Dissenting.
The district judge awarded fees because he found that “the Navy’s policy of discharging persons, based solely on their status as homosexuals, for the purported purpose of protecting the military mission, lacks any reasonable basis in law or fact.” Appendix at 10804.1 By affirming, the majority causes a conflict with four other circuits that have held the government may discharge members of the armed forces for saying they are homosexuals. See Steffan v. Perry, 41 F.3d 677, 686 (D.C.Cir.1994) (en banc); Ben-Shalom v. Marsh, 881 F.2d 454, 464 (7th Cir.1989); Woodward v. United States, 871 F.2d 1068, 1076 (Fed.Cir.1989); Rich v. Secretary of the Army, 735 F.2d 1220, 1229 (10th Cir.1984). Worse, the majority creates a conflict with the law of the case and the law of our circuit, as announced in Meinhold II, where we held that this is a close and difficult question which should not be decided in this case. Meinhold II, 34 F.3d 1469, 1479 (9th Cir.1994).
The majority does not acknowledge these conflicts, donning the mantle of deference to the district court. Maj. op. at 1278-80. But we review questions of law de novo. United States v. Gutierrez, 116 F.3d 412, 415 (9th Cir.1997). Where the district court premises its discretion on an error of law, we cannot affirm without adopting the district court’s error. See United States v. Willett, 90 F.3d 404, 406 (9th Cir.1996). Doubtless, this is why the majority tries so hard to explain away the wall of authority upholding the government’s right to discharge members of the armed forces who say they are homosexuals. Maj. op. at 1278-79. While my colleagues fail in this effort,2 it is beside the point. We held in Meinhold II that this is a close question-so close we chose to give the regulation a narrowing construction rather than confront it. Meinhold II, 34 F.3d at 1479. The district court was not free to ignore this holding and decide for itself that the issue was settled and the Navy acted unreasonably in even arguing about it.3
*1282The majority also seems to say that the fee award is justified because the Navy misinterpreted DOD Directive 1332.14, 32 C.F.R. Pt. 41, App. A (1981). Maj. op. at 1278-79. This rationale cannot support the award because it was not relied on by the district court. See United States v. Todd, 909 F.2d 395, 398-99 (9th Cir.1990). In any event, it stretches credulity to argue that the Navy was unreasonable in construing a regulation which authorizes the discharge of a serviceman who “has stated that he ... is a homosexual”, 32 C.F.R. Pt. 41, App. A(1)(H)(1)(c)(2), as authorizing the discharge of a serviceman who has stated that he is a homosexual. See Steffan, 41 F.3d at 694. In Meinhold II, we adopted a cramped interpretation of the regulation to avoid what we saw as a tough constitutional question, not to correct a misreading by the Navy. Meinhold II, 34 F.3d at 1476 (“Thus, it is incumbent on us to see whether there is some way the regulation can be construed to resolve the matter on a nonconstitutional ground.”).4 The interpretation we adopted in Meinhold II was so original, it was not suggested even by Meinhold. How can the Navy possibly be unreasonable in giving its regulation a literal construction, one that no other court of appeals has rejected?5
The majority creates yet another conflict, this time with Bay Area Peace Navy v. United States, 914 F.2d 1224, 1231 (9th Cir.1990), which reversed a fee award because “ ‘difficult questions’ were raised and there is an absence of adverse precedent on point.” While the majority nitpicks the authorities on which the government relies, maj. op. at 10789-91, my colleagues do not come to grips with the fact that the government has never lost on this issue in any other circuit or the Supreme Court. On the other side of the ledger, the government had won a contrary ruling squarely on point, maj. op. at 1278-79, citing Ben-Shalom, and a handful of other victories which-even if slightly distinguishable (which they are not)-weigh in on the side of the government, not Meinhold’s. Under the EAJA, the issue is not whether every case the government cites is on all fours with its position, but whether the weight of authority is so clearly against it that the government should have known better than to continue pressing the point. With not a single case before Meinhold II going the other way, there was indeed “an absence of adverse precedent on point.” Bay Area Peace Navy requires us to reverse.
While it comes to us in the mundane garb of an attorney’s fee application, the question presented is important: Is the government no longer entitled to argue in the courts of this circuit that it may remove self-proclaimed homosexuals from the armed services? The district court said quite clearly and explicitly-even vehemently-that the government is not entitled to litigate the question because the government’s position is utterly unreasonable: “The Navy placed Meinhold in a position of having to litigate this action to prove what the Navy already knew-that his homosexual status was irrelevant to his performance of the military mission. It is unconscionable for any branch of government to continue to enforce a regulation or policy, knowing that the regulation lacks any merit or rational basis for its exis*1283tence, and thereby, force citizens to litigate to enforce their constitutional rights.” Appendix at 10805. This does not reflect the law of the case or the law of the circuit; it does not reflect the law of anywhere at all. We may disagree with the law, but we are bound to follow it. See Watkins v. United States Army, 837 F.2d 1428, 1452 (9th Cir.1988) (Reinhardt, J., dissenting) (“I am bound, however, as a circuit judge to apply the [law] as it has been interpreted by the Supreme Court and our own circuit, whether or not I agree with those interpretations.”). The majority does not follow the law. I must respectfully dissent.
APPENDIX
Meinhold v. Department of Defense, No. CY 92-6044 TJH (JRx)
Memorandum Opinion, May 14,1996.
[Facts and Prevailing Party Determination Omitted]
Substantial Justification
In evaluating the Navy’s position that it was substantially justified in discharging Meinhold, the Court considered the Navy’s reasonableness. See Kali v. Bowen, 854 F.2d 329 (9th Cir.1988). A position is substantially justified if it has “a reasonable basis in both in law and fact.” Pierce v. Underwood, 487 U.S. 552, 565, 108 S.Ct. 2541, 2550, 101 L.Ed.2d 490, 504-505 (1988). Moreover, the Navy had the burden of demonstrating substantial justification. See Flores v. Shalala, 49 F.3d 562, 569 (9th Cir.1995). Since the Navy did not demonstrate that its position was substantially justified, the Court’s statement of supporting reasons need not be elaborate. United States v. 313.34 Acres of Land, 897 F.2d 1473 (9th Cir.1989). Nonetheless, the Court will elaborate.
The inquiry into the existence of substantial justification focuses on two questions: “[F]irst whether the [Navy] was substantially justified in taking its original action; and, second, whether the [Navy] was substantially justified in defending the validity of the action in court.” Kali, 854 F.2d at 332.
The Navy relied primarily on four cases decided in other circuits to support its position that it was substantially justified in discharging Meinhold. However, in doing so, the Navy disregarded the Ninth Circuit’s analysis in Pruitt v. Cheney, 963 F.2d 1160 (9th Cir.1991), a case more analogous to the facts here, where the Army discharged a homosexual on the basis of status. In Pruitt, the Circuit found that the Army officer had stated an equal protection claim, and held that there was no rational basis, established on the record, for the discrimination.
The Navy contends that it was substantially justified in disregarding Pruitt because the Ninth Circuit effectively overruled Pruitt in United States v. Harding, 971 F.2d 410 (9th Cir.1992). However, Harding was a criminal case involving the statutory classification of crack versus powder cocaine and the constitutionality of the sentencing guidelines as it related to those substances. Harding argued that the legislature, in establishing the statutory classification, needed to supply empirical evidence to support the rational relationship between the two substances and the disparate sentences for their possession. The Circuit held that the “burden falls on the party attempting to disprove the existences of a rational relationship between a statutory classification and a government objective.” Harding, 971 F.2d at 413. Here, “the military’s policy on homosexuality is not the act of the legislature but a set of administrative directives implemented by administrative rules.” (Defendant’s Memorandum of Points and Authorities in Opposition to Plaintiffs Motion for Summary Judgment, p. 12, In. 26—p.13, In. 1). Moreover, Harding did not involve the substantial justification standard imposed by the EAJA.
Although Harding is not applicable here, Meinhold, at his discharge hearing, met the burden imposed by Harding. Meinhold provided evidence, including testimony from Naval officers, that he was an exemplary and outstanding sailor, that “his professionalism and leadership abilities, both in the air and on the ground are unparalleled, exceeding the routine performance of a First Class Petty Officer” (Navy Administrative Discharge Board Hearing Record (“HR”) at 416), and that he was an “example for other [sailors] to emulate” (HR at 423). Additionally, Meinhold presented evidence that Naval officers “did not concern themselves *1284with the sexual orientation of their subordinates, and that high achievers who were known to be gay, usually were protected by their commanding officers.” (HR at 468). Furthermore, as long ago as 1957, reports commissioned by the Secretary of the Navy concluded that no factual data exists to support the Navy’s position that homosexuals cannot acceptably serve in the military.
Additionally, the Navy’s principal counsel at the hearing for reconsideration of the preliminary injunction, held on November 16, 1992, admitted that there was no scientific or sociological evidence that supports the Navy’s position that homosexuals impair the military objective. (Transcript of reconsideration healing, p. 22, In. 11 — 16). Therefore, the testimony of Naval officers, Meinhold’s evaluation reports, and the Navy’s admission that it lacked evidence to support its position is proof that, not only did Meinhold meet the burden imposed by Harding, but also that the Navy was not substantially justified in discharging Meinhold based solely on his status. Hence, the Navy’s policy bears no rational relationship to the Navy’s stated objectives of protecting the accomplishment of the military mission. Finally, the Navy’s contention that Harding overruled Pruitt, which was decided only one year prior, reads much more into Harding than the Ninth Circuit intended.
The Navy was not substantially justified in initiating the discharge proceeding against Meinhold merely because there were four other cases, decided in the Navy’s favor in other circuits, involving facts similar to the facts here. The Navy relies on Pierce, 487 U.S. 552, 108 S.Ct. 2541, 101 L.Ed.2d 490 (1988), for its contention that its string of successes in other circuits gave it substantial justification for bringing this action. The Navy asks this Court to rely on a test of objective indicia, which the Supreme Court rejected in Pierce:
Obviously, the fact that one other court agreed or disagreed with the Government does not establish whether its position was substantially justified. Conceivably, the Government could take a position that is not substantially justified, yet win; even more likely, it could take a position that is substantially justified, yet lose. Nevertheless, a string of losses can be indicative; and even more so a string of successes. Once again, however, we cannot say that this category of objective indicia is enough to decide the present case.
Pierce, 487 U.S. at 569, 108 S.Ct. at 2551, 101 L.Ed.2d at 507.
Additionally, whether the Ninth Circuit has addressed the issue is a component of an inquiry into substantial justification. Kali, 854 F.2d at 332 fn. 2. Prior to this action, the Ninth Circuit had decided Pruitt on equal protection grounds, yet, knowing this, the Navy proceeded against Meinhold in this circuit. Thus, the Navy’s contention that this Court should look to a “string of successes” in other circuits lacks merit.
The Navy relied on Dronenburg v. Zech, 741 F.2d 1388 (D.C.Cir.1984), Rich v. Secretary of the Army, 735 F.2d 1220 (10th Cir.1984), Woodward v. United States, 871 F.2d 1068 (Fed.Cir.1989), and Ben-Shalom v. Marsh, 881 F.2d 454 (7th Cir.1989), in support of its argument that it had a “string of successes” which gave it substantial justification to discharge Meinhold. Dronenburg and Rich are distinguishable in that both cases were based on homosexual conduct, not status. As such, it was not reasonable for the Navy to rely on them as evidence of a “string of successes” in this matter, which involved status alone.
The Navy’s reliance on Woodward and Ben-Shalom is, also, misplaced. Both cases relied upon Bowers v. Hardwick, 478 U.S. 186, 106 S.Ct. 2841, 92 L.Ed.2d 140 (1986), for their reasoning. The issue in Bowers was the constitutionality of a Georgia statue that criminalized consensual sodomy. The Supreme Court, in upholding the Georgia statute, held that the Constitution does not confer a fundamental right upon homosexuals to engage in sodomy. Bowers, 478 U.S. at 190, 106 S.Ct. at 2843, 92 L.Ed.2d at 145. Bowers never addressed the issue of whether a person’s status as a homosexual, as opposed to conduct, was constitutionally protected.
In Woodward, the Federal Circuit apparently misapplied the reasoning in Bowers to equate homosexual status to conduct, and concluded that one’s status as a homosexual implies that one engages in homosexual con-*1285duet. More importantly, the court in Woodward, in reviewing Woodward’s right to privacy claim, stated that homosexual conduct is not one of the fundamental rights subject to heightened review under the due process clause. Meinhold’s claim was based on the equal protection clauses' not the due process clause, and furthermore, Meinhold was not accused of homosexual conduct. Therefore, the Navy’s reliance on Woodward for substantial justification is misplaced.
Additionally, in Ben-Shalom, the Army refused to reenlist a reserve sergeant who was a lesbian. The refusal was based on the sergeant’s proclamation that she was a homosexual. Though the Seventh Circuit acknowledged that there was no evidence of homosexual conduct, and the trial judge found that Ben-Shalom’s status did not mean that she was likely to commit homosexual acts, the Circuit concluded that one can reasonably infer, from status alone, that Ben-Shalom had engaged, or probably will engage, in homosexual conduct. Here to, it appears that the court misapplied Bowers and assumed that status as a homosexual equates to constitutionally unprotected homosexual conduct.
Furthermore, the Seventh Circuit, in Ben-Shalom, recognized that the Ninth Circuit might not agree with its decision. Citing Judge Norris’s concurring opinion, which was joined by Judge Canby, in Watkins v. United States Army, 875 F.2d 699 (9th Cir.1989), the Seventh Circuit recognized that an equal protection argument could be made, but it refused to apply it to the military. “On this controversial subject [whether homosexuals have the protection of the equal protection clause] unanimity is not to be expected as respectable arguments may be made on both sides of the issue.” Ben-Shalom, 881 F.2d at 465. Therefore, it was not reasonable for the Navy to rely on Ben-Shalom in support of its argument for substantial justification, when the Seventh Circuit recognized that there was a difference of opinion with the Ninth Circuit on this issue.
In Cammermeyer v. Aspin, 850 F.Supp. 910, 918 (W.D.Wash.1994), a case where the Army cited Ben-Shalom in support of its argument that homosexual orientation can rationally and reasonably be viewed as reliable evidence of a desire and propensity to engage in homosexual conduct, the court found that the government’s reliance on Ben-Shalom was misplaced, because the case incorrectly relied upon Bowers.
This Court held that Meinhold’s statement, “I am in fact gay,” was not equivalent to an admission of conduct, nor did it demonstrate a propensity to commit homosexual acts, and an assumption that it does would lack any basis in law or in fact. Thus, the Navy lacked substantial justification to discharge Meinhold and to oppose his attempts to be reinstated.
Bad Faith
“[F]ederal courts, in the exercise of their equitable powers, may award attorney’s fees when the interests of justice so require. Thus, it is unquestioned that a federal court may award counsel fees to a successful party when his opponent has acted ‘in bad faith, vexatiously, wantonly, or for oppressive reasons.’ ” Hall v. Cole, 412 U.S. 1, 5, 93 S.Ct. 1943, 1946, 36 L.Ed.2d 702, 707 (1972). Here, the Navy’s policy of discharging persons, based solely on their status as homosexuals, for the purported purpose of protecting the military mission, lacks any reasonable basis in law or fact. Thus, the interest of justice requires that Meinhold be awarded attorney’s fees for bringing this action against an oppressive act of the Navy.
The Navy was aware that it lacked any basis for its position that one’s sexual orientation effected one’s ability to function in the military. As previously indicated, in 1957, the Secretary of the Navy commissioned a report concerning the revision of its policies, procedures and directives dealing with homosexuals. The report concluded that there was no “visible supporting data to support the conclusion that gays and lesbians cannot acceptably serve in the military.” Report of the Board Appointed to Prepare and Submit Recommendations to the Secretary of the Navy for the Revision of Policies, Procedures and Directives Dealing with Homosexuality (Mar. 15,1957).
In 1976, the Chief of Naval Personnel stated that “no empirical proof exists at this time *1286[to support the Navy’s contention that] homosexuality has an adverse effect upon the completion of the [military] mission.” Memorandum from Chief of Naval Personnel to Judge Advocate General (Aug. 2, 1976).
A 1988 Department of Defense (“DOD”) commissioned study concluded that “having a same-gender or an opposite gender orientation is unrelated to job performance in the same way as being left- or right-handed.” Theodore R. Sarbin & Kenneth K. Eoyang, Nonconforming Sexual Orientation and Military Suitability, p. 33 (1988).
In 1989, another DOD commissioned study found that “homosexuals more closely resemble those who successfully adjust to military life than those who are discharged for unsuitability ... [and that] homosexuals show pre-service suitability-related adjustment that is as good or better than the average heterosexual.” Michael A. McDaniel, Preservice Adjustment of Homosexual and Heterosexual Military Accessions, p. 19 (1989).
In 1990, Vice Admiral Joseph S. Donnell stated that “[experience has ... shown that the stereotypical female homosexual in the Navy is hardworking, career-oriented, willing to put in long hours on the job and among the command’s top professionals.” Administrative Message from Commander, Naval Surface Fleet, Atlantic, to the Naval Surface Fleet, Atlantic (July 2,1990).
Finally, Dr. Lawrence J. Korb, the former Assistant Secretary of Defense who was responsible for approving and implementing a policy of banning gays and lesbians, has declared “that there is no longer any justification for the armed services’ ... ban on homosexuals serving in the military ... [that] each of the justifications offered in support of this policy is without factual foundation ... [and, therefore,] there is no longer any rational basis [for it].” Meinhold’s exhibit 12 p. 874, ¶¶ 4 & 7.
Therefore, this Court found that the Navy’s discharge of Meinhold was done in bad faith and to oppress Meinhold. The Navy placed Meinhold in a position of having to litigate this action to prove what the Navy already knew — that his homosexual status was irrelevant to his performance of the military mission. It is unconscionable for any branch of government to continue to enforce a regulation or policy, knowing that the regulation lacks any merit or rational basis for its existence, and thereby, force citizens to litigate to enforce their constitutional rights.
Furthermore, the behavior of the Navy in this lawsuit evidences bad faith. For example, the Navy’s reluctance to obey this Court’s order to reinstate Meinhold forced this Court to hold a contempt hearing. Additionally, after the Navy filed an appeal to the Ninth Circuit, it filed a simultaneous appeal to the Federal Circuit. One wonders what strategy the Navy was pursuing at that time, and what the result might have been, had the Federal Circuit not wisely stayed its involvement pending the Ninth Circuit’s decision in this matter.
The Court recognizes that, traditionally, deference is given to the professional judgment of military authorities. Goldman v. Weinberger, 475 U.S. 503, 106 S.Ct. 1310, 89 L.Ed.2d 478 (1986). However, “one does not surrender his or her constitutional rights upon entering the military.” Better v. Middendorf, 632 F.2d 788, 810 (9th Cir.1980). The Navy’s policy of discharging homosexuals based solely on status is similar to the historic racial discrimination and segregation that existed in the Armed Forces. The only basis for the policy is that some individual military personnel do not want to be around homosexuals. Mere negative attitudes, or fear, are constitutionally impermissible bases for discriminatory governmental policies. Cleburne v. Cleburne Living Center, Inc., 473 U.S. 432, 105 S.Ct. 3249, 87 L.Ed.2d 313 (1985).
Furthermore, military regulations must follow basic constitutional imperatives. It is a basic tenet of our legal system that a government agency is not at liberty to ignore its own laws. See Vitarelli v. Seaton, 359 U.S. 535, 79 S.Ct. 968, 3 L.Ed.2d 1012 (1959). The legitimacy of America’s constitutional democracy derives, in part, from a fundamental rule of law and reason, that is, the government must obey its own laws and those laws must have some rational basis for their existence. The enforcement of rules and regulations that lack a rational basis, by any governmental agency, which *1287force citizens to bring the government into court to enforce their constitutional rights, can only be viewed as an attempt by the government to coerce citizens into obeying illegal regulations or face enormous expense to vindicate their constitutional rights. Such behavior cannot be condoned, and the promulgation and enforcement of such rules can only be viewed as having been done in bad faith.
[Attorney’s Fees Calculation Omitted]
. So as to avoid any misunderstanding, I reproduce the analysis portion of the district court’s order in an Appendix.
. The majority ignores Woodward and Rich because they did not “explicitly h[o]ld that it was constitutionally permissible to equate a statement of homosexuality with conduct,” maj. op. at 10789, as had the Seventh Circuit in Ben Shalom. This is word play. Woodward and Rich both involved statements, not conduct; while Ben-Shalom was more explicit, Woodward and Rich were clear victories for the government. The majority tries to distinguish Richenberg v. Perry, 97 F.3d 256 (8th Cir.1996), Able v. United States, 88 F.3d 1280 (2d Cir.1996), and Thomasson v. Perry, 80 F.3d 915 (4th Cir.) (en banc), cert. denied, — U.S. —, 117 S.Ct. 358, 136 L.Ed.2d 250 (1996), as cases involving 10 U.S.C. § 654(b)(2)-the so-called Don't Ask, Don’t Tell policj'. I don’t see why this matters. These cases all stand for the proposition that the government may constitutionally discharge members of the armed forces for saying they are gay, just as the government claimed in Meinhold II. Finally, the majority tries to distinguish Steffan because it was decided on standing grounds. This overlooks the lengthy passage in Steffan where the D.C. Circuit criticizes the reasoning of Meinhold II. 41 F.3d at 694 & n. 18. In determining whether the government was reasonable in defending Meinhold II, surely it is relevant what other circuits have said about our ruling there.
. The matter is made worse by the fact that in 1994 Congress adopted Don’t Ask, Don’t Tell, 10 U.S.C. § 654 (Supp.1994), which authorizes the discharge of servicemembers who say they are homosexuals. This policy raises the very legal issue decided by the district court in the fee application: whether it's OK to discharge a member of the service for saying he is gay. DADT lets the government infer homosexual conduct from an admission of homosexual status, precisely as the .Navy did in Meinhold’s case. The district court held that "the Navy’s discharge of Meinhold was done in bad faith and to oppress Meinhold,” based on its conclusion that homosexual status has no bearing on fitness for service in the armed forces. Appendix at 10804-5.
Will the opinion today be cited as striking down DADT insofar as it permits discharge based on status alone? I would not be surprised, as this is an area of the law where over-citing is endemic. For example, our ruling in Pruitt v. Cheney, 963 F.2d 1160, 1164 (9th Cir.1991), which held only that plaintiff’s equal protection claim would survive a 12(b)(6) challenge, was cited by Meinhold and the district court as holding that the Navy’s policy violates equal protection. Appellee's brief at 2, 22; Appendix at 10799, 10801.
. In the usual case, courts may strike down an agency’s interpretation of the statute it administers only if the agency’s interpretation is unreasonable or contrary to the statute’s plain meaning. See Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). Courts must also defer to an agency’s interpretation of a gap or ambiguity in its own regulations unless the interpretation is plainly wrong. See Bowles v. Seminole Rock & Sand Co., 325 U.S. 410, 414, 65 S.Ct. 1215, 1217, 89 L.Ed. 1700 (1945). See generally John F. Manning, Constitutional Structure and Judicial Deference to Agency Interpretations of Agency Rules, 96 Colum. L.Rev. 612 (1996). Meinhold II was not the usual case, however: Chevron deference does not apply when a narrowing construction avoids a tough constitutional question. See Mistretta v. United States, 488 U.S. 361, 373 n. 7, 109 S.Ct. 647, 655 n. 7, 102 L.Ed.2d 714 (1989); Williams v. Babbitt, 115 F.3d 657, 662-63 (9th Cir.1997).
. The majority violates the very rule it announces: "The government may avoid EAJA fees if it can prove that the regulation it violated was ambiguous, complex, or required exceptional analysis." Maj. op. at 1278 (emphasis added) (citing cases). Whatever else one may say about Meinhold II, surely the analysis there is nothing short of exceptional. See Steffan, 41 F.3d at 694 n. 18.