concurring in part and dissenting in part:
Although I agree with the court’s determination that the allegations of sex discrimination, if true, would establish that the State of Alaska, through its Governor’s Office, violated the Constitution’s Equal Protection Clause, I do not think the same can be said for the allegation of retaliatory discharge in violation of the First Amendment. In my view, that claim does not state an actual constitutional violation. We must therefore analyze the statute under which the claim is made, the Government Employee Rights Act of 1991 (“GERA”), to determine whether it is valid prophylactic legislation under section 5 of the Fourteenth Amendment. See generally City of Boerne v. Flores, 521 U.S. 507, 117 S.Ct. 2157, 138 L.Ed.2d 624 (1997). I believe GERA fails such scrutiny. With respect, I must dissent from the court’s opinion insofar as it holds that Alaska’s sovereign immunity does not preclude the claim of retaliatory discharge.
I
Section 5 of the Fourteenth Amendment grants Congress the “power to enforce, by appropriate legislation, the provisions of [the Fourteenth Amendment].” U.S. Const, amend. XIY, § 5. As the majority correctly explains, in order for Congress to abrogate state sovereign immunity pursuant to this enforcement power, it must “unequivocally express[ ] its intent to abrogate that immunity” and “act[] pursuant to a valid grant of constitutional authority.” Kimel v. Fla. Bd. of Regents, 528 U.S. 62, 73, 120 S.Ct. 631, 145 L.Ed.2d 522 (2000). Although I believe it to be a close question, and Judge Ikuta’s conscientious dissent to the contrary notwithstanding, it seems to me that Congress did express its intent to abrogate sovereign immunity in the GERA. With respect to the second requirement, Congress acts pursuant to a valid grant of constitutional authority if it either passes so-called “prophylactic legislation” or enacts remedies for actual violations of the Constitution. For “purportedly prophylactic legislation [to] constitute[ ] appropriate remedial legislation, ... ‘there must be a congruence and proportionality between the injury to be prevented or remedied and the means adopted to that end.’ ” Kimel, 528 U.S. at 81, 120 S.Ct. *1072631 (quoting Boerne, 521 U.S. at 520, 117 S.Ct. 2157). Actual remedial laws, as the majority points out, do not have to meet this test. See United States v. Georgia, 546 U.S. 151, 158, 126 S.Ct. 877, 163 L.Ed.2d 650 (2006) (“[N]o one doubts that § 5 grants Congress the power to enforce the provisions of the Amendment by creating private remedies against the States for actual violations of those provisions.” (internal quotations marks and alteration omitted)).
A crucial threshold question, then, is whether a given claim against a state alleges conduct that would violate the Fourteenth Amendment to the Constitution. The allegations of pay discrimination state straightforward violations of the Equal Protection Clause, and I join the majority’s opinion on that issue in its entirety. With respect to the claim of sex discrimination stemming from Jones’ sexual harassment complaint, I concur in the result but remain wary of some of the majority’s reasoning. Finally, I must disagree entirely with the majority’s analysis of the purported First Amendment claim of retaliatory discharge, an analysis that wrongly enlarges the constitutional implications of employment decisions at the highest levels of state government.
A
While I agree with the majority’s conclusion regarding the so-called sexual harassment issue, I wish to clarify that I read the majority opinion to hold no more than that it would violate the Equal Protection Clause if a state deliberately refused to protect its female employees from sexual harassment. In other words, it is not the sexual harassment that Jones allegedly suffered, as such, that generates her constitutional claim. After all, she does not claim that the State of Alaska, through the official acts of its agents, sexually harassed her. Rather, the constitutional claim properly rests on Jones’s allegation that the Governor’s Office, an arm of the State, responded to her formal complaint of sexual harassment by firing her.
It is worth pausing to consider this claim carefully, for it is not the ordinary instance of unconstitutional discrimination. In most suits against a state where sexual harassment is involved, it will make more sense to characterize the claim not as one for sexual harassment “but as a claim of failure to protect against such harassment.” Bohen v. City of East Chicago, 799 F.2d, 1180, 1189 (7th Cir.1986) (Posner, J., concurring). This is because, although states can pass laws or adopt policies that treat women differently in hiring, pay, or other official emoluments of employment, they do not often pass laws or adopt policies to harass women sexually. However, “[i]f a state or city deliberately refused to provide police protection for women, it would be violating the equal protection clause” regardless of whether those who harassed them were state actors. Id. at 1190. In order for such a claim to succeed, there would have to be “a policy of nonresponse to complaints of harassment, or an authoritative decision not to respond.” Id. In view of these considerations, I understand the majority to hold that Alaska’s “authoritative decision not to respond” to Jones’ formal complaint, but to fire her instead, if that is what happened, violated the Equal Protection Clause.
The logical corollary to this holding is that sexual harassment, as such, does not ordinarily violate the Equal Protection Clause.1 This is because discrimination *1073can only violate equal protection if it is intentional and done by the state. Cf. Washington v. Davis, 426 U.S. 229, 238-45, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976) (holding that evidence of discriminatory intent is necessary to make out an equal protection violation assuming the government action is neutral on its face). The actions of the Governor’s aide in this case would not bring liability on Alaska unless the state officially sanctioned them. As the majority recognizes, the Constitution cannot support liability against the state for constitutional torts on the agency law theory of respondeat superior. Maj. Op. at 1070.
Thus, there is a crucial limitation to the majority’s statement that sexual harassment can state a violation of the Equal Protection Clause, see Maj. Op. at 1069. This possibility is limited by the bedrock constitutional principle I have discussed: sexual harassment will state a violation only where there is intentional discrimination by the state.
I dwell on this caveat in order to emphasize that, without it, we would constitutionalize the type of claim employees might bring under Title VII. Such a result would be directly contrary to Supreme Court precedent. See Davis, 426 U.S. at 239, 96 S.Ct. 2040 (“We have never held that the constitutional standard for adjudicating claims of invidious racial discrimination is identical to the standards applicable under Title VII, and we decline to do so today.”). And the cases from sister circuits on which the majority relies followed the Court’s guidance. As the majority opinion stated in Bohen, “the ultimate inquiry,” where someone alleges sexual harassment as a violation of equal protection, “is whether the sexual harassment constitutes intentional discrimination. This differs from the inquiry under Title VII as to whether or not the sexual harassment altered the condition of the victim’s employment.” Bohen, 799 F.2d at 1187.
Thus, I agree with the majority that Jones’ claim — that the Governor’s Office fired her rather than respond to her complaint of sexual harassment — states a violation of the Equal Protection Clause. But it is not the alleged sexual harassment but rather the “authoritative decision not to respond,” Bohen, 799 F.2d at 1190, that justifies such conclusion.
B
Turning now to Ward’s allegation of retaliatory discharge in violation of the First Amendment, as incorporated against the states through the Fourteenth, I note that, at oral argument, counsel for Ward admitted that his client’s actual, First Amendment claim under the relevant case law was a “tough” one to make out. No wonder. This case, it seems to me, is a prototypical example of an employee’s attempt to “constitutionalize [an] employee grievance,” a practice that the Supreme Court has explicitly discouraged. See Connick v. Myers, 461 U.S. 138, 154, 103 S.Ct. 1684, 75 L.Ed.2d 708 (1983). But it goes farther even than that, for Ward attempts to constitutionalize a political spat over her loyalty to the administration of Alaska’s Governor. With respect, the majority’s approval of Ward’s novel theory opens up a new frontier in this area of constitutional law, which, I believe, contravenes the spirit, if not the letter, of the Supreme Court’s decisions on the subject.
*10741
In general, “[w]hen a citizen enters government service, the citizen by necessity must accept certain limitations on his or her freedom.” Garcetti v. Ceballos, 547 U.S. 410, 418, 126 S.Ct. 1951, 164 L.Ed.2d 689 (2006). This is because the “government as employer indeed has far broader powers than does the government as sovereign.” Id. (quoting Waters v. Churchill, 511 U.S. 661, 671, 114 S.Ct. 1878, 128 L.Ed.2d 686 (1994) (plurality opinion)). The reason the government would violate the First Amendment at all by firing one of its employees on account of what he or she said is that employees, as citizens, “retain the prospect of constitutional protection for their contributions to the civic discourse.” Id. at 422, 126 S.Ct. 1951. It is thus necessarily within that context that we apply the doctrinal test for whether a public employee has alleged a First Amendment violation for retaliatory discharge.
Such test has two parts. First, unless “the employee spoke as a citizen on a matter of public concern[,] .... the employee has no First Amendment cause of action based on his or her employer’s reaction to the speech.” Id. at 418, 126 S.Ct. 1951. Only if the employee passes this threshold does “the possibility of a First Amendment claim arise[].” Id. (emphasis added). A court must then evaluate that possibility under the balancing test of Pickering v. Board of Education of Township High School District 205. See 391 U.S. 563, 568, 88 S.Ct. 1731, 20 L.Ed.2d 811 (1968); Garcetti 547 U.S. at 418, 126 S.Ct. 1951.
At the threshold stage, as Garcetti illustrated, First Amendment protection attaches only to speech analogous to that which an ordinary citizen would make as part of public discourse. 547 U.S. at 423, 126 S.Ct. 1951 (“Employees who make public statements outside the course of performing their official duties retain some possibility of First Amendment protection because that is the kind of activity engaged in by citizens who do not work for the government.”). The First Amendment does not give more license to government employees than ordinary citizens. This is the meaning of the Supreme Court’s admonition that the First Amendment neither “invest[s] [public employees] with a right to perform their jobs however they see fit,” nor “empower[s] them to constitutionalize the employee grievance.” Id. at 420, 422, 126 S.Ct. 1951 (internal quotation marks omitted).2
2
As I read the precedents, Ward has failed to state a claim under the First Amendment for retaliatory discharge.
To state a First Amendment claim, the employee must show that he or she spoke not as an employee, but as a private citizen in public discourse. See Garcetti 547 U.S. at 418-25, 126 S.Ct. 1951. The majority’s analysis on this point addresses the requirements of “speech as a citizen” and “matter of public concern” in rather narrow terms. To be sure, they are two *1075distinct requirements for constitutional protection. Ceballos v. Garcetti, 361 F.3d 1168, 1186-87 (9th Cir.2004) (O’Scannlain, J., specially concurring) (insisting on both the “speech as a citizen” and “matter of public concern” prongs of the threshold inquiry), overruled by Garcetti, 547 U.S. at 426, 126 S.Ct. 1951. But the idea behind the caselaw is to ensure that public employees are still able to participate in public debate, not to provide them job security while they pursue their own ends. See, e.g., Pickering, 391 U.S. at 573, 88 S.Ct. 1731 (rejecting the school’s attempt to “limitf ] teachers’ opportunities to contribute to public debate”); see also Garcetti, 547 U.S. at 419, 126 S.Ct. 1951 (“The Court has acknowledged the importance of promoting the public’s interest in receiving the well-informed views of government employees engaging in civic discussion.”). We must keep our eye on the ball here, for the case before us throws something of a curve.
The typical situation requires a court to determine whether speech was primarily an internal office matter or a contribution to the public debate. See, e.g., Connick, 461 U.S. at 140, 103 S.Ct. 1684 (considering “whether the First and Fourteenth Amendments prevent the discharge of a state employee for circulating a questionnaire concerning internal office affairs”). But this case involves policymaking staff in the office of the chief executive of the State of Alaska. Thus, the internal office politics are also the politics of the state. In this context, it contravenes the spirit of Garcetti and its predecessors to hold that, even though Ward criticized the Governor on a subject of public interest the Governor cannot constitutionally fire her for disloyalty.
The majority spends time illustrating that Ward’s speech was not part of her official duties, which is surely correct. Garcetti does not squarely dictate the result in this case for that reason. Maj. Op. at---; Garcetti, 547 U.S. at 421, 126 S.Ct. 1951 (“We hold that when public employees make statements pursuant to their official duties, the employees are not speaking as citizens for First Amendment purposes.”). But the controlling ratio decidendi of Garcetti casts a longer shadow. The importance of the official nature of the speech in Garcetti lay in the distinction between speech that ordinary citizens make and speech that only occurs because of employment with the government. “Restricting speech that owes its existence to a public employee’s professional responsibilities,” the Court insisted, “does not infringe any liberties the employee might have enjoyed as a private citizen.” Id. at 421-22. Garcetti explicitly contrasted such speech with “the expressions made by the speaker in Pickering, whose letter to the newspaper had no official significance and bore similarities to letters submitted by numerous citizens every day.” Id. at 422, 126 S.Ct. 1951.
Thus, although Ward’s press conference was not strictly part of her official duties, her importance in the Governor’s administration necessarily means that more of her conduct came within the legitimate purview of her employer. The Supreme Court has emphasized the need to “afford!] governmental employers sufficient discretion to manage their operations.” Id. at 422, 126 S.Ct. 1951. Nowhere is such discretion more important than at the highest levels of state government. At those levels, loyalty to the administration on matters of public concern is in a sense the price of employment. Ward chose to go public with one side of an internal struggle among the Governor’s policy aides. The Governor’s Office considered this to be disloyal and fired her. Especially in the context of a governor’s office, this is a classic employment decision of the *1076kind the Supreme Court has warned should not “bec[o]me a constitutional matter.” Connick, 461 U.S. at 143, 103 S.Ct. 1684. We must remember that “a federal court is not the appropriate forum in which to review the wisdom of [such] a personnel decision.” Id. at 147, 103 S.Ct. 1684.
It helps to consider analogies. Take the example of an aide to a governor who criticizes publically the governor’s tax policy in a press conference. Such speech would be an undoubted contribution to the public debate, but would it violate the First Amendment if the governor fired the aide for disloyalty? I think not, and I imagine the majority would agree. And if the aide criticized not tax policy but the governor’s policy regarding internal complaints of sexual harassment? The result is the same, even though the subject of the criticism is a potentially illegal practice (ignoring sexual assault on female employees).3
Such a result would only seem harsh from the myopic perspective of the conviction that the Constitution must provide remedies for all harms. We can, and should, take allegations like those Ward made very seriously without invoking the First Amendment. “As the [Supreme] Court noted in Connick, public employers should, ‘as a matter of good judgment,’ be ‘receptive to constructive criticism offered by their employees.’ ” Garcetti 547 U.S. at 425, 126 S.Ct. 1951 (quoting Connick, 461 U.S. at 149, 103 S.Ct.1684). Indeed, “[t]he dictates of sound judgment are reinforced by the powerful network of legislative enactments — such as whistle-blower protection laws and labor codes — available to those who seek to expose wrongdoing.”4 Id. In the appropriate circumstances, we must rely on such customary and legislative protections if we are to avoid “constitutionalizing] the employee grievance.” Connick, 461 U.S. at 154, 103 S.Ct. 1684.
II
My conclusion that the allegation of retaliatory discharge does not state an actual violation of the Constitution compels me to address, insofar as the claim is remediable under the GERA, whether that statute constitutes valid “congruent and proportional” legislation under the Supreme Court’s Boeme test.
A
Congress’ power to enforce the Fourteenth Amendment under section 5 does not allow it “to decree the substance of the Fourteenth Amendment’s restrictions on the States. Legislation which alters the meaning of the [Fourteenth Amendment] cannot be said to be enforcing [it].” Boerne, 521 U.S. at 519, 117 S.Ct. 2157. Boerne requires that for “purportedly prophylactic legislation [to] constitute[] appropriate remedial legislation, ... ‘there must be a congruence and proportionality between the injury to be prevented or remedied and the means adopted to that end.’ ” Kimel, 528 U.S. at 81, 120 S.Ct. 631 (quoting Boene, 521 U.S. at 520, 117 S.Ct. 2157).
Because prophylactic legislation prohibits or regulates constitutional conduct that supposedly leads to unconstitutional eon-*1077duct, Congress must explain its belief that regulating the former will help to prevent the latter. See Boerne, 521 U.S. at 519-20, 117 S.Ct. 2157. Such requirement responds to the Supreme Court’s concern in Boeme that Congress not “decree the substance of the Fourteenth Amendment’s restrictions” under the guise of enforcing them. Id. at 519, 117 S.Ct. 2157. The Supreme Court has outlined a three-step test for determining congruence and proportionality. “The first step ... is to identify with some precision the scope of the constitutional right at issue.” Bd. of Trustees of the Univ. of Ala. v. Garrett, 531 U.S. 356, 365, 121 S.Ct. 955, 148 L.Ed.2d 866 (2001). Next, Congress must have identified a history and “pattern of constitutional violations” by the states. Fla. Prepaid Postsecondary Educ. Expense Bd. v. Coll. Sav. Bank, 527 U.S. 627, 639-640, 119 S.Ct. 2199, 144 L.Ed.2d 575 (1999); see also Garrett, 531 U.S. at 368, 121 S.Ct. 955. Finally, legislation must be in fact “congruent and proportional,” in light of Congress’ factual findings, “to the targeted violation.” Garnett, 531 U.S. at 374, 121 S.Ct. 955. It seems to me that the crucial step in this case is the second one — the requirement that Congress identify a pattern of constitutional violations. One way courts pursue this inquiry is “by examining the legislative record containing the reasons for Congress’ action.” Kimel, 528 U.S. at 88,120 S.Ct. 631.
The parties do not dispute that, when Congress enacted the GERA in 1991, it made no findings regarding discrimination against state employees at the policy-making level. When it passed the Equal Employment Opportunity Act in 1972, however, Congress did make extensive factual findings. H.R.Rep. No. 92-238, at 19 (1971), reprinted in 1972 U.S.C.C.A.N. 2137, 2152 (noting the existence of “widespread discrimination against minorities ... in State and local government employment, and that the existence of this discrimination is perpetuated by the presence of both institutional and overt discriminatory practices”). The EEOC would like us to consider the latter findings in evaluating the former law for purposes of the Boeme test.
The EEOC’s theory is that, with the GERA, Congress merely finished the job it started in 1972 when it amended Title VII to cover the States as employers. It cites extensive language from Supreme Court opinions and congressional records to show that gender discrimination persisted in state government as of 1991 the way it had existed in 1972.5 But all of the passages the EEOC quotes speak of gender discrimination in general, not at the policymaking levels of state government to which the GERA applies. The EEOC argues that it need not present such particularized congressional findings. It cites for support Justice Powell’s concurrence in Fullilove v. Klutznick, 448 U.S. 448, 100 S.Ct. 2758, 65 L.Ed.2d 902 (1980). But the question is what the more recent Boeme requires. A concurrence published seventeen years before Boeme is not probative of that question.
Furthermore, even if one could consider the 1972 findings, they do not pertain to the policymaking staff covered by the GERA. In 1972, Congress did find widespread discrimination in state and local *1078government and it acted to prevent it, but it specifically excluded personal and policymaking staff. That is part of the background against which Congress legislated when it enacted the GERA. If one wants to impute congressional intent to that exclusion, the only responsible imputation is that Congress did not believe a remedy was necessary with respect to policy-making employees. That is to say, if there is any reason to believe what one reads in committee reports, the official position of the relevant House Committee belies the EEOC’s position. In a section of the committee report entitled “Need for the Bill,” it declared that the “time ha[d] come to bring an end to job discrimination once and for all,” and that “[i]t is essential that ... effective enforcement procedures be provided the [EEOC] to strengthen its efforts to reduce discrimination in employment.” H.R.Rep. No. 92-238, at 2139-41. It would follow logically from this language that whatever levels of state and local employment Congress exempted from the 1972 Act’s reach did not suffer from the job discrimination that so concerned the House Committee.
The background against which Congress enacted the GERA, therefore, does not illustrate that Congress had already found a pattern of unconstitutional discrimination at the policymaking level of state and local employment. Instead it shows that Congress had excluded employees at that level from protection. Because Congress explicitly excluded policymaking employees from Title VII’s reach in 1972, I do not believe this court would be justified in using the findings Congress made in doing so to support its decision in 1991 to repeal that very exclusion.
Without the 1972 findings, the EEOC can point to no evidence that Congress identified, as the Supreme Court has required it to do, a history and pattern of violations of the constitutional rights of the states against high-level personal and policy-making employees. This compels me to conclude that the GERA is not “congruent and proportional” legislation within the meaning of Boeme. It therefore cannot constitute a valid abrogation of state sovereign immunity.
Neither the EEOC nor the federal courts are empowered to entertain the non-constitutional claim against the State of Alaska, which, as I have explained, is precisely Ward’s retaliation claim. I must respectfully dissent from the majority’s conclusion to the contrary.
. I hasten to add that the conduct Jones complained of is outrageous and unsavory; I do not mean to condone it in any way. At the same time, we should acknowledge that the *1073State of Alaska has not conceded the truth of the allegations either of Jones or of Ward. Indeed, there seems to be vigorous dispute about the facts underlying their dismissal from the Governor’s Office.
. We have also recently clarified that, at the second, balancing stage, "the plaintiff bears the burden of showing the state took adverse employment action and that the speech was a substantial or motivating factor in the adverse action.” Eng v. Cooley, 552 F.3d 1062, 1071 (9th Cir.2009) (internal quotation marks and alterations omitted). If the plaintiff carries that burden, then the government must show that, under Pickering, its "legitimate administrative interests outweigh the employee's First Amendment rights” or that it would have made the same decision without the employee's protected speech. Id. at 1071-72 (internal quotation marks omitted). In my view, for the plaintiff to state a First Amendment claim, he or she must meet at least the initial burden under Eng.
. The majority conflates the supposed policy of ignoring illegality with the illegality itself, quoting back to me my observation that states sometimes adopt policies to treat women differently but not usually to harass them sexually. Maj. Op. 1070 n. 7. This only confuses the issue. My point is that the scope of an employee grievance, as opposed to a contribution as a citizen to public debate, necessarily widens the higher one climbs up the ladder of government.
. Alaska, in fact, has such a whistle-blower protection law. Alaska Stat. 39.90.100-.150.
. There is no attempt to show congressional concern for the violation of First Amendment rights per se. The EEOC's and Intervenor Ward's argument seems to be that preventing retaliatory discharges against state employees for complaining about sexual harassment is part of Congress' prophylactic remedy for unconstitutional gender discrimination. It therefore stands or falls with the legitimacy of prophylactic remedies for employment discrimination.