Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
Concurring opinion filed by Circuit Judge RANDOLPH.
Dissenting opinion filed by Circuit Judge SENTELLE.
STEPHEN F. WILLIAMS, Circuit Judge:
In § 501(b) of the Ethics in Government Act, 5 U.S.C. app. § 501 et seq., Congress provided that “[a]n individual may not receive any honorarium while that individual is a Member [of Congress, or] officer or employee [of the federal government].” Congress defined “honorarium” as “a payment of money or anything of value for an appearance, speech or article (including a series of appearances, speeches, or articles if the subject matter is directly related to the individual’s official duties or the payment is made because of the individual’s status with the Government) ... excluding any actual and necessary travel expenses.” Id. § 505(3). The Office of Government Ethics has promulgated regulations implementing the Act for officers and employees of the executive branch. See 56 Fed.Reg. 1721 (January 17, 1991) (to be codified at 5 CFR § 2636.101ff.); 57 Fed.Reg. 601 (January 8, 1992) (amending 5 C.F.R. § 2636.-203).
Employees of the executive branch, and several unions of such employees, responded to enactment of the honorarium ban by challenging it in district court as a violation of their rights under the First Amendment. The National Treasury Employees Union was certified as the class representative for all affected executive branch employees below the grade of GS-16,1 and the various cases were consolidated.
On cross motions for summary judgment, the district court found the ban a violation of the First Amendment in so far as it affected the speech of executive branch employees.2 It enjoined enforcement, but stayed its judgment pending appeal. 788 F.Supp. 4. The government appeals from the judgment and injunction, and plaintiffs appeal from the stay. We affirm the judgment of the district court on the merits; this moots the problem of the stay.
Because the case involves a government burden on the speech of its own employees, Pickering v. Board of Education, 391 U.S. 563, 88 S.Ct. 1731, 20 L.Ed.2d 811 (1968), supplies the standard for judicial review of the congressional action. In Pickering a county had dismissed a teacher for publishing in a newspaper a letter criticizing the county school board’s allocation of funds. [3]*3While saying that the state could not make public employment conditional upon relinquishment of “the First Amendment rights [employees] would otherwise enjoy as citizens to comment on matters of public interest”, id. at 568, 88 S.Ct. at 1734, the Court also said that “the State has interests as an employer in regulating the speech of its employees that differ significantly from those it possesses in connection with regulation of the speech of the citizenry in general.” Id. It identified the “problem” as being “to arrive at a balance between the interests of the [employee], as a citizen, in commenting upon matters of public concern and the interest of the State, as an employer, in promoting the efficiency of the public services it performs through its employees.” Id.
As Pickering defines the employees’ speech interests in terms of “matter[s] of public concern”, see also Connick v. Myers, 461 U.S. 138, 146-47, 103 S.Ct. 1684, 1689-90, 75 L.Ed.2d 708 (1983), we pause briefly to consider whether this case involves such matters. In Connick the Court spoke broadly of expression “relating to any matter of political, social, or other concern to the community.” Viewing the idea of “public concern” in the abstract, one might suppose it excluded some of the topics on which plaintiffs have spoken or written — such as the technology of Civil War ironclads. See Joint Appendix (“J.A.”) at 119.
But Connick makes clear that the “public concern” criterion does not require any great intensity or breadth of public interest in the subject. It is thus far broader than the sort of “public questions” in which a person must be involved for application of New York Times v. Sullivan, 376 U.S. 254, 84 S.Ct. 710, 11 L.Ed.2d 686 (1964). See, e.g., Gertz v. Robert Welch, Inc., 418 U.S. 323, 351, 94 S.Ct. 2997, 3012, 41 L.Ed.2d 789 (1974); Waldbaum v. Fairchild Publications, Inc., 627 F.2d 1287, 1296 (D.C.Cir.1980). In Connick the employee had circulated a questionnaire asking fellow employees for their views on such matters as the level of office morale, their confidence in various supervisors and the need for a grievance committee. These the Court wrote off as “mere extensions of Myers’ dispute over her transfer”. 461 U.S. at 148, 103 S.Ct. at 1690. In contrast, it found a question on whether employees ever felt “pressured to work in political campaigns on behalf of office supported candidates” to be of interest to the community. Id. at 149, 103 S.Ct. at 1691. The contrast, then, was between issues of external interest as opposed to ones of internal office management. See also id. at 146, 103 S.Ct. at 1689 (invoking need for officials to “enjoy wide latitude in managing their offices, without intrusive oversight by the judiciary”). Accordingly, we read the “public concern” criterion as referring not to the number of interested listeners or readers but to whether the expression relates to some issue of interest beyond the employee’s bureaucratic niche. None of the samples of past or intended expression mentioned by plaintiffs involves such a parochial concern.
Although § 501(b) prohibits no speech, it places a financial burden on speech — denial of compensation. While the employees’ First Amendment interest is therefore somewhat less weighty than under a flat ban, there can be no doubt that the burden counts for purposes of the Pickering balance. In Simon & Schuster, Inc. v. New York State Crime Victims Board, — U.S. -, 112 S.Ct. 501, 116 L.Ed.2d 476 (1991), the Supreme Court considered a statute that singled out compensation for writings by a person convicted or accused of a crime on the subject of his crime. Though at times seeming to characterize the statute as content-based, see, e.g., id. at-, 112 S.Ct. at 508, the Court ultimately declined to say whether it was or not, and invalidated it as not “narrowly tailored” enough even under “the more lenient tailoring standards applied” to content-neutral provisions, id. at---n. **, 112 S.Ct. at 511-12 n.**. The point that the burden was simply denial of compensation played no apparent role in the Court’s “tailoring” analysis. Similarly, the financial character of the limitation here affects only the “weight” of the employees’ interest in the Pickering balance.
[4]*4Neither party disputes that the government has a strong interest in protecting the integrity and efficiency of public service and in avoiding even the appearance of impropriety created by abuse of the practice of receiving honoraria. Indeed, in Keeffe v. Library of Congress, 777 F.2d 1573, 1581 (D.C.Cir.1985), we identified “prevent[ing] erosion of congressional and public confidence in the integrity of the [Congressional Research] Service” as a compelling justification for limits on its analysts’ participation in political activities. We can safely assume for the purposes of this opinion that the interest in avoiding the appearance of impropriety is strong enough to outweigh government employees’ interest in engaging in speech for compensation where the compensation creates such an appearance, and so is strong enough to justify a ban on compensation in those circumstances. Thus, for some of § 501(b)’s applications — perhaps many of them — the Pickering balance supports its constitutionality.
That conclusion will not save § 501(b), however, if it either is “overbroad” or manifests a want of “narrow tailoring.” Plaintiffs in their attack on the statute use the terms more or less interchangeably. So has the Supreme Court, on occasion, when speaking of the substance of the doctrines, i.e., what they demand of a statute in terms of focus on a genuine evil. Thus in Austin v. Michigan Chamber of Commerce, 494 U.S. 652, 110 S.Ct. 1391, 108 L.Ed.2d 652 (1990), the Court started its analysis of the subject by posing the question whether the statute was “sufficiently narrowly tailored to achieve its goal”, id. at 660, 110 S.Ct. at 1397, and ended by finding that it was “not substantially overbroad”, id. at 661, 110 S.Ct. at 1398. See also Secretary of State of Maryland v. J.H. Munson Co., 467 U.S. 947, 965-66 n. 13, 104 S.Ct. 2839, 2851-52 n. 13, 81 L.Ed.2d 786 (1984) (“‘Overbreadth’ has also been used to describe a challenge to a statute that ... does not employ means narrowly tailored to serve a compelling governmental interest”); cf. Simon & Schuster, Inc. v. New York State Crime Victims Board, — U.S. at-, 112 S.Ct. at 511-12 (addressing whether law “is significantly overinclusive” and concluding that it “is ... not narrowly tailored to achieve the State’s objective”). Moreover, the terms used by the Court to describe the doctrines’ substance do not seem to suggest a material difference. As to overbreadth, the Court said in Broadrick v. Oklahoma, 413 U.S. 601, 93 S.Ct. 2908, 37 L.Ed.2d 830 (1973), “[Particularly where conduct and not merely speech is involved, we believe that the overbreadth of a statute must not only be real [i.e., not based on speculative constructions of the statute] but substantial as well, judged in relation to the statute’s plainly legitimate sweep.” Id. at 615, 93 S.Ct. at 2917. And in its most explicit treatment of narrow tailoring, Ward v. Rock Against Racism, 491 U.S. 781, 109 S.Ct. 2746, 105 L.Ed.2d 661 (1989), it said:
[T]he requirement of narrow tailoring is satisfied “so long as the ... regulation promotes a substantial government interest that would be achieved less effectively absent the regulation.” ... To be sure, this standard does not mean that a ... regulation may burden substantially more speech than is necessary to further the government’s legitimate interests. Government may not regulate expression in such a manner that a substantial portion of the burden on speech does not serve to advance its goals.
Id. at 799, 109 S.Ct. at 2758 (citations and footnote omitted). Although Ward involved a time, place, and manner restriction, the Court has since indicated that this concept of “tailoring” applies to any content-neutral restriction on speech. See Simon & Schuster, — U.S. at-n. **, 112 S.Ct. at 511 n.**. Thus both doctrines invalidate a statute that imposes “substantial” burdens that are not supported by the statute’s justifications; i.e., both invalidate statutes that are unduly “overinclusive.” Conceivably the quality of fit demanded by the two doctrines differs. The Court has stressed that overbreadth is “strong medicine”, Broadrick, 413 U.S. at 613, 93 S.Ct. at 2916, possibly suggesting that a plaintiff can prevail under that doctrine only by showing more drastic overinclusive[5]*5ness than would be necessary to prevail on a narrow tailoring claim. Such a view might make some sense, as overbreadth (in its classic form) gives the plaintiff a procedural advantage — the ability to challenge a statute on the basis of its effect on others — see, e.g., Thornhill v. Alabama, 310 U.S. 88, 97-98, 60 S.Ct. 736, 741-742, 84 L.Ed. 1093 (1940); City Council v. Taxpayers for Vincent, 466 U.S. 789, 796-98, 104 S.Ct. 2118, 2124-25, 80 L.Ed.2d 772 (1984); Board of Trustees v. Fox, 492 U.S. 469, 482-83, 109 S.Ct. 3028, 3035-36, 106 L.Ed.2d 388 (1989) — in exchange for which he might be expected to surmount an exceptional substantive hurdle. But, as developed below, we cannot see that § 501(b) satisfies even the most lenient form of the requirement.
Although the Supreme Court has occasionally suggested that courts may proceed to consider a facial overbreadth challenge only after having determined that it is valid as applied to the challengers, see, e.g., Fox, 492 U.S. at 484-85, 109 S.Ct. at 3037, in fact the Court allows facial overinclusiveness claims by parties whose conduct may well be constitutionally protected — such as many of the plaintiffs here. In Austin v. Michigan Chamber of Commerce, for example, the Chamber of Commerce mounted a pre-enforcement facial challenge to a state restriction on corporate political expenditures, maintaining that its speech was in fact protected. 494 U.S. at 661-65. Yet the Court addressed the facial attack on the statute’s scope.3 See also New York State Club Ass’n v. New York City, 487 U.S. 1, 13-15, 108 S.Ct. 2225, 2234-35, 101 L.Ed.2d 1 (1988) (applying standard over-breadth analysis without even considering whether members of the Association could be legitimately subject to the ordinance); Boos v. Barry, 485 U.S. 312, 329-32, 108 S.Ct. 1157, 1168-69, 99 L.Ed.2d 333 (1988) (applying overbreadth analysis without either a concession by plaintiffs that their speech could be constitutionally restricted or a finding to that effect, and indeed in the face of indications that they claimed their speech was protected, see id. at 315-16, 108 S.Ct. at 1160-61), affirming in relevant part Finzer v. Barry, 798 F.2d 1450, 1472 (D.C.Cir.1986) (addressing facial attack on the statute’s scope in the face of the challengers’ assertion that their speech was protected); but cf. Sanjour v. EPA, 984 F.2d 434, 444 & n. 10 (D.C.Cir.1993).
Accordingly we reach the merits of the plaintiffs’ overinclusiveness claims. To create the sort of impropriety or appearance of impropriety at which the statute is evidently aimed, there would have to be some sort of nexus between the employee’s job and either the subject matter of the expression or the character of the payor. But as to many of the plaintiffs, the government identifies no such nexus. These plaintiffs include a Nuclear Regulatory Commission lawyer who writes on Russian history of the late Romanov era, see J.A. at 51; a Postal Service mailhandler who writes and gives speeches on the Quaker religion, id. at 63; a Department of Labor lawyer who lectures on Judaism, id. at 70; a Department of Health and Human Services employee who reviews art, musical, and theater performances for local newspapers, id. at 95; and a civilian Navy electronics technician who writes on Civil War ironclad vessel technology, id. at 119. The topics appear not to be such that the employee could have used information acquired in the course of his government work; there is no suggestion of any use of government time, word processors, paper or ink; there is no suggestion that the institutions that have paid or are likely to pay for the speeches or writings would have some relationship with the employee’s agency that would make them wish to curry its favor.
Of course the ban also covers payments for speeches and articles that may well create an appearance of impropriety. In fact, even some of the plaintiffs receive payments that might at least raise an eyebrow. For example, a business editor at [6]*6the Voice of America also receives payment for business analysis that he provides various media. J.A. 57-62. From his own account it seems at least quite possible that he uses information acquired on the job; and it is possible that the media to whom he sells might believe that his favor could improve the chances that VO A would use some of their materials. (The editor suggests that banning compensation for his writings and talks will make government service far less attractive to hard-working, intelligent, creative persons. Surely this is so. We do not here address whether such values might count in the Pickering balance.) Another plaintiff is a GS-7 “tax examining assistant”. See J.A. 85-86. In view of the universality of citizens’ subjection to the Internal Revenue Service, we can understand some anxiety about receipt of payment by such an employee. However, even assuming arguendo that § 501(b) could be lawfully applied to these last two plaintiffs, it is clear that the ban reaches a lot of compensation that has no nexus to government work that could give rise to the slightest concern.4
There remains the possibility that these apparent excesses might be legitimized by the enforcement difficulties — including the problem of government scrutiny of subject matter — that any narrower restriction would involve. (Ironically, one plaintiff objects to even the screening that is involved in enforcement of § 501(b). J.A. 64.) If manageable lines are available to limit the ban to genuinely troubling compensation, then excesses, even if they affected few speakers, would appear gratuitous.
In fact, however, the government points neither to improprieties in the pre-§ 501(b) era that would have been prevented by § 501(b) but not by the prior regulations, nor to any serious enforcement or line-drawing costs associated with those regulations. Cf. Boos v. Barry, 485 U.S. at 324-29, 108 S.Ct. 1165-68 (stating that the “most useful starting point” for assessing whether a statute is narrowly tailored is to “compare it with an analogous statute”). As summarized in testimony before the Senate Committee on Governmental Affairs, the prior regulations allowed honoraria so long as the speaker or writer held a rank lower than GS-16 and all of the following questions could be answered negatively:
(1) Is the honorarium offered for carrying out government duties or for an activity that focuses specifically on the employing agency’s responsibilities, policies and programs?
(2) Is the honorarium offered to the government employee or family member because of the official position held by the employee?
(3) Is the honorarium offered because of the government information that is being imparted?
(4) Is the honorarium offered by someone who does business with or wishes to do business with the employee in his or her official capacity?
(5) Were any government resources or time used by the employee to produce the materials for the article or speech or make the appearance?
S.Rep. No. 29, 102d Cong., 1st Sess., at 8 (1991). While some of the limits may have an amorphous quality about them (such as the one purporting to probe the motive of the honorarium’s offeror), there appears no actual experience of difficulty, and one can hypothesize rules of thumb that could con[7]*7strain government discretion. Indeed, Congress’s use of similar criteria to cover a “series of appearances, speeches, or articles”, which § 501(b) as amended allows unless “the subject matter is directly related to the individual’s official duties or the payment is made because of the individual’s status with the Government”, suggests that it regards such lines as entirely workable.
While no one questions the authority of Congress to enact broad prophylactic rules, see United Public Workers v. Mitchell, 330 U.S. 75, 102, 67 S.Ct. 556, 570, 91 L.Ed. 754 (1947), a mere “hypothetical possibili: ty” of a corrupt “exchange of political favors” is not enough. FEC v. National Conservative Political Action Committee, 470 U.S. 480, 498, 105 S.Ct. 1459, 1469, 84 L.Ed.2d 455 (1985); see also id. at 500-01, 105 S.Ct. at 1470 (government evidence trying to link corruption to independent expenditures by PACs fails to pass “rigorous” First Amendment standard of review); First Nat’l Bank of Boston v. Bellotti, 435 U.S. 765, 789, 98 S.Ct. 1407, 1422, 55 L.Ed.2d 707 (1978). Of course where it is in the nature of the evil to be averted that it will be concealed, the court will necessarily expect less evidence. Thus, in upholding the Hatch Act, 5 U.S.C. § 7324 et seq., the Court pointed out that one important concern was to ensure that “Government employees would be free from pressure and from express or tacit invitation to vote in a certain way or perform political chores in order to curry favor with their superiors rather than to act out their own beliefs.” United States Civil Service Comm’n v. National Ass’n of Letter Carriers, 413 U.S. 548, 566, 93 S.Ct. 2880, 2890, 37 L.Ed.2d 796 (1973). The potential for such subtle pressure is not only pervasive but inherently difficult to demonstrate or assess; thus, the absence of episodes coming to light is quite consistent with the congressional concern. Similarly, in Buckley v. Valeo, 424 U.S. 1, 26-27, 96 S.Ct. 612, 638, 46 L.Ed.2d 659 (1976), the Court noted that “the full scope of such pernicious practices [as political quid pro quos for large campaign contributions] can never be reliably ascertained,” and then went on to observe that “deeply disturbing examples” of such arrangements had surfaced in the 1972 elections. Evidently the Court believed that the surreptitious character of a possible evil could explain the absence of evidence, at least if the evil’s existence could reasonably be inferred (as from human nature and the character of the political process). In contrast, here the government does not suggest that the public actually perceives a risk of corruption in receipt by low-level government employees of remuneration for talks on topics that are wholly unrelated to their function, to audiences that are equally unrelated. Nor does it suggest any reason to infer the likelihood of such perceptions, nor any reason why we would not see evidence of such perceptions if they existed.
As the apparently excess sweep of § 501(b) is supported by no more than the theoretical possibilities viewed as inadequate in National Conservative Political Action Committee, we cannot find § 501(b) “narrowly tailored”.5
Our final step is to determine the proper remedy. In general, a court should “refrain from invalidating more of the statute than is necessary.” Alaska Airlines, Inc. v. Brock, 480 U.S. 678, 684, 107 S.Ct. 1476, 1479, 94 L.Ed.2d 661 (1987) (quoting Regan v. Time, Inc., 468 U.S. 641, 652, 104 S.Ct. 3262, 3268, 82 L.Ed.2d 487 (1984) (plurality opinion)). But the language of § 501(b) — “[a]n individual may not receive any [payment for an appearance, speech or article]” (emphasis added) — does not seem to admit of any construction that would trim off all or even most of the invalid applications to executive branch employees; indeed the government proposes no limiting construction. Articulation of some appropriate nexus test would seem a purely legislative act.
This leaves open the possibility, however, that § 501(b)’s application to executive [8]*8branch employees may be severable from the remainder of the statute. No party here has argued that § 501(b) is unconstitutional as applied to members of Congress, officers or employees of Congress, or judicial officers or employees, and any such claim would raise quite different considerations. Legislators and judges and their staffs are likely to have to deal with a wide range of issues, so that a party purportedly paying for speech or writing by any of them is more likely to anticipate gaining some advantage, or at least to be seen as hoping for such an advantage. The tradeoff between preventing excess applications and keeping administrative costs low would be different from what it is for executive branch employees.
Whether an'unconstitutional provision is severable “is largely a question of legislative intent, but the presumption is in favor of severability.” Regan v. Time, Inc., 468 U.S. at 653, 104 S.Ct. at 3269. “ ‘Unless it is evident that the Legislature would not have enacted those provisions which are within its power, independently of that which is not, the invalid part may be dropped if what is left is fully operative as a law.’ ” Buckley v. Valeo, 424 U.S. at 108-09, 96 S.Ct. at 677 (quoting Champlin Refining Co. v. Corporation Comm’n of Oklahoma, 286 U.S. 210, 234, 52 S.Ct. 559, 564, 76 L.Ed. 1062 (1932)).
Section 501(b) does not contain a severability clause, and the legislative history yields no direct evidence of intent concerning severability. Thus, we must ask whether there is anything in the history indicating that Congress would have enacted the honorarium ban if it had been aware of its unconstitutionality as applied to executive branch employees. Alaska Airlines, Inc. v. Brock, 480 U.S. at 685, 107 S.Ct. at 1480.
In this case the evidence is that it clearly would have gone forward as to the legislative branch and in all probability as to the judicial. First, the floor debates indicate that Congress was principally concerned that the receipt of honoraria by Members of Congress created the appearance of influence-buying. Speakers throughout the debates refer to “Members of Congress” in explaining why the honorarium ban was necessary. See, e.g., 135 Cong.Rec. H8756 (daily ed. Nov. 16, 1989). For example, at one point a congressman noted that “the elimination of honoraria will have a beneficial impact on the public’s perception of the integrity of Congress as an institution.” Id. at H8763 (emphasis added). Another congressman noted that the statute addressed “the underlying sources of abuse in the current income system for public employees, in particular for Members of Congress.” Id. at H8767 (emphasis added).
Similarly, the Report of the Bipartisan Task Force, issued after the Act was passed, also indicates a primary concern with the receipt of honoraria by Members of Congress. In describing the background of the ban the report states that “substantial payments to a Member of Congress for rendering personal services to outside organizations presents a significant and avoidable potential for conflict of interest.” Id. at H9256. The report continues by describing how the increase in Members’ honoraria income in recent years “has heightened the public perception that honoraria is [sic] a way for special interests to try to gain influence or buy access to Members of Congress,” id. at H9257, and noting the “growing concern that the practice of acceptance of honoraria by Members ... creates serious conflict of interest problems and threatens to undermine the institutional integrity of Congress.” Id. Nowhere did members of Congress display any specific concern with the receipt of honoraria by executive branch employees, much less indicate that the application of the ban to them was a condition of the bill’s passage.
Further, the honorarium ban was adopted as part of a package of which a key ingredient was a sharp increase in the salary of members of Congress, judges, and a limited class of senior executive branch officials. See id. at H9254, H926869 (describing Title III of the Act). The one clearly detectable interdependency between segments of the statute was between the ban and the salary increase. [9]*9See, e.g., id. at H8756 (statement of Congressman Packard); id. at H8766/1-2; id. at H8767; id. at H9254 (stating that the pay raise would be given “as part of the ban on honoraria”). Striking down the application of the honorarium ban as to members of Congress and judges — while leaving their salary increases in place (a constitutional necessity for the latter) — would truly violate the intent of Congress.
Nominally, the invalidation of the honorarium ban as to executive branch employees upsets some of the intended balance (the salary increase for senior officials survives, the honorarium ban falls). But as other, severe restrictions have applied to senior executive branch officials anyway, see, e.g., E.O. 12674 (April 12, 1989)6 (providing that “[no] employee who is appointed by the President to a full-time noncareer position in the executive branch ... shall receive any earned income for any outside employment or activity performed during that Presidential appointment”), the effect on senior officials benefitting from the salary increase may well be nil.
We cannot, as a technical matter, achieve the intended severance simply by striking the words “officer or employee” from § 501(b), as that would invalidate the ban beyond the executive branch. See § 505(3) (defining “officer or employee” as “any officer or employee of the government”, and in context indisputably encompassing employees of Congress and judicial officers and employees). However, given the far greater congressional interest in banning honoraria for the legislative and judicial branches, we think it a proper form of severance to strike “officer or employee” from § 501(b) except in so far as those terms encompass members of Congress, officers and employees of Congress, judicial officers and judicial employees. Compare 5 U.S.C. app. 6, § 101(f)(9)-(12) (definition of “officers and employees” in related ethics legislation, encompassing persons in all three branches but distinguishing between them). This severance is similar to the type employed by the Court in Brockett v. Spokane Arcades, Inc., 472 U.S. 491, 105 S.Ct. 2794, 86 L.Ed.2d 394 (1985). There the Court “pretermitfed]” the issue of . whether “lust” might be construed as referring only to morbid and shameful interests (rather than also encompassing “ ‘good, old-fashioned, healthy’ interest in sex”), id. at 498-501, 105 S.Ct. at 2798-2800, and instead severed from the statute any meaning of lust other than shameful and morbid interests, id. at 506-07, 105 S.Ct. at 2803.
The decision of the district court is
Affirmed.