with whom The Chief Justice, Justice O’Connor, and Justice Thomas join, dissenting.
Section 504(a)(16) of the Omnibus Consolidated Rescis-sions and Appropriations Act of 1996 (Appropriations Act) defines the scope of a federal spending program. It does not directly regulate speech, and it neither establishes a public forum nor discriminates on the basis of viewpoint. The Court agrees with all this, yet applies a novel and unsupportable interpretation of our public-forum precedents to declare § 504(a)(16) facially unconstitutional. This holding not only has no foundation in our jurisprudence; it is flatly contradicted by a recent decision that is on all fours with the present cases. Having found the limitation upon the spending program unconstitutional, the Court then declines to consider the question of severability, allowing a judgment to stand that lets the program go forward under a version of *550the statute Congress never enacted. I respectfully dissent from both aspects of the judgment.
I
The Legal Services Corporation Act of 1974 (LSC Act), 42 U. S. C. § 2996 et seq., is a federal subsidy program, the stated purpose of which is to “provid[e] financial support for legal assistance in noncriminal proceedings or matters to persons financially unable to afford legal assistance.” §2996b(a). Congress, recognizing that the program could not serve its purpose unless it was “kept free from the influence of or use by it of political pressures,” § 2996(5), has from the program’s inception tightly regulated the use of its funds. See ante, at 537-538. No Legal Services Corporation (LSC) funds may be used, for example, for “encouraging ... labor or antilabor activities,” §2996f(b)(6), for “litigation relating to the desegregation of any elementary or secondary school or school system,” § 2996f(b)(9), or for “litigation which seeks to procure a nontherapeutic abortion,” § 2996f(b)(8). Congress discovered through experience, however, that these restrictions did not exhaust the politically controversial uses to which LSC funds could be put.
Accordingly, in 1996 Congress added new restrictions to the LSC Act and strengthened existing restrictions. Among the new restrictions is the one at issue here. Section 504(a)(16) of the Appropriations Act, 110 Stat. 1321-55 to 1321-56, withholds LSC funds from every entity that “participates in any . . . way ... in litigation, lobbying, or rulemaking . . . involving an effort to reform a Federal or State welfare system.” It thus bans LSC-funded entities from participating on either side of litigation involving such statutes, from participating in rulemaking relating to the implementation of such legislation, and from lobbying Congress itself regarding any proposed changes to such legislation. See 45 CFR § 1639.3 (2000).
*551The restrictions relating to rulemaking and lobbying are superfluous; they duplicate general prohibitions on the use of LSC funds for those activities found elsewhere in the Appropriations Act. See §§ 504(a)(2), (3), (4). The restriction on litigation, however, is unique, and it contains a proviso specifying what the restriction does not cover. Funding recipients may “represent an individual eligible client who is seeking specific relief from a welfare agency if such relief does not involve an effort to amend or otherwise challenge existing law in effect on the date of the initiation of the representation.” The LSC declares in its brief, and respondents do not deny, that under these provisions the LSC can sponsor neither challenges to nor defenses of existing welfare reform law, Brief for Petitioner in No. 99-603, p. 29. The litigation ban is symmetrical: Litigants challenging the covered ■ statutes or regulations do not receive LSC funding, and neither do litigants defending those laws against challenge.
If a suit for benefits raises a claim outside the scope of the LSC program, the LSC-funded lawyer may not participate in the suit. As the Court explains, if LSC-funded lawyers anticipate that a forbidden claim will arise in a prospective client’s suit, they “may not undertake [the] representation,” ante, at 544. Likewise, if a forbidden claim arises unexpectedly at trial, “LSC-funded attorney[s] must cease the representation at once,” ante, at 544-545. See also Brief for Petitioner in No. 99-603, at 7, n. 4 (if the issue arises at trial, “the lawyer should discontinue the representation 'consistent with the applicable rules of professional responsibility’ ”). The lawyers may, however, and indeed must explain to the client why they cannot represent him. See 164 F. 3d 757, 765 (CA2 1999). They are also free to express their views of the legality of the welfare law to the client, and they may refer the client to another attorney who can accept the representation, ibid. See 985 F. Supp. 323, 335-336 (EDNY 1997).
*552II
The LSC Act is a federal subsidy program, not a federal regulatory program, and “[t]here is a basic difference between [the two].” Maher v. Roe, 432 U. S. 464, 475 (1977). Regulations directly restrict speech; subsidies do not. Subsidies, it is true, may indirectly abridge speech, but only if the funding scheme is “ ‘manipulated’ to have a ‘coercive effect’ ” on those who do not hold the subsidized position. National Endowment for Arts v. Finley, 524 U. S. 569, 587 (1998) (quoting Arkansas Writers’ Project, Inc. v. Ragland, 481 U. S. 221, 237 (1987) (Scalia, J., dissenting)). Proving unconstitutional coercion is difficult enough when the spending program has universal coverage and excludes only certain speech — such as a tax exemption scheme excluding lobbying expenses. The Court has found such programs unconstitutional only when the exclusion was “aimed at the suppression of dangerous ideas.” Speiser v. Randall, 357 U. S. 513, 519 (1958) (internal quotation marks omitted); see also Regan v. Taxation With Representation of Wash., 461 U. S. 540, 550 (1983). Proving the requisite coercion is harder still when a spending program is not universal but limited, providing benefits to a restricted number of recipients, see Rust v. Sullivan, 500 U. S. 173, 194-195 (1991). The Court has found such selective spending unconstitutionally coercive only once, when the government created a public forum with the spending program but then discriminated in distributing funding within the forum on the basis of viewpoint. See Rosenberger v. Rector and Visitors of Univ. of Va., 515 U. S. 819, 829-830 (1995). When the limited spending program does not create a public forum, proving coercion is virtually impossible, because simply denying a subsidy “does not ‘coerce’ belief,” Lyng v. Automobile Workers, 485 U. S. 360, 369 (1988), and because the criterion of unconstitutionality is whether denial of the subsidy threatens “to drive certain ideas or viewpoints from the marketplace,” National Endowment for Arts v. Finley, supra, at 587 (internal quota*553tion marks omitted). Absent such a threat, “the Government may allocate . . . funding according to criteria that would be impermissible were direct regulation of speech or a criminal penalty at stake.” 524 U. S., at 587-588.
In Rust v. Sullivan, supra, the Court applied these principles to a statutory scheme that is in all relevant respects indistinguishable from § 504(a)(16). The statute in Rust authorized grants for the provision of family planning services, but provided that “[n]one of the funds . . . shall be used in programs where abortion is a method of family planning.” Id., at 178. Valid regulations implementing the statute required funding recipients to refer pregnant clients “for appropriate prenatal. . . services by furnishing a list of available providers that promote the welfare of mother and unborn child,” but forbade them to refer a pregnant woman specifically to an abortion provider, even upon request. Id., at 180. We rejected, a First Amendment free-speech challenge to the funding scheme, explaining that “[t]he Government can, without violating the Constitution, selectively fund a program to encourage certain activities it believes to be in the public interest, without at the same time funding an alternative program which seeks to deal with the problem another way.” Id., at 193. This was not, we said, the type of “discriminat[ion] on the basis of viewpoint” that triggers strict scrutiny, ibid., because the “ 'decision not to subsidize the exercise of a fundamental right does not infringe the right,’ ” ibid, (quoting Regan v. Taxation With Representation of Wash., supra, at 549).
The same is true here. The LSC Act, like the scheme in Rust, see 500 U. S., at 200, does not create a public forum. Far from encouraging a diversity of views, it has always, as the Court accurately states, “placed restrictions on its use of funds,” ante, at 537. Nor does §504(a)(16) discriminate on the basis of viewpoint, since it funds neither challenges to nor defenses of existing welfare law. The provision simply declines to subsidize a certain class of litigation, and under *554Rust that decision “does not infringe the right” to bring such litigation. Cf. Ortwein v. Schwab, 410 U. S. 656, 658-660, and n. 5 (1978) (per curiam) (government not required by First Amendment or Due Process Clause to waive filing fee for welfare benefits litigation). The Court’s repeated claims that § 504(a)(16) “restricts” and “prohibits” speech, see, e. g., ante, at 545, 546, and “insulates” laws from judicial review, see, e. g., ante, at 547, are simply baseless. Ño litigant who, in the absence of LSC funding, would bring a suit challenging existing welfare law is deterred from doing so by §504(a)(16). Rust thus controls these cases and compels the conclusion that § 504(a)(16) is constitutional.
The Court contends that Rust is different because the program at issue subsidized government speech, while the LSC funds private speech. See ante, at 541-542. This is so unpersuasive it hardly needs response. If the private doctors’ confidential advice to their patients at issue in Rust constituted “government speech,” it is hard to imagine what subsidized speech would not be government speech. Moreover, the majority’s contention that the subsidized speech in these cases is not government speech because the lawyers have a professional obligation to represent the interests of their clients founders on the reality that the doctors in Rust had a professional obligation to serve the interests of their patients, see 500 U. S., at 214 (Blackmun, J., dissenting) (“ethical responsibilities of the medical profession”) — which at the time of Rust we had held to be highly relevant to the permissible scope of federal regulation, see Thornburgh v. American College of Obstetricians and Gynecologists, 476 U. S. 747, 763 (1986) (“professional responsibilities” of physicians), overruled in part on other grounds, Planned Parenthood of Southeastern Pa. v. Casey, 505 U. S. 833 (1992). Even respondents agree that “the true speaker in Rust was not the government, but a doctor.” Brief for Respondents 19, n. 17.
The Court further asserts that these cases are different from Rust because the welfare funding restriction “seeks to *555use an existing medium of expression and to control it... in ways which distort its usual functioning,” ante, at 543. This is wrong on both the facts and the law. It is wrong on the law because there is utterly no precedent for the novel and facially implausible proposition that the First Amendment has anything to do with government funding that — though it does not actually abridge anyone’s speech — “distorts an existing medium of expression.” None of the three cases cited by the Court mentions such an odd principle. In Ro-senberger v. Rector and Visitors of Unir, of Va., the point critical to the Court’s analysis was not, as the Court would have it, that it is part of the “usual functioning” of student newspapers to “expres[s] many different points of view,” ante, at 543 (it surely is not), but rather that the spending program itself Yis,& been created “to encourage a diversity of views from private speakers,” 515 U. S., at 834. What could not be distorted was the public forum that the spending program had created. As for Arkansas Ed. Television Comm’n v. Forbes, 523 U. S. 666 (1998), that case discussed the nature of television broadcasting, not to determine whether government regulation would alter its “usual functioning” and thus violate the First Amendment (no government regulation was even at issue in the case), but rather to determine whether state-owned television is a “public forum” under our First Amendment jurisprudence. Id., at 673-674. And finally, the passage the Court cites from FCC v. League of Women Voters of Cal., 468 U. S. 364, 396-397 (1984), says nothing whatever about “using the forum [of public radio] in an unconventional way to suppress speech inherent in the nature of the medium,” ante, at 543. It discusses why the Government’s asserted interest in “preventing [public radio] stations from becoming a privileged outlet for the political and ideological opinions of station owners and managers,” 468 U. S., at 396 (internal quotation marks omitted), was insubstantial and thus could not justify the statute’s restriction on editorializing. Even worse for the Court, after invalidat*556ing the restriction on this conventional First Amendment ground, League of Women Voters goes on to say that “[o]f course,” the restriction on editorializing “would plainly be valid” if “Congress were to adopt a revised version of [the statute] that permitted [public radio] stations to establish ‘affiliate’ organizations which could then use the station’s facilities to editorialize with nonfederal funds.” Id., at 400. But of course that is the case here. Regulations permit funding recipients to establish affiliate organizations to conduct litigation and other activities that fall outside the scope of the LSC program. See 45 CFR pt. 1610 (2000). Far from supporting the Court’s nondistortion analysis, League of Women Voters dooms the Court’s case.
The Court’s “nondistortion” principle is also wrong on the facts, since there is no basis for believing that §504(a)(16), by causing “eases [to] be presented by LSC attorneys who [can]not advise the courts of serious questions of statutory validity,” ante, at 545, will distort the operation of the courts. It may well be that the bar of § 504(a)(16) will cause LSC-funded attorneys to decline or to withdraw from cases that' involve statutory validity. But that means at most that fewer statutory challenges to welfare laws will be presented to the courts because of the unavailability of free legal services for that purpose. So what? The same result would ensue from excluding LSC-funded lawyers from welfare litigation entirely. It is not the mandated, nondistortable function of the courts to inquire into all “serious questions of statutory validity” in all eases. Courts must consider only those questions of statutory validity that are presented by litigants, and if the Government chooses not to subsidize the presentation of some such questions, that in no way “distorts” the courts’ role. It is remarkable that a Court that has so studiously avoided deciding whether Congress could entirely eliminate federal jurisdiction over certain matters, see, e. g., Webster v. Doe, 486 U. S. 592, 603 (1988); Bowen v. Michigan Academy of Family Physicians, 476 U. S. 667, *557681, n. 12 (1986), would be so eager to hold the much lesser step of declining to subsidize the litigation unconstitutional under the First Amendment.
Nor will the judicial opinions produced by LSC cases systematically distort the interpretation of welfare laws. Judicial decisions do not stand as binding “precedent” for points that were not raised, not argued, and hence not analyzed. See, e. g., United States v. Verdugo-Urquidez, 494 U. S. 259, 272 (1990); Hagans v. Lavine, 415 U. S. 528, 533, n. 5 (1974); United States v. L. A. Tucker Truck Lines, Inc., 344 U. S. 33, 37-38 (1952); United States v. More, 3 Cranch 159, 172 (1805) (Marshall, C. J.). The statutory validity that courts assume in LSC cases will remain open for full determination in later cases.
Finally, the Court is troubled “because in eases where the attorney withdraws from a representation, the client is unlikely to find other counsel.” Ante, at 546. That is surely irrelevant, since it leaves the welfare recipient in no worse condition than he would have been in had the LSC program never been enacted. Respondents properly concede that even if welfare claimants cannot obtain a lawyer anywhere else, the Government is not required to provide one. Brief for Respondents 16; accord, Goldberg v. Kelly, 397 U. S. 254, 270 (1970) (government not required to provide counsel at hearing regarding termination of welfare benefits). It is hard to see how providing free legal services to some welfare claimants (those whose claims do not challenge the applicable statutes) while not providing it to others is beyond the range of legitimate legislative choice. Rust rejected a similar argument:
“Petitioners contend, however, that most Title X clients are effectively precluded by indigency and poverty from seeing a health-care provider who will provide abortion-related services. But once again, even these Title X clients are in no worse position than if Congress had never enacted Title X. The financial constraints *558that restrict an indigent woman’s ability to enjoy the full range of constitutionally protected freedom of choice are the product not of governmental restrictions on access to abortion, but rather of her indigency.” 500 U. S., at 203 (internal quotation marks omitted).
The only conceivable argument that can be made for distinguishing Rust is that there even patients who wished to receive abortion counseling could receive the nonabortion services that the Government-funded clinic offered, whereas here some potential LSC clients who wish to receive representation on a benefits claim that does not challenge the statutes will be unable to do so because their eases raise a reform claim that an LSC lawyer may not present. This difference, of course, is required by the same ethical canons that the Court elsewhere does not wish to distort. Rather than sponsor “truncated representation,” ante, at 546, Congress chose to subsidize only those cases in which the attorneys it subsidized could work freely. See, e. g., 42 U. S. C. § 2996(6) (“[Ajttorneys providing legal assistance must have full freedom to protect the best interests of their clients”). And it is impossible to see how this difference from Rust has any bearing upon the First Amendment question, which, to repeat, is whether the funding scheme is “'manipulated’ to have a ‘coercive effect’ ” on those who do not hold the subsidized position. National Endowment for Arts v. Finley, 524 U. S., at 587 (quoting Arkansas Writers’ Project, Inc. v. Ragland, 481 U. S., at 237 (Scalia, J., dissenting)). It could be claimed to have such an effect if the client in a ease ineligible for LSC representation could eliminate the ineligibility by waiving the claim that the statute is invalid; but he cannot. No conceivable coercive effect exists.
This has been a very long discussion to make a point that is embarrassingly simple: The LSC subsidy neither prevents anyone from speaking nor coerces anyone to change speech, and is indistinguishable in all relevant respects from the sub*559sidy upheld in Rust v. Sullivan, supra. There is no legitimate basis for declaring § 504(a)(16) facially unconstitutional.
III
Even were I to accept the Court s First Amendment analysis, I could not join its decision to conclude this litigation without reaching the issue of severability. That issue, although decided by the Second Circuit, was not included within the question on which certiorari was granted, and, as the Court points out, was not briefed or argued here. I nonetheless think it an abuse of discretion to ignore it.
The Court has said that “[w]e may consider questions outside the scope of the limited order [granting certiorari] when resolution of those questions is necessary for the proper disposition of the case.” Piper Aircraft Co. v. Reyno, 454 U. S. 235, 246-247, n. 12 (1981). I think it necessary to a “proper disposition” here because the statute concocted by the Court of Appeals bears little resemblance to what Congress enacted, funding without restriction welfare-benefits litigation that Congress funded only under the limitations of § 504(a)(16). Although no party briefed severability in Denver Area Ed. Telecommunications Consortium, Inc. v. FCC, 518 U. S. 727 (1996), the Justices finding partial unconstitutionality considered it necessary to address the issue. Id., at 767 (plurality opinion) (“[W]e must ask whether § 10(a) is severable”); accord, New York v. United States, 505 U. S. 144, 186 (1992). I think we have that same obligation here. Moreover, by exercising our “discretion” to leave the sever-ability question open, we fail to resolve the basic, real-world dispute at issue: whether LSC attorneys may represent welfare claimants who challenge the applicable welfare laws. Indeed, we leave the LSC program subject to even a greater uncertainty than the one we purport to have eliminated, since other circuits may conclude (as I do) that if the limitation upon welfare representation is unconstitutional, LSC attorneys cannot engage in welfare litigation at all.
*560“The inquiry into whether a statute is severable is essentially an inquiry into legislative intent.” Minnesota v. Mille Lacs Band of Chippewa Indians, 526 U. S. 172, 191 (1999). If Congress “would not have enacted those provisions which are within its power, independently of that which is not,” then courts must strike the provisions as a piece. Alaska Airlines, Inc. v. Brock, 480 U. S. 678, 684 (1987) (internal quotation marks omitted). One determines what Congress would have done by examining what it did. Perhaps the most that can be said on the subject is contained in a passage written by Chief Justice Shaw of the Supreme Judicial Court of Massachusetts that we have often quoted:
“[I]f [a statute’s provisions] are so mutually connected with and dependent on each other, as conditions, considerations or compensations for each other, as to warrant a belief that the legislature intended them as a whole, and that, if all could not be carried into effect, the legislature would not pass the residue independently, and some parts are unconstitutional, all the provisions which as thus dependent, conditional or connected, must fall with them.” Warren v. Mayor and Aldermen of Charlestown, 68 Mass. 84, 99 (1854).
It is clear to me that the LSC Act’s funding of welfare benefits suits and its prohibition on suits challenging or defending the validity of existing law are “conditions, considerations [and] compensations for each other” that cannot be severed. Congress through the LSC Act intended “to provide high quality legal assistance to those who would be otherwise unable to afford adequate legal counsel,” 42 U. S. C. §2996(2), but only if the program could at the same time “be kept free from the influence of or use by it of political pressures,” §2996(5). More than a dozen times in § 504(a) Congress made the decision that certain activities could not be funded at all without crippling the LSC program with political pressures. See, e.g., § 504(a)(1) (reapportionment *561litigation); § 504(a)(4) (local, state, and federal lobbying); § 504(a)(7) (class-action lawsuits); §504(a)(12) (training programs for, inter alia, boycotts, picketing, and demonstrations); § 504(a)(14) (litigation with respect to abortion). The severability question here is, essentially, whether, without the restriction that the Court today invalidates, the permission for conducting welfare litigation would have been accorded. As far as appears from the best evidence (which is the structure of the statute), I think the answer must be no.
We have in some cases stated that when an “excepting proviso is found unconstitutional the substantive provisions which it qualifies cannot stand,” for “to hold otherwise would be to extend the scope of the law ... so as to embrace [situations] which the legislature passing the statute had, by its very terms, expressly excluded.” Frost v. Corporation Comm’n of Okla., 278 U. S. 515, 525 (1929); see also Davis v. Wallace, 257 U. S. 478, 484 (1922) (“Where an excepting provision in a statute is found unconstitutional, courts very generally hold that this does not work an enlargement of the scope or operation of other provisions with which that provision was enacted, and which it was intended to qualify or restrain”). I frankly doubt whether this approach has been followed consistently enough to be called the “general” rule, but if there were ever an instance in which it is appropriate it is here. To strike the restriction on welfare benefits suits is to void § 504(a)(16) altogether. Subsection (a)(16) prohibits involvement in three types of activities with respect to welfare reform: lobbying, rulemaking, and litigation. But the proscriptions against using LSC funds to participate in welfare lobbying and rulemaking are superfluous, since as described above subsections (a)(2), (a)(3), and (a)(4) of §504 withhold LSC funds from those activities generally. What is unique about subsection (a)(16) — the only thing it achieves — is its limit on litigation. To remove that limit is to repeal subsection (a)(16) altogether, and thus to eliminate a significant quid pro quo of the legislative compromise. We *562have no authority to “rewrite [the] statute and give it an effect altogether different” from what Congress agreed to. Railroad Retirement Bd. v. Alton R. Co., 295 U. S. 330, 362 (1935) (quoted in Carter v. Carter Coal Co., 298 U. S. 238, 313 (1936)).
* * *
It is illuminating to speculate how these cases would have been decided if Congress had enacted § 504(a)(16) without its proviso (prescribing only the general ban against “litigation, lobbying, or rulemaking, involving an effort to reform a Federal or State welfare system”), and if the positions of the parties before us here were reversed. If the LSC-funded lawyers were here arguing that the statute permitted representation of individual welfare claimants who did not challenge existing law, I venture to say that the Court would endorse their argument — perhaps with stirring language about the importance of aid to welfare applicants and the Court’s unwillingness to presume without clear indication that Congress would want to eliminate it. And I have little doubt that in that context the Court would find its current First Amendment musings as unpersuasive as I find them today.
Today’s decision is quite simply inexplicable on the basis of our prior law. The only difference between Rust and the present cases is that the former involved “distortion” of (that is to say, refusal to subsidize) the normal work of doctors, and the latter involves “distortion” of (that is to say, refusal to subsidize) the normal work of lawyers. The Court’s decision displays not only an improper special solicitude for our own profession; it also displays, I think, the very fondness for “reform through the courts” — the making of innumerable social judgments through judge-pronounced constitutional imperatives — that prompted Congress to restrict publicly funded litigation of this sort. The Court says today, through an unprecedented (and indeed previously rejected) interpretation of the First Amendment, that we will not allow this *563restriction — and then, to add insult to injury, permits to stand a judgment that awards the general litigation funding that the statute does not contain. I respectfully dissent.