2548) (“Two funda*1002mental ideas lie at the core of the national labor policy: (1) freedom of employee self-organization; and (2) the voluntary private adjustment of conflicts of interest over wages, hours, and other conditions of employment through the negotiation and administration of collective bargaining agreements.”) (emphasis added). Because the statute creates a wall between publically funded activity and private activity, it does not make sense to hold that the California statute has violated Machinists simply by restricting the use of state funds.
The majority’s critical error is its conclusion that a state’s restricting employers’ use of the state’s own funds for labor-related advocacy must require the conclusion that it is regulating the labor field. This leap of faith is not supported by precedent or logic, and can only be maintained through an unwarranted vision of the statute’s meddling effect on labor relations. As to the health and vitality of the labor relations debate itself, there is no evidence in the record that AB 1889’s core restriction will do anything to undermine it.
Here is an illustrative example. Imagine that California had a certain amount of money that it wished to grant to hospitals to create more nurse positions. However, California wanted the money to go toward creating nurse positions — not toward funding a campaign to convince the new nurses not to unionize. So long as the recipient hospitals are still free to lobby the nurses to whatever extent they please with their existing treasuries, why should California be forced to fund such lobbying with the money that the state wishes to go toward funding the positions themselves?
The majority, then, has a problem: it purports to rely on the statute’s “lever-ag[ing]of state funds” alone as requiring preemption, Maj. Op. at 980 n. 6, but cannot demonstrate how such leveraging alone can have anything but a negligible and incidental effect (at worst) on employer/employee union-related debates. The majority deals with this problem in two ways. First, it relies on the Supreme Court’s opinion in Wisconsin Dept. of Industry, Labor and Human Relations v. Gould, Inc., 475 U.S. 282, 289, 106 S.Ct. 1057, 89 L.Ed.2d 223 (1986), for the proposition that just because California “has chosen to use its spending power rather than its police power does not significantly lessen the inherent potential for conflict [with federal labor relations].” Second, it focuses on the statute’s enforcement provisions as having an improper effect on employer speech.
A. Gould, Rust and State Spending Power
In Gould, the Supreme Court found preempted a statute that used a state’s spending power to directly regulate labor relations. Id. at 291, 106 S.Ct. 1057. The statute held that a company found to have violated the NLRA three times within any five year period would be barred from doing any business with the state. In doing so, the statute made an employer’s union-related conduct (in this case, whether or not it had violated the NLRA) a condition for the receipt of state funds. If an employer violated the statute, it could not receive the state funds (or, of course, spend them) for any purpose. Id. at 283-84, 106 S.Ct. 1057.
Likewise, in Golden State, the Court found preempted a city’s decision to condition renewal of a cab franchise on settlement of a labor dispute. Golden State, 475 U.S. at 616-18, 106 S.Ct. 1395. Once again, the employer’s labor-related activity (in this case, settling a labor dispute) was made a condition for the receipt of state funds. If the employer did not act in the labor arena in the way the city demanded, *1003the employer could not receive the state funding (or franchise). In both cases, the Supreme Court concluded that such conditioning of funds constituted direct regulation of the activity at issue — and was therefore preempted.
If California had made employer neutrality a condition of the receipt of state grants or funds- — that is, if it had made employer neutrality a condition of doing business with the state — I have little doubt that such a restriction would be preempted by the NLRA. As in Gould and Golden State, a condition on the receipt of funds would encourage employers to modify the spending of their private funds in labor disputes in order to become eligible for state funding, and would therefore be a transparent effort to use the spending power to “introduce some standard of properly balanced bargaining power” and alter employers’ private spending decisions. Machinists, 427 U.S. at 149-150, 96 S.Ct. 2548 (internal quotation marks and citation omitted).
The California statute at issue here has a crucial difference. A restriction on the use of state funds is different from a restriction imposed as a condition for the receipt of state funds. In the former instance, the employer has absolute freedom to spend its own funds however it wishes, so long as it does not spend state funds on its union-related advocacy; in the latter, the employer’s use of its own funds is curtailed by the restriction. Preventing private employers from using state funds on pro- or anti-union activity does not, in and of itself, undermine the NLRA bargaining process protected by Machinists. Such a restriction affects that process only to the extent that it allows the state to maintain its own neutrality even as it funds private employers.
As such, the difference between the statutes at issue in Gould and Golden State and AB 1889 can be analogized to the circumstances of Rust v. Sullivan, 500 U.S. 173, 111 S.Ct. 1759, 114 L.Ed.2d 233 (1991), where the Supreme Court distinguished between statutes that impose conditions on the receipt of state funds from statutes that simply limit the uses to which those funds may be put. Id. at 198, 111 S.Ct. 1759. I agree with the majority that “the question before us ... is not whether AB 1889 is consistent with the First Amendment.” Maj. Op. at 994. However, Rust is about more than just the First Amendment. It also reveals the Supreme Court’s understanding of what constitutes direct regulation through a state’s spending power. The Court held in Rust that “[b]y requiring that the ... grantee engage in abortion-related activity separately from activity receiving federal funding, Congress has ... not denied it the right to engage in abortion-related activities.” Rust, 500 U.S. at 198, 111 S.Ct. 1759. With these words, the Court signaled what the majority misses today: such a restriction on the use of government funds is not direct regulation of the activity. Instead, “Congress has merely refused to fund such activities out of the public fisc.” Id.
Likewise, here, California has not “denied” employers the “right to engage in [unionj-related activity,” but “merely refused to fund such activities out of the public fisc.” Id. This conclusion operates independently of First Amendment doctrine: it simply follows from the Court’s observations about what is regulation and what isn’t. The issue in Rust was whether a statute was invalidated by the First Amendment; here, it is whether a statute is invalidated by federal labor preemption doctrine. These are different inquiries, but they share the same predicate inquiry into what activity is being regulated. In Rust, the Supreme Court offered guidance *1004on this separate inquiry, and we should not ignore its obvious relevance here.
The Supreme Court was clear in Gould that “[n]o other purpose [than regulating labor relations] could credibly be ascribed [to the statute], given the rigid and undiscriminating manner in which [it] operates: firms adjudged to have violated the NLRA three times are automatically deprived of the opportunity to compete for the State’s business.” Gould, 475 U.S. at 287-88, 106 S.Ct. 1057. By contrast, AB 1889 is both flexible and discriminating: it discriminates between employers’ use of their own funds, which is permitted, and of state funds, which is not; and it allows employers the flexibility to use their own funds for union-related advocacy. Further, a very different purpose could credibly be ascribed to it: to help the state remain neutral in labor disputes.
Rust’s analysis of what kinds of exercises of the state spending power constitute direct regulation of activity is arguably controlling here. At the very least, it illustrates how the majority has misunderstood the nature of the regulation at issue.
B. AB 1889’s Enforcement Provisions
The majority further argues that AB 1889’s enforcement provisions have effects that require preemption. As I will explain, I agree with several of the majority’s conclusions in this regard — but these conclusions provide no basis whatsoever for holding, as the majority has, that the restriction on the use of state funds is itself preempted.
My own analysis leads me to conclude that although AB 1889’s core restriction survives preemption, several of its enforcement provisions may in their effect go beyond assuring California’s neutrality in labor disputes and instead pressure employers themselves to remain neutral in labor disputes. The majority identifies these problems as well, arguing that “[b]y creating exacting compliance burdens, strict accounting requirements, the threat of lawsuits, and onerous penalties, the statute chills employer speech on the merits of unionism, and adds that” “[t]he potential cost of litigation, plus the threat of severe penalties threaten to effectively halt employer campaigns ...” Maj. Op. at 978. The majority then details what it considers to be the evils of “the compliance provisions,” the “documentation demands” and other enforcement provisions. Maj. Op. at 979.
In certain respects, I share the majority’s concerns. Some of the statute’s enforcement provisions appear to have an impermissibly intrusive effect on the NLRA’s balance of private actions between employer and employee, by exposing employers to the risk of significant litigation costs and punitive sanctions if they support or oppose unionization, even without using state funds.
The majority, however, uses these intrusive effects, which derive solely from a few specific enforcement provisions, as a proxy for finding the entire statute preempted. The majority has held that the restriction on the use of state funds must be preempted, but only marshaled arguments against a narrower subset of the statute’s provisions — those that implement and enforce the restriction.
C. Dalton and Partial Preemption
“In a pre-emption case such as this, state law is displaced only to the extent that it actually conflicts with federal law. ... The rule is that a federal court should not extend its invalidation of a statute further than necessary to dispose of the case before it.” Dalton v. Little Rock Family Planning Servs., 516 U.S. 474, 476, 116 S.Ct. 1063, 134 L.Ed.2d 115 (1996) (inter*1005nal citations and quotation marks omitted). Thus, in order to resolve this case, we have an obligation to determine which parts of the statute are preempted and which are not.
The majority attempts to avoid this obligation by concluding that regardless of how the statute is enforced, “the balance of power as between labor unions and employers would still be improperly disturbed.” Maj. Op. at 989. But how? The majority fails to establish that the restriction on use of state fund is itself preempted. Once one strips from the majority opinion all of its warnings about the effects of the enforcement provisions, there is nothing left to trigger preemption.
Instead, the majority is content to point to vague “risks” — the risk that employers will “be accused of misspending state funds” or “of violating state law.”6 Aside from the few enforcement provisions the majority and I agree upon, these generalized risks are not sufficiently concrete effects to warrant striking down a sovereign act of the state of California — as the Supreme Court has made clear in Dalton and before. See Boston Harbor, 507 U.S. at 224, 113 S.Ct. 1190 (cautioning us to be “reluctant to infer preemption”); Maryland v. Louisiana, 451 U.S. 725, 746, 101 S.Ct. 2114, 68 L.Ed.2d 576 (1981) (stating that we must begin any preemption analysis with the “basic assumption that Congress did not intend to displace state law.”).
Because I believe only certain enforcement provisions sufficiently risk intrusion into areas preempted under Machinists, I would hold that only partial preemption is warranted. On the record before us, however, we are not in a position to determine with certainty which parts of AB 1889 must be found preempted. As made clear by the majority, both in the district court and before us, the parties limited their preemption arguments to discussion of the statute as a whole. It would not be appropriate for us on this record to try to divide the statute’s enforcement provisions into preempted and nonpreempted portions. For example, certain enforcement provisions, such as the requirement that employers segregate state and nonstate funds, may be designed to do no more than permit the state to erect a “wall” between state and non-state funds and to enforce a purely compensatory remedy when state funds are misused. To conclude that such provisions are preempted could call into question a state’s power to require routine record-keeping requirements that apply in many different contexts to nonprofit corporations and other recipients of state grants.
Even on this record, however, it seems likely that at least some of the enforcement provisions — particularly those that relate to record-keeping — should survive preemption. Therefore, we should leave it to the district court to resolve such questions in the first instance on an appropriate record. We should remand to the district court so the parties can present arguments (and evidence, if necessary) about the separate statutory provisions in light of the preemption principles I have explained today. See, e.g., Arizona Libertarian Party, Inc. v. Bayless, 351 F.3d 1277, 1283 (9th Cir.2003) (“Although sever-ability is a question of state law that we review de novo, we nonetheless consider it prudent to remand to the district court where we believe the district court is better able to decide the question in the first *1006instance.”) (internal quotation marks and citations omitted).
For the reasons set forth above, I respectfully dissent.
. Amicus curiae the National Labor Relations Board ("NLRB”) — the very entity whose primary jurisdiction over labor related issues is protected by Gannon — makes no such claim as to 8(c)'s supposed affirmative grant of speech rights. See NLRB Amicus Brief at 4, 21-26 (echoing interpretation of the Act offered above, and noting Section 8(c)'s "exemption” for noncoercive speech).
. Even if one were to disagree with my reading of Section 8(c), I explain below in the Machinists analysis how AB 1889 does not directly regulate employer speech in any case — meaning that even if Section 8(c) did grant employer speech rights, AB 1889 would not be preempted under Garmon because it does not regulate such speech. See infra Section II.
. In Radcliffe, defendants argued that "the validity of plaintiffs' claims ... turns on whether the union activities carried on by the *999plaintiffs were ... protected by § 7 of the NLRA.” Id. at 785.
. There is no merit to the Chamber of Commerce's claim that the California statute provides an additional remedy for NLRA violations. The statute's damages provisions are intended to remedy the misuse of state funds under the statute, regardless of whether any NLRA violation has occurred.
. The majority's extensive arguments as to the "real" purpose of AB 1889 are beside the point. Neutrality means not taking sides. Even if the majority is right that California is effectively preventing employers from using state funds to advocate against unionization only, California is still remaining neutral— and it is simply irrelevant whether the money would otherwise be spent to support unionization, oppose it, or in some combination.
. The majority also stresses that employers would suffer the indignity of "needpng] to maintain [financial] records,'1 something that should not be entirely unfamiliar to private entities that receive money from state coffers.