delivered the opinion of the Court.

In 2004, voters in the State of Washington passed an initiative changing the State’s primary election system. The People’s Choice Initiative of 2004, or Initiative 872 (1-872), provides that candidates for office shall be identified on the ballot by their self-designated “party preference”; that voters may vote for any candidate; and that the top two votegetters for each office, regardless of party preference, advance to the general election. The Court of Appeals for the Ninth Circuit held 1-872 facially invalid as imposing an unconstitutional burden on state political parties’ First Amendment rights. Because 1-872 does not on its face impose a severe burden on political parties’ associational rights, and because respondents’ arguments to the contrary rest on factual assumptions about voter confusion that can be evaluated only in the context of an as-applied challenge, we reverse.

*445I

For most of the past century, Washington voters selected nominees for state and local offices using a blanket primary.1 From 1935 until 2003, the State used a blanket primary that placed candidates from all parties on one ballot and allowed voters to select a candidate from any party. See 1935 Wash. Laws §§1-5, pp. 60-64. Under this system, the candidate who won a plurality of votes within each major party became that party’s nominee in the general election. See 2003 Wash. Laws § 919, p. 775.

California used a nearly identical primary in its own elections until our decision in California Democratic Party v. Jones, 530 U. S. 567 (2000). In Jones, four political parties challenged California’s blanket primary, arguing that it unconstitutionally burdened their associational rights by forcing them to associate with voters who did not share their beliefs. We agreed and struck down the blanket primary as inconsistent with the First Amendment. In so doing, we emphasized the importance of the nomination process as “ ‘the crucial juncture at which the appeal to common principles may be translated into concerted action, and hence to political power in the community.’” Id., at 575 (quoting Tashjian v. Republican Party of Conn., 479 U. S. 208, 216 (1986)). We observed that a party’s right to exclude is central to its freedom of association, and is never “more important than in the process of selecting its nominee.” 530 U. S., at 575. California’s blanket primary, we concluded, severely burdened the parties’ freedom of association because it *446forced them to allow nonmembers to participate in selecting the parties’ nominees. That the parties retained the right to endorse their preferred candidates did not render the burden any less severe, as “[t]here is simply no substitute for a party’s selecting its own candidates.” Id., at 581.

Because California’s blanket primary severely burdened the parties’ associational rights, we subjected it to strict scrutiny, carefully examining each of the state interests offered by California in support of its primary system. We rejected as illegitimate three of the asserted interests: “producing elected officials who better represent the electorate,” “expanding candidate debate beyond the scope of partisan concerns,” and ensuring “the right to an effective vote” by allowing nonmembers of a party to vote in the majority party’s primary in “ ‘safe’ ” districts. Id., at 582-584. We concluded that the remaining interests — promoting fairness, affording voters greater choice, increasing voter participation, and protecting privacy — were not compelling on the facts of the case. Even if they were, the partisan California primary was not narrowly tailored to further those interests because a nonpartisan blanket primary, in which the top two votegetters advance to the general election regardless of party affiliation, would accomplish each of those interests without burdening the parties’ associational rights. Id., at 585-586. The nonpartisan blanket primary had “all the characteristics of the partisan blanket primary, save the constitutionally crucial one: Primary voters [were] not choosing a party’s nominee.” Ibid.

After our decision in Jones, the Court of Appeals for the Ninth Circuit struck down Washington’s primary as “materially indistinguishable from the California scheme.” Democratic Party of Washington State v. Reed, 343 F. 3d 1198, 1203 (2003). The Washington State Grange2 promptly pro*447posed 1-872 as a replacement.3 It passed with nearly 60% of the vote and became effective in December 2004.

Under 1-872, all elections for “partisan offices”4 are conducted in two stages: a primary and a general election. To participate in the primary, a candidate must file a “declaration of candidacy” form, on which he declares his “major or minor party preference, or independent status.” Wash. Rev. Code §29A.24.030 (Supp. 2005). Each candidate and his party preference (or independent status) is in turn designated on the primary election ballot. A political party cannot prevent a candidate who is unaffiliated with, or even repugnant to, the party from designating it as his party of preference. See App. 396-397, 595 (declaration of James K. Pharris, Exhibit C: Ruling Order, May 18, 2005, Wash. Admin. Code §434-215-015). In the primary election, voters may select “any candidate listed on the ballot, regardless of the party preference of the candidates or the voter.” Id., at 606, §434-262-012.

The candidates with the highest and second-highest vote totals advance to the general election, regardless of their *448party preferences. Ibid. Thus, the general election may pit two candidates with the same party preference against one another.5 Each candidate’s party preference is listed on the general election ballot, and may not be changed between the primary and general elections. See id., at 601, § 434-230-040.

Immediately after the State enacted regulations to implement 1-872, the Washington State Republican Party filed suit against a number of county auditors challenging the law on its face. The party contended that the new system violates its associational rights by usurping its right to nominate its own candidates and by forcing it to associate with candidates it does not endorse. The Washington State Democratic Central Committee and Libertarian Party of Washington State joined the suit as plaintiffs. The Washington State Grange joined as a defendant, and the State of Washington was substituted for the county auditors as defendant. The United States District Court for the Western District of Washington granted the political parties’ motions for summary judgment and enjoined the implementation of 1-872. See Washington State Republican Party v. Logan, 377 F. Supp. 2d 907, 932 (2005).

The Court of Appeals affirmed. 460 F. 3d 1108, 1125 (CA9 2006). It held that the 1-872 primary severely burdens the political parties’ associational rights because the party-preference designation on the ballot creates a risk that primary winners will be perceived as the parties’ nominees and produces an “impression of associatio[n]” between a candidate and his party of preference even when the party does not associate, or wish to be associated, with the candidate. Id., at 1119. The Court of Appeals noted a “constitutionally *449significant distinction between ballots and other vehicles for political expression,” reasoning that the risk of perceived association is particularly acute when ballots include party labels because such labels are typically used to designate candidates’ views on issues of public concern. Id., at 1121. And it determined that the State’s interests underlying 1-872 were not sufficiently compelling to justify the severe burden on the parties’ association. Concluding that the provisions of 1-872 providing for the party-preference designation on the ballot were not severable, the court struck down 1-872 in its entirety.

We granted certiorari, 549 U. S. 1251 (2007), to determine whether 1-872, on its face, violates the political parties’ associational rights.

II

Respondents object to 1-872 not in the context of an actual election, but in a facial challenge. Under United States v. Salerno, 481 U. S. 739 (1987), a plaintiff can only succeed in a facial challenge by “establish[ing] that no set of circumstances exists under which the Act would be valid,” i. e., that the law is unconstitutional in all of its applications. Id., at 745. While some Members of the Court have criticized the Salerno formulation, all agree that a facial challenge must fail where the statute has a “‘plainly legitimate sweep.’” Washington v. Glucksberg, 521 U. S. 702, 739-740, and n. 7 (1997) (Stevens, J., concurring in judgments). Washington’s primary system survives under either standard, as we explain below.6 In determining whether a law is facially in*450valid, we must be careful not to go beyond the statute’s facial requirements and speculate about “hypothetical” or “imaginary” cases. See United States v. Raines, 362 U. S. 17, 22 (1960) (“The delicate power of pronouncing an Act of Congress unconstitutional is not to be exercised with reference to hypothetical cases thus imagined”). The State has had no opportunity to implement 1-872, and its courts have had no occasion to construe the law in the context of actual disputes arising from the electoral context, or to accord the law a limiting construction to avoid constitutional questions. Cf. Yazoo & Mississippi Valley R. Co. v. Jackson Vinegar Co., 226 U. S. 217, 220 (1912) (“How the state court may apply [a statute] to other cases, whether its general words may be treated as more or. less restrained, and how far parts of it may be sustained if others fail are matters upon which we need not speculate now”). Exercising judicial restraint in a facial challenge “frees the Court not only from unnecessary pronouncement on constitutional issues, but also from premature interpretations of statutes in areas where their constitutional application might be cloudy.” Raines, supra, at 22.

Facial challenges are disfavored for several reasons. Claims of facial invalidity often rest on speculation. As a consequence, they raise the risk of “premature interpretation of statutes on the basis of factually barebones records.” Sabri v. United States, 541 U. S. 600, 609 (2004) (internal quotation marks and brackets omitted). Facial challenges also run contrary to the fundamental principle of judicial restraint that courts should neither “ ‘anticipate a question of constitutional law in advance of the necessity of deciding it’ ” nor “ ‘formulate a rule of constitutional law broader than is required by the precise facts to which it is to be applied.’” Ashwander v. TVA, 297 U. S. 288, 346-347 (1936) (Brandeis, J., concurring) (quoting Liverpool, New York & Philadelphia *451S. S. Co. v. Commissioners of Emigration, 113 U. S. 33, 39 (1885)). Finally, facial challenges threaten to short circuit the democratic process by preventing laws embodying the will of the people from being implemented in a manner consistent with the Constitution. We must keep in mind that “ ‘[a] ruling of unconstitutionality frustrates the intent of the elected representatives of the people.’ ” Ayotte v. Planned Parenthood of Northern New Eng., 546 U. S. 320, 329 (2006) (quoting Regan v. Time, Inc., 468 U. S. 641, 652 (1984) (plurality opinion)). It is with these principles in view that we turn to the merits of respondents’ facial challenge to 1-872.

A

The States possess a “‘broad power to prescribe the “Times, Places and Manner of holding Elections for Senators and Representatives,” Art. I, § 4, cl. 1, which power is matched by state control over the election process for state offices.’ ” Clingman v. Beaver, 544 U. S. 581, 586 (2005) (quoting Tashjian, 479 U. S., at 217); Timmons v. Twin Cities Area New Party, 520 U. S. 351, 358 (1997) (same). This power is not absolute, but is “subject to the limitation that [it] may not be exercised in a way that violates . . . specific provisions of the Constitution.” Williams v. Rhodes, 393 U. S. 23, 29 (1968). In particular, the State has the “ ‘responsibility to observe the limits established by the First Amendment rights of the State’s citizens,’ ” including the freedom of political association. Eu v. San Francisco County Democratic Central Comm., 489 U. S. 214, 222 (1989) (quoting Tashjian, supra, at 217).

Election regulations that impose a severe burden on associational rights are subject to strict scrutiny, and we uphold them only if they are “narrowly tailored to serve a compelling state interest.” Clingman, supra, at 586; see also Rhodes, supra, at 31 (“ ‘[0]nly a compelling state interest in the regulation of a subject within the State’s constitutional power to regulate can justify limiting First Amendment *452freedoms’ ” (quoting NAACP v. Button, 371 U. S. 415, 438 (1963))). If a statute imposes only modest burdens, however, then “the State’s important regulatory interests are generally sufficient to justify reasonable, nondiscriminatory restrictions” on election procedures. Anderson v. Celebrezze, 460 U. S. 780, 788 (1983). “Accordingly, we have repeatedly upheld reasonable, politically neutral regulations that have the effect of channeling expressive activity at the polls.” Burdick v. Takushi, 504 U. S. 428, 438 (1992).

The parties do not dispute these general principles; rather, they disagree about whether 1-872 severely burdens respondents’ associational rights. That disagreement begins with Jones. Petitioners argue that the 1-872 primary is indistinguishable from the alternative Jones suggested would be constitutional. In Jones we noted that a nonpartisan blanket primary, where the top two votegetters proceed to the general election regardless of their party, was a less restrictive alternative to California’s system because such a primary does not nominate candidates. 530 U. S., at 585-586 (The nonpartisan blanket primary “has all the characteristics of the partisan blanket primary, save the constitutionally crucial one: Primary voters are not choosing a party’s nominee”). Petitioners are correct that we assumed that the nonpartisan primary we described in Jones would be constitutional. But that is not dispositive here because we had no occasion in Jones to determine whether a primary system that indicates each candidate’s party preference on the ballot, in effect, chooses the parties’ nominees.

That question is now squarely before us. Respondents argue that 1-872 is unconstitutional under Jones because it has the same “constitutionally crucial” infirmity that doomed California’s blanket primary: It allows primary voters who are unaffiliated with a party to choose the party’s nominee. Respondents claim that candidates who progress to the general election under 1-872 will become the de facto nominees *453of the parties they prefer, thereby violating the parties’ right to choose their own standard bearers, see Timmons, supra, at 359, and altering their messages. They rely on our statement in Jones reaffirming “the special place the First Amendment reserves for, and the special protection it accords, the process by which a political party ‘select[s] a standard bearer who best represents the party’s ideologies and preferences.’” Jones, 530 U.S., at 575 (quoting Eu, supra, at 224).

The flaw in this argument is that, unlike the California primary, the 1-872 primary does not, by its terms, choose parties’ nominees. The essence of nomination — the choice of a party representative — does not occur under 1-872. The law never refers to the candidates as nominees of any party, nor does it treat them as such. To the contrary, the election regulations specifically provide that the primary “does not serve to determine the nominees of a political party but serves to winnow the number of candidates to a final list of two for the general election.” App. 606, Wash. Admin. Code §434-262-012. The top two candidatés from the primary election proceed to the general election regardless of their party preferences. Whether parties nominate their own candidates outside the state-run primary is simply irrelevant. In fact, parties may now nominate candidates by whatever mechanism they choose because 1-872 repealed Washington’s prior regulations governing party nominations.7

*454Respondents counter that, even if the 1-872 primary does not actually choose parties’ nominees, it nevertheless burdens their associational rights because voters will assume that candidates on the general election ballot are the nominees of their preferred parties. This brings us to the heart of respondents’ case — and to the fatal flaw in their argument. At bottom, respondents’ objection to 1-872 is that voters will be confused by candidates’ party-preference designations. Respondents’ arguments are largely variations on this theme. Thus, they argue that even if voters do not assume that candidates on the general election ballot are the nominees of their parties, they will at least assume that the parties associate with, and approve of, them. This, they say, compels them to associate with candidates they do not endorse, alters the messages they wish to convey, and forces them to engage in counterspeech to disassociate themselves from the candidates and their positions on the issues.

We reject each of these contentions for the same reason: They all depend, not on any facial requirement of 1-872, but on the possibility that voters will be confused as to the meaning of the party-preference designation. But respondents’ assertion that voters will misinterpret the party-preference designation is sheer speculation. It “depends upon the belief that voters can be ‘misled’ by party labels. But ‘[o]ur cases reflect a greater faith in the ability of individual voters to inform themselves about campaign issues.’” Tashjian, 479 U. S., at 220 (quoting Anderson, swpra, at 797). There is simply no basis to presume that a well-informed electorate will interpret a candidate’s party-preference designation to mean that the candidate is the party’s chosen nominee or representative or that the party associates with or approves of the candidate. See New York State Clubn Assn., Inc. v. City of New York, 487 U. S. 1, 13-14 (1988) *455(rejecting a facial challenge to a law regulating club membership and noting that “[w]e could hardly hold otherwise on the record before us, which contains no specific evidence on the characteristics of any club covered by the [l]aw”). This strikes us as especially true here, given that it was the voters of Washington themselves, rather than their elected representatives, who enacted 1-872.

Of course, it is possible that voters will misinterpret the candidates’ party-preference designations as reflecting endorsement by the parties. But these cases involve a facial challenge, and we cannot strike down 1-872 on its face based on the mere possibility of voter confusion. See Yazoo, 226 U. S., at 219 (“[T]his court must deal with the case in hand and not with imaginary ones”); Pullman Co. v. Knott, 235 U. S. 23, 26 (1914) (A statute “is not to be upset upon hypothetical and unreal possibilities, if it would be good upon the facts as they are”). Because respondents brought their suit as a facial challenge, we have no evidentiary record against which to assess their assertions that voters will be confused. See Timmons, 520 U. S., at 375-376 (Stevens, J., dissenting) (rejecting judgments based on “imaginative theoretical sources of voter confusion” and “entirely hypothetical” outcomes). Indeed, because 1-872 has never been implemented, we do not even have ballots indicating how party preference will be displayed. It stands to reason that whether voters will be confused by the party-preference designations will depend in significant part on the form of the ballot. The Court of Appeals assumed that the ballot would not place abbreviations like “ ‘D’ ” and “ ‘R,’ ” or “ ‘Dem.’ ” and “‘Rep.’” after the names of candidates, but would instead “clearly state that a particular candidate ‘prefers’ a particular party.” 460 F. 3d, at 1121, n. 20. It thought that even such a clear statement did too little to eliminate the risk of voter confusion.

But we see no reason to stop there. As long as we are speculating about the form of the ballot — and we can do no *456more than speculate in this facial challenge — we must, in fairness to the voters of the State of Washington who enacted 1-872 and in deference to the executive and judicial officials who are charged with implementing it, ask whether the ballot could conceivably be printed in such a way as to eliminate the possibility of widespread voter confusion and with it the perceived threat to the First Amendment. See Ayotte, 546 U. S., at 329 (noting that courts should not nullify more of a state law than necessary so as to avoid frustrating the intent of the people and their duly elected representatives); Ward v. Rock Against Racism, 491 U. S. 781, 795-796 (1989) (“ ‘[I]n evaluating a facial challenge to a state law, a federal court must... consider any limiting construction that a state court or enforcement agency has proffered’ ” (quoting Hoffman Estates v. Flipside, Hoffman Estates, Inc., 455 U. S. 489, 494, n. 5 (1982))).

It is not difficult to conceive of such a ballot. For example, petitioners propose that the actual 1-872 ballot could include prominent disclaimers explaining that party preference reflects only the self-designation of the candidate and not an official endorsement by the party. They also suggest that the ballots might note preference in the form of a candidate statement that emphasizes the candidate’s personal determination rather than the party’s acceptance of the candidate, such as “my party preference is the Republican Party.” Additionally, the State could decide to educate the public about the new primary ballots through advertising or explanatory materials mailed to voters along with their ballots.8 We are satisfied that there are a variety of ways in which the State could implement 1-872 that would eliminate any real threat of voter confusion. And without the spec*457ter of widespread voter confusion, respondents’ arguments about forced association9 and compelled speech10 fall flat.

Our conclusion that these implementations of 1-872 would be consistent with the First Amendment is fatal to respondents’ facial challenge. See Schall v. Martin, 467 U. S. 253, 264 (1984) (a facial challenge fails where “at least some” constitutional applications exist). Each of their arguments rests on factual assumptions about voter confusion, and each fails for the same reason: In the absence of evidence, we cannot assume that Washington’s voters will be misled. See Jones, 530 U. S., at 600 (Stevens, J., dissenting) (“[A]n empirically debatable assumption ... is too thin a reed to support a credible First Amendment distinction” between permissible and impermissible burdens on association). That *458factual determination must await an as-applied challenge. On its face, 1-872 does not impose any severe burden on respondents’ associational rights.

B

Because we have concluded that 1-872 does not severely burden respondents, the State need not assert a compelling interest. See Clingman, 544 U. S., at 593 (“When a state electoral provision places no heavy burden on associational rights, ‘a State’s important regulatory interests will usually be enough to justify reasonable, nondiscriminatory restrictions’ ” (quoting Timmons, 520 U. S., at 358)). The State’s asserted interest in providing voters with relevant information about the candidates on the ballot is easily sufficient to sustain 1-872. See Anderson, 460 U. S., at 796 (“There can be no question about the legitimacy of the State’s interest in fostering informed and educated expressions of the popular will in a general election”).11

Ill

Respondents ask this Court to invalidate a popularly enacted election process that has never been carried out. Immediately after implementing regulations were enacted, respondents obtained a permanent injunction against the enforcement of 1-872. The First Amendment does not require this extraordinary and precipitous nullification of the will of the people. Because 1-872 does not on its face provide for *459the nomination of candidates or compel political parties to associate with or endorse candidates, and because there is no basis in this facial challenge for presuming that candidates’ party-preference designations will confuse voters, 1-872 does not on its face severely burden respondents’ associational rights. We accordingly hold that 1-872 is facially constitutional. The judgment of the Court of Appeals is reversed.

It is so ordered.

with whom Justice Alito joins, concurring.

I share Justice Scalia’s concern that permitting a candidate to identify his political party preference on an official election ballot — regardless of whether the candidate is endorsed by the party or is even a member — may effectively force parties to accept candidates they do not want, amounting to forced association in violation of the First Amendment.

I do think, however, that whether voters perceive the candidate and the party to be associated is relevant to the constitutional inquiry. Our other forced-association cases indicate as much. In Boy Scouts of America v. Dale, 530 U. S. 640, 653 (2000), we said that Dale’s presence in the Boy Scouts would “force the organization to send a message . . . [to] the world” that the Scouts approved of homosexuality. In other words, accepting Dale would lead outsiders to believe the Scouts endorsed homosexual conduct. Largely for that reason, we held that the First Amendment entitled the Scouts to exclude Dale. Id., at 659. Similarly, in Hurley v. Irish-American Gay, Lesbian and .Bisexual Group of Boston, Inc., 515 U. S. 557 (1995), we allowed the organizers of Boston’s St. Patrick’s Day Parade to exclude a pro-gay rights float because the float’s presence in the parade might create the impression that the organizers agreed with the float sponsors’ message. See id., at 575-577.

*460Voter perceptions matter, and if voters do not actually believe the parties and the candidates are tied together, it is hard to see how the parties’ associational rights are adversely implicated. See Rumsfeld v. Forum for Academic and Institutional Rights, Inc., 547 U. S. 47,65 (2006) (rejecting law schools’ First Amendment objection to military recruiters on campus because no reasonable person would believe the “law schools agree[d] with any speech by recruiters”). After all, individuals frequently claim to favor this or that political party; these preferences, without more, do not create an unconstitutional forced association.

What makes these cases different, as Justice Scalia explains, is the place where the candidates express their party preferences: on the ballot. See post, at 465 (dissenting opinion) (noting “the special role that a state-printed ballot plays in elections”). And what makes the ballot “special” is precisely the effect it has on voter impressions. See Cook v. Gralike, 531 U. S. 510, 532 (2001) (Rehnquist, C. J., concurring in judgment) (“[T]he ballot ... is the last thing the voter sees before he makes his choice”); Anderson v. Martin, 375 U. S. 399, 402 (1964) (“[Directing the citizen’s attention to the single consideration of race . . . may decisively influence the citizen to cast his ballot along racial lines”).

But because respondents brought this challenge before the State of Washington had printed ballots for use under the new primary regime, we have no idea what those ballots will look like. Petitioners themselves emphasize that the content of the ballots in the pertinent respect is yet to be determined. See Reply Brief for Washington State Grange 2-4, 7-13.

If the ballot is designed in such a manner that no reasonable voter would believe that the candidates listed there are nominees or members of, or otherwise associated with, the parties the candidates claimed to “prefer,” the 1-872 primary system would likely pass constitutional muster. I cannot say on the present record that it would be impossible for *461the State to design such a ballot. Assuming the ballot is so designed, voters would not regard the listed candidates as “party” candidates, any more than someone saying “I like Campbell’s soup” would be understood to be associated with Campbell’s. Voters would understand that the candidate does not speak on the party’s behalf or with the party’s approval. On the other hand, if the ballot merely lists the candidates’ preferred parties next to the candidates’ names, or otherwise fails clearly to convey that the parties and the candidates are not necessarily associated, the 1-872 system would not survive a First Amendment challenge.

Justice Scalia complains that “[i]t is hard to know how to respond” to such mistaken views, post, at 467 (dissenting opinion), but he soldiers on nonetheless. He would hold that a party is burdened by a candidate’s statement of preference even if no reasonable voter believes from the ballot that the party and the candidate are associated. I take his point to be that a particular candidate’s “endorsement” of a party might alter the party’s message, and this violates the party’s freedom of association. See post, at 468 (dissenting opinion).

But there is no general right to stop an individual from saying, “I prefer this party,” even if the party would rather he not. Normally, the party protects its message in such a case through responsive speech of its own. What makes these cases different of course is that the State controls the content of the ballot, which we have never considered a public forum. See Timmons v. Twin Cities Area New Party, 520 U. S. 351, 363 (1997) (ballots are not “forums for political expression”). Neither the candidate nor the party dictates the message conveyed by the ballot. In such a case, it is important to know what the ballot actually says — both about the candidate and about the party’s association with the candidate. It is possible that no reasonable voter in Washington State will regard the listed candidates as members of, or otherwise associated with, the political parties the candidates claim to prefer. Nothing in my analysis requires the *462parties to produce studies regarding voter perceptions on this score, but I would wait to see what the ballot says before deciding whether it is unconstitutional.

Still, I agree with Justice Scalia that the history of the challenged law suggests the State is not particularly interested in devising ballots that meet these constitutional requirements. See post, at 468 (dissenting opinion). But this record simply does not allow us to say with certainty that the election system created by 1-872 is unconstitutional. Accordingly, I agree with the Court that respondents’ present challenge to the law must fail, and I join the Court’s opinion.

with whom Justice Kennedy joins, dissenting.

The electorate’s perception of a political party’s beliefs is colored by its perception of those who support the party; and a party’s defining act is the selection of a candidate and advocacy of that candidate’s election by conferring upon him the party’s endorsement. When the state-printed ballot for the general election causes a party to be associated with candidates who may not fully (if at all) represent its views, it undermines both these vital aspects of political association. The views of the self-identified party supporter color perception of the party’s message, and that self-identification on the ballot, with no space for party repudiation or party identification of its own candidate, impairs the party’s advocacy of its standard bearer. Because Washington has not demonstrated that this severe burden upon parties’ associational rights is narrowly tailored to serve a compelling interest— indeed, because it seems to me Washington’s only plausible interest is precisely to reduce the effectiveness of political parties — I would find the law unconstitutional.

I

I begin with the principles on which the Court and I agree. States may not use election regulations to undercut political *463parties’ freedoms of speech or association. See U S. Term Limits, Inc. v. Thornton, 514 U. S. 779, 833-834 (1995). Thus, when a State regulates political parties as a part of its election process, we consider “the ‘character and magnitude’” of the burden imposed on the party’s associational rights and “the extent to which the State’s concerns make the burden necessary.” Timmons v. Twin Cities Area New Party, 520 U. S. 351, 358 (1997). Regulations imposing severe burdens must be narrowly tailored to advance a compelling state interest. Ibid.

Among the First Amendment rights that political parties possess is the right to associate with the persons whom they choose and to refrain from associating with persons whom they reject. Democratic Party of United States v. Wisconsin ex rel. La Follette, 450 U. S. 107, 122 (1981). Also included is the freedom to choose and promote the “ ‘standard bearer who best represents the party’s ideologies and preferences.’” Eu v. San Francisco County Democratic Central Comm., 489 U. S. 214, 224 (1989).

When an expressive organization is compelled to associate with a person whose views the group does not accept, the organization’s message is undermined; the organization is understood to embrace, or at the very least tolerate, the views of the persons linked with them. We therefore held, for example, that a State severely burdened the right of expressive association when it required the Boy Scouts to accept an openly gay scoutmaster. The scoutmaster’s presence “would, at the very least, force the organization to send a message, both to the youth members and the world, that the Boy Scouts accepts homosexual conduct as a legitimate form of behavior.” Boy Scouts of America v. Dale, 530 U. S. 640, 653 (2000).

A political party’s expressive mission is not simply, or even primarily, to persuade voters of the party’s views. Parties seek principally to promote the election of candidates who will implement those views. See, e. g., Tashjian v. Republi*464can Party of Conn., 479 U. S. 208, 216 (1986); Storer v. Brown, 415 U. S. 724, 745 (1974); M. Hershey & P. Beck, Party Politics in America 13 (10th ed. 2003). That is achieved in large part by marking candidates with the party’s seal of approval. Parties devote substantial resources to making their names trusted symbols of certain approaches to governance. See, e.g., App. 239 (declaration of Democratic Committee Chair Paul J. Berendt); J. Aldrich, Why Parties? 48-49 (1995). They then encourage voters to cast their votes for the candidates that carry the party name. Parties’ efforts to support candidates by marking them with the party trademark, so to speak, have been successful enough to make the party name, in the words of one commentator, "the most important resource that the party possesses.” Cain, Party Autonomy and Two-Party Electoral Competition, 149 U. Pa. L. Rev. 793,804 (2001). And all evidence suggests party labels are indeed a central consideration for most voters. See, e. g., id., at 804, n. 34; Rahn, The Role of Partisan Stereotypes in Information Processing About Political Candidates, 37 Am. J. Pol. Sci. 472 (1993); Klein & Baum, Ballot Information and Voting Decisions in Judicial Elections, 54 Pol. Research Q. 709 (2001).

II

A

The State of Washington need not like, and need not favor, political parties. It is entirely free to decline running primaries for the selection of party nominees and to hold nonpartisan general elections in which party labels have no place on the ballot. See California Democratic Party v. Jones, 530 U. S. 567, 585-586 (2000). Parties would then be left to their own devices in both selecting and publicizing their candidates. But Washington has done more than merely decline to make its electoral machinery available for party building. Recognizing that parties draw support for their candidates by giving them the party imprimatur, Washing*465ton seeks to reduce the effectiveness of that endorsement by allowing any candidate to use the ballot for drawing upon the goodwill that a party has developed, while preventing the party from using the ballot to reject the claimed association or to identify the genuine candidate of its choice. This does not merely place the ballot off limits for party building; it makes the ballot an instrument by which party building is impeded, permitting unrebutted associations that the party itself does not approve.

These cases cannot be decided without taking account of the special role that a state-printed ballot plays in elections. The ballot comes into play “at the most crucial stage in the electoral process — the instant before the vote is cast.” Anderson v. Martin, 375 U. S. 399, 402 (1964). It is the only document that all voters are guaranteed to see, and it is “the last thing the voter sees before he makes his choice,” Cook v. Gralike, 531 U. S. 510, 532 (2001) (Rehnquist, C. J., concurring in judgment). Thus, we have held that a State cannot elevate a particular issue to prominence by making it the only issue for which the ballot sets forth the candidates’ positions. Id., at 525-526 (opinion of the Court). And we held unconstitutional California’s election system, which listed as the party’s candidate on the general-election ballot the candidate selected in a state-run “blanket primary” in which all citizens could determine who would be the party’s nominee. Jones, 530 U. S., at 586. It was not enough to sustain the law that the party remained free to select its preferred candidate through another process, and could denounce or campaign against the candidate carrying the party’s name on the general-election ballot. Forced association with the party on the general-election ballot was fatal. Id., at 575-577.

The Court makes much of the fact that the party names shown on the Washington ballot may be billed as mere statements of candidate “preference.” See ante, at 454-457. To be sure, the party is not itself forced to display favor for someone it does not wish to associate with, as the Boy Scouts *466were arguably forced to do by employing the homosexual scoutmaster in Dale, and as the political parties were arguably forced to do by lending their ballot endorsement as party nominee in Jones. But thrusting an unwelcome, self-proclaimed association upon the party on the election ballot itself is amply destructive of the party’s associational rights. An individual’s endorsement of a party shapes the voter’s view of what the party stands for, no less than the party’s endorsement of an individual shapes the voter’s view of what the individual stands for. That is why party nominees are often asked (and regularly agree) to repudiate the support of persons regarded as racial extremists. On Washington’s ballot, such repudiation is impossible. And because the ballot is the only document voters are guaranteed to see, and the last thing they see before casting their vote, there is “no means of replying” that “would be equally effective with the voter.” Cook, supra, at 532 (Rehnquist, C. J., concurring in judgment).

Not only is the party’s message distorted, but its goodwill is hijacked. There can be no dispute that candidate acquisition of party labels on Washington’s ballot — even if billed as self-identification — is a means of garnering the support of those who trust and agree with the party. The “I prefer the D’s” and “I prefer the R’s” will not be on the ballot for esthetic reasons; they are designed to link candidates to unwilling parties (or at least parties who are unable to express their revulsion) and to encourage voters to cast their ballots based in part on the trust they place in the party’s name and the party’s philosophy. These harms will be present no matter how Washington’s law is implemented. There is therefore “no set of circumstances” under which Washington’s law would not severely burden political parties, see United States v. Salerno, 481 U. S. 739, 745 (1987), and no good reason to wait until Washington has undermined its political parties to declare that it is forbidden to do so.

*467B

The Chief Justice would wait to see if the law is implemented in a manner that no more harms political parties than allowing a person to state that he “ ‘like[s] Campbell’s soup’ ” would harm the Campbell Soup Company. See ante, at 461 (concurring opinion). It is hard to know how to respond. First and most fundamentally, there is simply no comparison between statements of “preference” for an expressive association and statements of “preference” for soup. The robust First Amendment freedom to associate belongs only to groups “engage[d] in ‘expressive association,’” Dale, 530 U. S., at 648. The Campbell Soup Company does not exist to promote a message, and “there is only minimal constitutional protection of the freedom of commercial association,” Roberts v. United States Jay cees, 468 U. S. 609, 634 (1984) (O’Connor, J., concurring in part and concurring in judgment).

Second, I assuredly do not share The Chief Justice’s view that the First Amendment will be satisfied so long as the ballot “is designed in such a manner that no reasonable voter would believe that the candidates listed there are nominees or members of, or otherwise associated with, the parties the candidates claimed to ‘prefer.’” Ante, at 460. To begin with, it seems to me quite impossible for the ballot to satisfy a reasonable voter that the candidate is not “associated with” the party for which he has expressed a preference. He has associated himself with the party by his very expression of a preference — and that indeed is the whole purpose of allowing the preference to be expressed. If all The Chief Justice means by “associated with” is that the candidate “does not speak on the party’s behalf or with the party’s approval,” ante, at 461, none of my analysis in this opinion relies upon that misperception, nor upon the misperception that the candidate is a member or the nominee of the party. Avoiding those misperceptions is far from enough. *468Is it enough to say on the ballot that a notorious and despised racist who says that the party is his choice does not speak with the party’s approval? Surely not. His unrebutted association of that party with his views distorts the image of the party nonetheless. And the fact that the candidate who expresses a “preference” for one or another party is shown not to be the nominee of that party does not deprive him of the boost from the party’s reputation which the party wishes to confer only on its nominee. The Chief Justice claims that “the content of the ballots in the pertinent respect is yet to be determined,” ante, at 460. I disagree. We know all we need to know about the form of ballot. When pressed, Washington’s attorney general assured us at oral argument that the ballot will not say whether the party for whom the candidate expresses a preference claims or disavows him. (Of course it will not, for that would enable the party expression that it is the very object of this legislation to impair.)

And finally, while The Chief Justice earlier expresses his awareness that the special character of the ballot is what makes these cases different, ante, at 460, his Campbell’s Soup example seems to forget that. If we must speak in terms of soup, Washington’s law is like a law that encourages Oscar the Grouch (Sesame Street’s famed bad-taste resident of a garbage can) to state a “preference” for Campbell’s at every point of sale, while barring the soup company from disavowing his endorsement, or indeed using its name at all, in those same crucial locations. Reserving the most critical communications forum for statements of “preference” by a potentially distasteful speaker alters public perceptions of the entity that is “preferred”; and when this privileged connection undermines not a company’s ability to identify and promote soup but an expressive association’s ability to identify and promote its message and its standard bearer, the State treads on the constitutionally protected freedom of association.

*469The majority opinion and The Chief Justice’s concurrence also endorse a wait-and-see approach on the grounds that it is not yet evident how the law will affect voter perception of the political parties. But contrary to the Court’s suggestion, it is not incumbent on the political parties to adduce “evidence,” ante, at 457, that forced association affects their ability to advocate for their candidates and their causes. We have never put expressive groups to this perhaps-impossible task. Rather, we accept their own assessments of the matter. The very cases on which The Chief Justice relies for a wait-and-see approach, ante, at 459-460, establish as much. In Dale, for example, we did not require the Boy Scouts to prove that forced acceptance of the openly homosexual scoutmaster would distort their message. See 530 U. S., at 653 (citing La Follette, 450 U. S., at 123-124). Nor in Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston, Inc., 515 U. S. 557 (1995), did we require the organizers of the St. Patrick’s Day Parade to demonstrate that including a gay contingent in the parade would distort their message. See id., at 577. Nor in Jones, 530 U. S. 567, did we require the political parties to demonstrate either that voters would incorrectly perceive the “nominee” labels on the ballot to be the products of party elections or that the labels would change voter perceptions of the party. It does not take a study to establish that when statements of party connection are the sole information listed next to candidate names on the ballot, those statements will affect voters’ perceptions of what the candidate stands for, what the party stands for, and whom they should elect.

III

Since I conclude that Washington’s law imposes a severe burden on political parties’ associational rights, I would uphold the law only if it were “narrowly tailored” to advance “a compelling state interest.” Timmons, 520 U. S., at 358. Neither the Court’s opinion nor the State’s submission claims that Washington’s law passes such scrutiny. The State ar*470gues only that it “has a rational basis” for “providing voters with a modicum of relevant information about the candidates,” Brief for Petitioners in No. 06-730, pp. 48-49. This is the only interest the Court’s opinion identifies as well. Ante, at 458.

But “rational basis” is the least demanding of our tests; it is the same test that allows individuals to be taxed at different rates because they are in different businesses. See Allied Stores of Ohio, Inc. v. Bowers, 358. U. S. 522, 526-527 (1959). It falls far, far short of establishing the compelling state interest that the First Amendment requires. And to tell the truth, here even the existence of a rational basis is questionable. Allowing candidates to identify themselves with particular parties on the ballot displays the State’s view that adherence to party philosophy is “an important— perhaps paramount — consideration in the citizen’s choice.” Anderson, 375 U. S., at 402. If that is so, however, it seems to me irrational not to allow the party to disclaim that self-association, or to identify its own endorsed candidate.

It is no mystery what is going on here. There is no state interest behind this law except the Washington Legislature’s dislike for bright-colors partisanship, and its desire to blunt the ability of political parties with noncentrist views to endorse and advocate their own candidates. That was the purpose of the Washington system that this enactment was adopted to replace — a system indistinguishable from the one we invalidated in Jones, which required parties to allow nonmembers to join in the selection of the candidates shown as their nominees on the election ballot. (The system was held unconstitutional in Democratic Party of Washington State v. Reed, 343 F. 3d 1198 (CA9 2003).) And it is the obvious purpose of Washington legislation enacted after this law, which requires political parties to repeat a candidate’s self-declared party “preference” in electioneering communications concerning the candidate — even if the purpose of the communication is to criticize the candidate and to disavow *471any connection between him and the party. Wash. Rev. Code § 42.17.510(1) (2006); see also Wash. Admin. Code § 390-18-020 (2007).

Even if I were to assume, however, that Washington has a legitimate interest in telling voters on the ballot (above all other things) that a candidate says he favors a particular political party; and even if I were further to assume (per impossibile) that that interest was a compelling one; Washington would still have to “narrowly tailor” its law to protect that interest with minimal intrusion upon the parties’ associational rights. There has been no attempt to do that here. Washington could, for example, have permitted parties to disclaim on the general-election ballot the asserted association or to designate on the ballot their true nominees. The course the State has chosen makes sense only as an effort to use its monopoly power over the ballot to undermine the expressive activities of the political parties.

The right to associate for the election of candidates is fundamental to the operation of our political system, and state action impairing that association bears a heavy burden of justification. Washington’s electoral system permits individuals to appropriate the parties’ trademarks, so to speak, at the most crucial stage of election, thereby distorting the parties’ messages and impairing their endorsement of candidates. The State’s justification for this (to convey a “modicum of relevant information”) is not only weak but undeserving of credence. We have here a system which, like the one it replaced, does not merely refuse to assist, but positively impairs, the legitimate role of political parties. I dissent from the Court’s conclusion that the Constitution permits this sabotage.