This case presents the question whether the State of California may, consistent with the First Amendment to the United States Constitution, use a so-called "blanket" primary to determine a political party's nominee for the general election.

* Under California law, a candidate for public office has two routes to gain access to the general ballot for most state and federal elective offices. He may receive the nomination of a qualified political party by winning its primary,1 see Cal. Elec. Code Ann. §§15451, 13105(a) (West 1996); or he may file as an independent by obtaining (for a statewide race) the signatures of one percent of the State's electorate or (for other races) the signatures of three percent of the voting population of the area represented by the office in contest, see §8400.

Until 1996, to determine the nominees of qualified parties California held what is known as a "closed" partisan primary, in which only persons who are members of the political party-i.e., who have declared affiliation with that party when they register to vote, see Cal. Elec. Code Ann. §§2150, 2151 (West 1996 and Supp. 2000)-can vote on its nominee, see Cal. Elec. Code Ann. §2151 (West 1996). In 1996 the citizens of California adopted by initiative Proposition 198. Promoted largely as a measure that would "weaken" party "hard-liners" and ease the way for "moderate problem-solvers," App. 89-90 (reproducing ballot pamphlet distributed to voters), Proposition 198 changed California's partisan primary from a closed primary to a blanket primary. Under the new system, "[a]ll persons entitled to vote, including those not affiliated with any political party, shall have the right to vote - for any candidate regardless of the candidate's political affiliation." Cal. Elec. Code Ann. §2001 (West Supp. 2000); see also §2151. Whereas under the closed primary each voter received a ballot limited to candidates of his own party, as a result of Proposition 198 each voter's primary ballot now lists every candidate regardless of party affiliation and allows the voter to choose freely among them. It remains the case, however, that the candidate of each party who wins the greatest number of votes "is the nominee of that party at the ensuing general election." Cal. Elec. Code Ann. §15451 (West 1996).2

Petitioners in this case are four political parties-the California Democratic Party, the California Republican Party, the Libertarian Party of California, and the Peace and Freedom Party-each of which has a rule prohibiting persons not members of the party from voting in the party's primary.3 Petitioners brought suit in the United States District Court for the Eastern District of California against respondent California Secretary of State, alleging, inter alia, that California's blanket primary violated their First Amendment rights of association, and seeking declaratory and injunctive relief. The group Californians for an Open Primary, also respondent, intervened as a party defendant. The District Court recognized that the new law would inject into each party's primary substantial numbers of voters unaffiliated with the party. 984 F. Supp. 1288, 1298-1299 (1997). It further recognized that this might result in selection of a nominee different from the one party members would select, or at the least cause the same nominee to commit himself to different positions. Id., at 1299. Nevertheless, the District Court held that the burden on petitioners' rights of association was not a severe one, and was justified by state interests ultimately reducing to this: "enhanc[ing] the democratic nature of the election process and the representativeness of elected officials." Id., at 1301. The Ninth Circuit, adopting the District Court's opinion as its own, affirmed. 169 F.3d 646 (1999). We granted certiorari. 528 U.S. 1133 (2000).

Respondents rest their defense of the blanket primary upon the proposition that primaries play an integral role in citizens' selection of public officials. As a consequence, they contend, primaries are public rather than private proceedings, and the States may and must play a role in ensuring that they serve the public interest. Proposition 198, respondents conclude, is simply a rather pedestrian example of a State's regulating its system of elections.

We have recognized, of course, that States have a major role to play in structuring and monitoring the election process, including primaries. See Burdick v. Takushi, 504 U.S. 428, 433 (1992); Tashjian v. Republican Party of Conn., 479 U.S. 208, 217 (1986). We have considered it "too plain for argument," for example, that a State may require parties to use the primary format for selecting their nominees, in order to assure that intraparty competition is resolved in a democratic fashion. American Party of Tex. v. White, 415 U.S. 767, 781 (1974); see also Tashjian, supra, at 237 (Scalia, J., dissenting). Similarly, in order to avoid burdening the general election ballot with frivolous candidacies, a State may require parties to demonstrate "a significant modicum of support" before allowing their candidates a place on that ballot. See Jenness v. Fortson, 403 U.S. 431, 442 (1971). Finally, in order to prevent "party raiding"-a process in which dedicated members of one party formally switch to another party to alter the outcome of that party's primary-a State may require party registration a reasonable period of time before a primary election. See Rosario v. Rockefeller, 410 U.S. 752 (1973). Cf. Kusper v. Pontikes, 414 U.S. 51 (1973) (23-month waiting period unreasonable).

What we have not held, however, is that the processes by which political parties select their nominees are, as respondents would have it, wholly public affairs that States may regulate freely.4 To the contrary, we have continually stressed that when States regulate parties' internal processes they must act within limits imposed by the Constitution. See, e.g., Eu v. San Francisco County Democratic Central Comm., 489 U.S. 214 (1989); Democratic Party of United States v. Wisconsin ex rel. La Follette, 450 U.S. 107 (1981). In this regard, respondents' reliance on Smith v. Allwright, 321 U.S. 649 (1944), and Terry v. Adams, 345 U.S. 461 (1953), is misplaced. In Allwright, we invalidated the Texas Democratic Party's rule limiting participation in its primary to whites; in Terry, we invalidated the same rule promulgated by the Jaybird Democratic Association, a "self-governing voluntary club," 345 U.S., at 463. These cases held only that, when a State prescribes an election process that gives a special role to political parties, it "endorses, adopts and enforces the discrimination against Negroes," that the parties (or, in the case of the Jaybird Democratic Association, organizations that are "part and parcel" of the parties, see id., at 482 (Clark, J., concurring)) bring into the process-so that the parties' discriminatory action becomes state action under the Fifteenth Amendment. Allwright, supra, at 664; see also Terry, 345 U.S., at 484 (Clark, J., concurring); id., at 469 (opinion of Black, J.). They do not stand for the proposition that party affairs are public affairs, free of First Amendment protections-and our later holdings make that entirely clear.5 See, e.g., Tashjian, supra.

Representative democracy in any populous unit of governance is unimaginable without the ability of citizens to band together in promoting among the electorate candidates who espouse their political views. The formation of national political parties was almost concurrent with the formation of the Republic itself. See Cunningham, The Jeffersonian Republican Party, in 1 History of U.S. Political Parties 239, 241 (A. Schlesinger ed., 1973). Consistent with this tradition, the Court has recognized that the First Amendment protects "the freedom to join together in furtherance of common political beliefs," Tashjian, supra, at 214-215, which "necessarily presupposes the freedom to identify the people who constitute the association, and to limit the association to those people only," La Follette, 450 U.S., at 122. That is to say, a corollary of the right to associate is the right not to associate. " `Freedom of association would prove an empty guarantee if associations could not limit control over their decisions to those who share the interests and persuasions that underlie the association's being.' " Id., at 122, n. 22 (quoting L. Tribe, American Constitutional Law 791 (1978)). See also Roberts v. United States Jaycees, 468 U.S. 609, 623 (1984).

In no area is the political association's right to exclude more important than in the process of selecting its nominee. That process often determines the party's positions on the most significant public policy issues of the day, and even when those positions are predetermined it is the nominee who becomes the party's ambassador to the general electorate in winning it over to the party's views. See Timmons v. Twin Cities Area New Party, 520 U.S. 351, 372 (1997) (Stevens, J., dissenting) ("But a party's choice of a candidate is the most effective way in which that party can communicate to the voters what the party represents and, thereby, attract voter interest and support"). Some political parties-such as President Theodore Roosevelt's Bull Moose Party, the La Follette Progressives of 1924, the Henry Wallace Progressives of 1948, and the George Wallace American Independent Party of 1968-are virtually inseparable from their nominees (and tend not to outlast them). See generally E. Kruschke, Encyclopedia of Third Parties in the United States (1991).

Unsurprisingly, our cases vigorously affirm 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." Eu, supra, at 224 (internal quotation marks omitted). The moment of choosing the party's nominee, we have said, is "the crucial juncture at which the appeal to common principles may be translated into concerted action, and hence to political power in the community." Tashjian, 479 U.S., at 216; see also id., at 235-236 (Scalia, J., dissenting) ("The ability of the members of the Republican Party to select their own candidate - unquestionably implicates an associational freedom"); Timmons, 520 U.S., at 359 ("[T]he New Party, and not someone else, has the right to select the New Party's standard bearer" (internal quotation marks omitted)); id., at 371 (Stevens, J., dissenting) ("The members of a recognized political party unquestionably have a constitutional right to select their nominees for public office").

In La Follette, the State of Wisconsin conducted an open presidential preference primary.6 Although the voters did not select the delegates to the Democratic Party's National Convention directly-they were chosen later at caucuses of party members-Wisconsin law required these delegates to vote in accord with the primary results. Thus allowing nonparty members to participate in the selection of the party's nominee conflicted with the Democratic Party's rules. We held that, whatever the strength of the state interests supporting the open primary itself, they could not justify this "substantial intrusion into the associational freedom of members of the National Party."7 450 U.S., at 126.

California's blanket primary violates the principles set forth in these cases. Proposition 198 forces political parties to associate with-to have their nominees, and hence their positions, determined by-those who, at best, have refused to affiliate with the party, and, at worst, have expressly affiliated with a rival. In this respect, it is qualitatively different from a closed primary. Under that system, even when it is made quite easy for a voter to change his party affiliation the day of the primary, and thus, in some sense, to "cross over," at least he must formally become a member of the party; and once he does so, he is limited to voting for candidates of that party.8

The evidence in this case demonstrates that under California's blanket primary system, the prospect of having a party's nominee determined by adherents of an opposing party is far from remote-indeed, it is a clear and present danger. For example, in one 1997 survey of California voters 37 percent of Republicans said that they planned to vote in the 1998 Democratic gubernatorial primary, and 20 percent of Democrats said they planned to vote in the 1998 Republican United States Senate primary. Tr. 668-669. Those figures are comparable to the results of studies in other States with blanket primaries. One expert testified, for example, that in Washington the number of voters crossing over from one party to another can rise to as high as 25 percent, id., at 511, and another that only 25 to 33 percent of all Washington voters limit themselves to candidates of one party throughout the ballot, App. 136. The impact of voting by nonparty members is much greater upon minor parties, such as the Libertarian Party and the Peace and Freedom Party. In the first primaries these parties conducted following California's implementation of Proposition 198, the total votes cast for party candidates in some races was more than double the total number of registered party members. California Secretary of State, Statement of Vote, Primary Election, June 2, 1998, http://primary98.ss.ca.gov/Final/ Official-Results.htm; California Secretary of State, Re-port of Registration, May 1998, http://www.ss.ca.gov/ elections/elections-u.htm.

The record also supports the obvious proposition that these substantial numbers of voters who help select the nominees of parties they have chosen not to join often have policy views that diverge from those of the party faithful. The 1997 survey of California voters revealed significantly different policy preferences between party members and primary voters who "crossed over" from another party. Pl. Exh. 8 (Addendum to Mervin Field Report). One expert went so far as to describe it as "inevitable [under Proposition 198] that parties will be forced in some circumstances to give their official designation to a candidate who's not preferred by a majority or even plurality of party members." Tr. 421 (expert testimony of Bruce Cain).

In concluding that the burden Proposition 198 imposes on petitioners' rights of association is not severe, the Ninth Circuit cited testimony that the prospect of malicious crossover voting, or raiding, is slight, and that even though the numbers of "benevolent" crossover voters were significant, they would be determinative in only a small number of races.9 169 F.3d, at 656-657. But a single election in which the party nominee is selected by nonparty members could be enough to destroy the party. In the 1860 presidential election, if opponents of the fledgling Republican Party had been able to cause its nomination of a pro-slavery candidate in place of Abraham Lincoln, the coalition of intraparty factions forming behind him likely would have disintegrated, endangering the party's survival and thwarting its effort to fill the vacuum left by the dissolution of the Whigs. See generally, 1 Political Parties & Elections in the United States: An Encyclopedia 398-408, 587 (L. Maisel ed. 1991). Ordinarily, however, being saddled with an unwanted, and possibly antithetical, nominee would not destroy the party but severely transform it. "[R]egulating the identity of the parties' leaders," we have said, "may - color the parties' message and interfere with the parties' decisions as to the best means to promote that message." Eu, 489 U.S., at 231, n. 21.

In any event, the deleterious effects of Proposition 198 are not limited to altering the identity of the nominee. Even when the person favored by a majority of the party members prevails, he will have prevailed by taking somewhat different positions-and, should he be elected, will continue to take somewhat different positions in order to be renominated. As respondents' own expert concluded, "[t]he policy positions of Members of Congress elected from blanket primary states are - more moderate, both in an absolute sense and relative to the other party, and so are more reflective of the preferences of the mass of voters at the center of the ideological spectrum." App. 109 (expert report of Elisabeth R. Gerber). It is unnecessary to cumulate evidence of this phenomenon, since, after all, the whole purpose of Proposition 198 was to favor nominees with "moderate" positions. Id., at 89. It encourages candidates-and officeholders who hope to be renominated- to curry favor with persons whose views are more "centrist" than those of the party base. In effect, Proposition 198 has simply moved the general election one step earlier in the process, at the expense of the parties' ability to perform the "basic function" of choosing their own leaders. Kusper, 414 U.S., at 58.

Nor can we accept the Court of Appeals' contention that the burden imposed by Proposition 198 is minor because petitioners are free to endorse and financially support the candidate of their choice in the primary. 169 F.3d, at 659. The ability of the party leadership to endorse a candidate is simply no substitute for the party members' ability to choose their own nominee. In Eu, we recognized that party-leadership endorsements are not always effective-for instance, in New York's 1982 gubernatorial primary, Edward Koch, the Democratic Party leadership's choice, lost out to Mario Cuomo. 489 U.S., at 228, n. 18. One study has concluded, moreover, that even when the leadership-endorsed candidate has won, the effect of the endorsement has been negligible. Ibid. (citing App. in Eu v. San Francisco County Democratic Central Comm., O. T. 1988, No. 87-1269, pp. 97-98). New York's was a closed primary; one would expect leadership endorsement to be even less effective in a blanket primary, where many of the voters are unconnected not only to the party leadership but even to the party itself. In any event, the ability of the party leadership to endorse a candidate does not assist the party rank and file, who may not themselves agree with the party leadership, but do not want the party's choice decided by outsiders.

We are similarly unconvinced by respondents' claim that the burden is not severe because Proposition 198 does not limit the parties from engaging fully in other traditional party behavior, such as ensuring orderly internal party governance, maintaining party discipline in the legislature, and conducting campaigns. The accuracy of this assertion is highly questionable, at least as to the first two activities. That party nominees will be equally observant of internal party procedures and equally respectful of party discipline when their nomination depends on the general electorate rather than on the party faithful seems to us improbable. Respondents themselves suggest as much when they assert that the blanket primary system " `will lead to the election of more representative `problem solvers' who are less beholden to party officials.' " Brief for Respondents 41 (emphasis added) (quoting 169 F.3d, at 661). In the end, however, the effect of Proposition 198 on these other activities is beside the point. We have consistently refused to overlook an unconstitutional restriction upon some First Amendment activity simply because it leaves other First Amendment activity unimpaired. See, e.g., Spence v. Washington, 418 U.S. 405, 411, n. 4 (1974) (per curiam); Kusper, 414 U.S., at 58. There is simply no substitute for a party's selecting its own candidates.

In sum, Proposition 198 forces petitioners to adulterate their candidate-selection process-the "basic function of a political party," ibid.-by opening it up to persons wholly unaffiliated with the party. Such forced association has the likely outcome-indeed, in this case the intended outcome-of changing the parties' message. We can think of no heavier burden on a political party's associational freedom. Proposition 198 is therefore unconstitutional unless it is narrowly tailored to serve a compelling state interest. See Timmons, 520 U.S., at 358 ("Regulations imposing severe burdens on [parties'] rights must be narrowly tailored and advance a compelling state interest"). It is to that question which we now turn.

Respondents proffer seven state interests they claim are compelling. Two of them-producing elected officials who better represent the electorate and expanding candidate debate beyond the scope of partisan concerns-are simply circumlocution for producing nominees and nominee positions other than those the parties would choose if left to their own devices. Indeed, respondents admit as much. For instance, in substantiating their interest in "representativeness," respondents point to the fact that "officials elected under blanket primaries stand closer to the median policy positions of their districts" than do those selected only by party members. Brief for Respondents 40. And in explaining their desire to increase debate, respondents claim that a blanket primary forces parties to reconsider long standing positions since it "compels [their] candidates to appeal to a larger segment of the electorate." Id., at 46. Both of these supposed interests, therefore, reduce to nothing more than a stark repudiation of freedom of political association: Parties should not be free to select their own nominees because those nominees, and the positions taken by those nominees, will not be congenial to the majority.

We have recognized the inadmissibility of this sort of "interest" before. In Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston, Inc., 515 U.S. 557 (1995), the South Boston Allied War Veterans Council refused to allow an organization of openly gay, lesbian, and bisexual persons (GLIB) to participate in the council's annual St. Patrick's Day parade. GLIB sued the council under Massachusetts' public accommodation law, claiming that the council impermissibly denied them access on account of their sexual orientation. After noting that parades are expressive endeavors, we rejected GLIB's contention that Massachusetts' public accommodation law overrode the council's right to choose the content of its own message. Applying the law in such circumstances, we held, made apparent that its "object [was] simply to require speakers to modify the content of their expression to whatever extent beneficiaries of the law choose to alter it with messages of their own. - [I]n the absence of some further, legitimate end, this object is merely to allow exactly what the general rule of speaker's autonomy forbids." Id., at 578.

Respondents' third asserted compelling interest is that the blanket primary is the only way to ensure that disenfranchised persons enjoy the right to an effective vote. By "disenfranchised," respondents do not mean those who cannot vote; they mean simply independents and members of the minority party in "safe" districts. These persons are disenfranchised, according to respondents, because under a closed primary they are unable to participate in what amounts to the determinative election-the majority party's primary; the only way to ensure they have an "effective" vote is to force the party to open its primary to them. This also appears to be nothing more than reformulation of an asserted state interest we have already rejected-recharacterizing nonparty members' keen desire to participate in selection of the party's nominee as "disenfranchisement" if that desire is not fulfilled. We have said, however, that a "nonmember's desire to participate in the party's affairs is overborne by the countervailing and legitimate right of the party to determine its own membership qualifications." Tashjian, 479 U.S., at 215-216, n. 6 (citing Rosario v. Rockefeller, 410 U.S. 752 (1973), and Nader v. Schaffer, 417 F. Supp. 837 (Conn.), summarily aff'd, 429 U.S. 989 (1976)). The voter's desire to participate does not become more weighty simply because the State supports it. Moreover, even if it were accurate to describe the plight of the non-party-member in a safe district as "disenfranchisement," Proposition 198 is not needed to solve the problem. The voter who feels himself disenfranchised should simply join the party. That may put him to a hard choice, but it is not a state-imposed restriction upon his freedom of association, whereas compelling party members to accept his selection of their nominee is a state-imposed restriction upon theirs.

Respondents' remaining four asserted state interests-promoting fairness, affording voters greater choice, increasing voter participation, and protecting privacy-are not, like the others, automatically out of the running; but neither are they, in the circumstances of this case, compelling. That determination is not to be made in the abstract, by asking whether fairness, privacy, etc., are highly significant values; but rather by asking whether the aspect of fairness, privacy, etc., addressed by the law at issue is highly significant. And for all four of these asserted interests, we find it not to be.

The aspect of fairness addressed by Proposition 198 is presumably the supposed inequity of not permitting nonparty members in "safe" districts to determine the party nominee. If that is unfair at all (rather than merely a consequence of the eminently democratic principle that-except where constitutional imperatives intervene-the majority rules), it seems to us less unfair than permitting nonparty members to hijack the party. As for affording voters greater choice, it is obvious that the net effect of this scheme-indeed, its avowed purpose-is to reduce the scope of choice, by assuring a range of candidates who are all more "centrist." This may well be described as broadening the range of choices favored by the majority-but that is hardly a compelling state interest, if indeed it is even a legitimate one. The interest in increasing voter participation is just a variation on the same theme (more choices favored by the majority will produce more voters), and suffers from the same defect. As for the protection of privacy: The specific privacy interest at issue is not the confidentiality of medical records or personal finances, but confidentiality of one's party affiliation. Even if (as seems unlikely) a scheme for administering a closed primary could not be devised in which the voter's declaration of party affiliation would not be public information, we do not think that the State's interest in assuring the privacy of this piece of information in all cases can conceivably be considered a "compelling" one. If such information were generally so sacrosanct, federal statutes would not require a declaration of party affiliation as a condition of appointment to certain offices. See, e.g., 47 U.S.C. § 154(b)(5) ("[M]aximum number of commissioners [of the Federal Communications Commission] who may be members of the same political party shall be a number equal to the least number of commissioners which constitutes a majority of the full membership of the Commission"); 47 U.S.C. § 396(c)(1) (1994 ed., Supp. III) (no more than five members of Board of Directors of Corporation for Public Broadcasting may be of same party); 42 U.S.C. § 2000e-4(a) (no more than three members of Equal Employment Opportunity Commission may be of same party).

Finally, we may observe that even if all these state interests were compelling ones, Proposition 198 is not a narrowly tailored means of furthering them. Respondents could protect them all by resorting to a nonpartisan blanket primary. Generally speaking, under such a system, the State determines what qualifications it requires for a candidate to have a place on the primary ballot-which may include nomination by established parties and voter-petition requirements for independent candidates. Each voter, regardless of party affiliation, may then vote for any candidate, and the top two vote getters (or however many the State prescribes) then move on to the general election. This system has all the characteristics of the partisan blanket primary, save the constitutionally crucial one: Primary voters are not choosing a party's nominee. Under a nonpartisan blanket primary, a State may ensure more choice, greater participation, increased "privacy," and a sense of "fairness"-all without severely burdening a political party's First Amendment right of association.

* * *

Respondents' legitimate state interests and petitioners' First Amendment rights are not inherently incompatible. To the extent they are in this case, the State of California has made them so by forcing political parties to associate with those who do not share their beliefs. And it has done this at the "crucial juncture" at which party members traditionally find their collective voice and select their spokesman. Tashjian, 479 U.S., at 216. The burden Proposition 198 places on petitioners' rights of political association is both severe and unnecessary. The judgment for the Court of Appeals for the Ninth Circuit is reversed.

It is so ordered.

Justice Kennedy, concurring.

Proposition 198, the product of a statewide popular initiative, is a strong and recent expression of the will of California's electorate. It is designed, in part, to further the object of widening the base of voter participation in California elections. Until a few weeks or even days before an election, many voters pay little attention to campaigns and even less to the details of party politics. Fewer still participate in the direction and control of party affairs, for most voters consider the internal dynamics of party organization remote, partisan, and of slight interest. Under these conditions voters tend to become disinterested, and so they refrain from voting altogether. To correct this, California seeks to make primary voting more responsive to the views and preferences of the electorate as a whole. The results of California's blanket primary system may demonstrate the efficacy of its solution, for there appears to have been a substantial increase in voter interest and voter participation. See Brief for Respondents 45-46.

Encouraging citizens to vote is a legitimate, indeed essential, state objective; for the constitutional order must be preserved by a strong, participatory democratic process. In short, there is much to be said in favor of California's law; and I might find this to be a close case if it were simply a way to make elections more fair and open or addressed matters purely of party structure.

The true purpose of this law, however, is to force a political party to accept a candidate it may not want and, by so doing, to change the party's doctrinal position on major issues. Ante, at 14. From the outset the State has been fair and candid to admit that doctrinal change is the intended operation and effect of its law. See, e.g., Brief for Respondents 40, 46. It may be that organized parties, controlled-in fact or perception-by activists seeking to promote their self-interest rather than enhance the party's long term support, are shortsighted and insensitive to the views of even their own members. A political party might be better served by allowing blanket primaries as a means of nominating candidates with broader appeal. Under the First Amendment's guarantee of speech through free association, however, this is an issue for the party to resolve, not for the State. Political parties advance a shared political belief, but to do so they often must speak through their candidates. When the State seeks to direct changes in a political party's philosophy by forcing upon it unwanted candidates and wresting the choice between moderation and partisanship away from the party itself, the State's incursion on the party's associational freedom is subject to careful scrutiny under the First Amendment. For these reasons I agree with the Court's opinion.

I add this separate concurrence to say that Proposition 198 is doubtful for a further reason. In justification of its statute California tells us a political party has the means at hand to protect its associational freedoms. The party, California contends, can simply use its funds and resources to support the candidate of its choice, thus defending its doctrinal positions by advising the voters of its own preference. To begin with, this does not meet the parties' First Amendment objection, as the Court well explains. Ante, at 13. The important additional point, however, is that, by reason of the Court's denial of First Amendment protections to a political party's spending of its own funds and resources in cooperation with its preferred candidate, see Colorado Republican Federal Campaign Comm. v. Federal Election Comm'n, 518 U.S. 604 (1996), the Federal Government or the State has the power to prevent the party from using the very remedy California now offers up to defend its law.

Federal campaign finance laws place strict limits on the manner and amount of speech parties may undertake in aid of candidates. Of particular relevance are limits on coordinated party expenditures, which the Federal Election Campaign Act of 1971 deems to be contributions subject to specific monetary restrictions. See 90 Stat. 488, 2 U.S.C. § 441a(a)(7)(B)(i) ("[E]xpenditures made by any person in cooperation, consultation, or concert, with, or at the request or suggestion of, a candidate, his authorized political committees, or their agents, shall be considered to be a contribution to such candidate"). Though we invalidated limits on independent party expenditures in Colorado Republican, the principal opinion did not question federal limits placed on coordinated expenditures. See 518 U.S., at 624-625 (opinion of Breyer, J.). Two Justices in dissent said that "all money spent by a political party to secure the election of its candidate" would constitute coordinated expenditures and would have upheld the statute as applied in that case. See id., at 648 (opinion of Stevens, J.). Thus, five Justices of the Court subscribe to the position that Congress or a State may limit the amount a political party spends in direct collaboration with its preferred candidate for elected office.

In my view, as stated in both Colorado Republican, supra, at 626 (opinion concurring in judgment and dissenting in part), and in Nixon v. Shrink Missouri Government PAC, 528 U.S. ---, --- (2000) (dissenting opinion), these recent cases deprive political parties of their First Amendment rights. Our constitutional tradition is one in which political parties and their candidates make common cause in the exercise of political speech, which is subject to First Amendment protection. There is a practical identity of interests between parties and their candidates during an election. Our unfortunate decisions remit the political party to use of indirect or covert speech to support its preferred candidate, hardly a result consistent with free thought and expression. It is a perversion of the First Amendment to force a political party to warp honest, straightforward speech, exemplified by its vigorous and open support of its favored candidate, into the covert speech of soft money and issue advocacy so that it may escape burdensome spending restrictions. In a regime where campaign spending cannot otherwise be limited-the structure this Court created on its own in Buckley v. Valeo, 424 U.S. 1 (1976) (per curiam)-restricting the amounts a political party may spend in collaboration with its own candidate is a violation of the political party's First Amendment rights.

Were the views of those who would uphold both California's blanket primary system and limitations on coordinated party expenditures to become prevailing law, the State could control political parties at two vital points in the election process. First, it could mandate a blanket primary to weaken the party's ability to defend and maintain its doctrinal positions by allowing nonparty members to vote in the primary. Second, it could impose severe restrictions on the amount of funds and resources the party could spend in efforts to counteract the State's doctrinal intervention. In other words, the First Amendment injury done by the Court's ruling in Colorado Republican would be compounded were California to prevail in the instant case.

When the State seeks to regulate a political party's nomination process as a means to shape and control political doctrine and the scope of political choice, the First Amendment gives substantial protection to the party from the manipulation. In a free society the State is directed by political doctrine, not the other way around. With these observations, I join the opinion of the Court.

Justice Stevens, with whom Justice Ginsburg joins as to Part I, dissenting.

Today the Court construes the First Amendment as a limitation on a State's power to broaden voter participation in elections conducted by the State. The Court's holding is novel and, in my judgment, plainly wrong. I am convinced that California's adoption of a blanket primary pursuant to Proposition 198 does not violate the First Amendment, and that its use in primary elections for state offices is therefore valid. The application of Proposition 198 to elections for United States Senators and Representatives, however, raises a more difficult question under the Elections Clause of the United States Constitution, Art. I, §4, cl. 1. I shall first explain my disagreement with the Court's resolution of the First Amendment issue and then comment on the Elections Clause issue.

* A State's power to determine how its officials are to be elected is a quintessential attribute of sovereignty. This case is about the State of California's power to decide who may vote in an election conducted, and paid for, by the State.1 The United States Constitution imposes constraints on the States' power to limit access to the polls, but we have never before held or suggested that it imposes any constraints on States' power to authorize additional citizens to participate in any state election for a state office. In my view, principles of federalism require us to respect the policy choice made by the State's voters in approving Proposition 198.

The blanket primary system instituted by Proposition 198 does not abridge "the ability of citizens to band together in promoting among the electorate candidates who espouse their political views." Ante, at 6.2 The Court's contrary conclusion rests on the premise that a political party's freedom of expressive association includes a "right not to associate," which in turn includes a right to exclude voters unaffiliated with the party from participating in the selection of that party's nominee in a primary election. Ante, at 6-7. In drawing this conclusion, however, the Court blurs two distinctions that are critical: (1) the distinction between a private organization's right to define itself and its messages, on the one hand, and the State's right to define the obligations of citizens and organiza- tions performing public functions, on the other; and (2) the distinction between laws that abridge participation in the political process and those that encourage such participation.

When a political party defines the organization and composition of its governing units, when it decides what candidates to endorse, and when it decides whether and how to communicate those endorsements to the public, it is engaged in the kind of private expressive associational activity that the First Amendment protects. Timmons v. Twin Cities Area New Party, 520 U.S. 351, 354-355, n. 4, 359 (1997) (recognizing party's right to select its own standard-bearer in context of minor party that selected its candidate through means other than a primary); id., at 371 (Stevens, J., dissenting); Eu v. San Francisco County Democratic Central Comm., 489 U.S. 214 (1989); Democratic Party of United States v. Wisconsin ex rel. La Follette, 450 U.S. 107, 124 (1981) ("A political party's choice among the various ways of determining the makeup of a State's delegation to the party's national convention is protected by the Constitution"); Cousins v. Wigoda, 419 U.S. 477, 491 (1975) ("Illinois' interest in protecting the integrity of its electoral process cannot be deemed compelling in the context of the selection of delegates to the National Party Convention" (emphasis added)).3 A political party could, if a majority of its members chose to do so, adopt a platform advocating white supremacy and opposing the election of any non-Caucasians. Indeed, it could decide to use its funds and oratorical skills to support only those candidates who were loyal to its racist views. Moreover, if a State permitted its political parties to select their candidates through conventions or caucuses, a racist party would also be free to select only candidates who would adhere to the party line.

As District Judge Levi correctly observed in an opinion adopted by the Ninth Circuit, however, the associational rights of political parties are neither absolute nor as comprehensive as the rights enjoyed by wholly private associations. 169 F.3d 646, 654-655 (1999); cf. Timmons, 520 U.S., at 360 (concluding that while regulation of endorsements implicates political parties' internal affairs and core associational activities, regulation of access to election ballot does not); La Follette, 450 U.S., at 120-121 (noting that it "may well be correct" to conclude that party associational rights are not unconstitutionally infringed by state open primary); id., at 131-132 (Powell, J., dissenting) (concluding that associational rights of major political parties are limited by parties' lack of defined ideological orientation and political mission). I think it clear-though the point has never been decided by this Court-"that a State may require parties to use the primary format for selecting their nominees." Ante, at 4. The reason a State may impose this significant restriction on a party's associational freedoms is that both the general election and the primary are quintessential forms of state action.4 It is because the primary is state action that an organization-whether it calls itself a political party or just a "Jaybird" association-may not deny non-Caucasians the right to participate in the selection of its nominees. Terry v. Adams, 345 U.S. 461 (1953); Smith v. Allwright, 321 U.S. 649, 663-664 (1944). The Court is quite right in stating that those cases "do not stand for the proposition that party affairs are [wholly] public affairs, free of First Amendment protections." Ante, at 6. They do, however, stand for the proposition that primary elections, unlike most "party affairs," are state action.5 The protections that the First Amendment affords to the "internal processes" of a political party, ibid., do not encompass a right to exclude nonmembers from voting in a state-required, state-financed primary election.

The so-called "right not to associate" that the Court relies upon, then, is simply inapplicable to participation in a state election. A political party, like any other association, may refuse to allow non-members to participate in the party's decisions when it is conducting its own affairs;6 California's blanket primary system does not infringe this principle. Ante, at 2-3, n. 2. But an election, unlike a convention or caucus, is a public affair. Although it is true that we have extended First Amendment protection to a party's right to invite independents to participate in its primaries, Tashjian v. Republican Party of Conn., 479 U.S. 208 (1986), neither that case nor any other has held or suggested that the "right not to associate" imposes a limit on the State's power to open up its primary elections to all voters eligible to vote in a general election. In my view, while state rules abridging participation in its elections should be closely scrutinized,7 the First Amendment does not inhibit the State from acting to broaden voter access to state-run, state-financed elections. When a State acts not to limit democratic participation but to expand the ability of individuals to participate in the democratic process, it is acting not as a foe of the First Amendment but as a friend and ally.

Although I would not endorse it, I could at least understand a constitutional rule that protected a party's associational rights by allowing it to refuse to select its candidates through state-regulated primary elections. See Marchioro v. Chaney, 442 U.S. 191, 199 (1979) ("There can be no complaint that [a] party's [First Amendment] right to govern itself has been substantially burdened by [state regulation] when the source of the complaint is the party's own decision to confer critical authority on the [party governing unit being regulated]"); cf. Tashjian, 479 U.S., at 237 (Scalia, J., dissenting) ("It is beyond my understanding why the Republican Party's delegation of its democratic choice [of candidates] to a Republican Convention [rather than a primary] can be proscribed [by the State], but its delegation of that choice to nonmembers of the Party cannot"). A meaningful "right not to associate," if there is such a right in the context of limiting an electorate, ought to enable a party to insist on choosing its nominees at a convention or caucus where non-members could be excluded. In the real world, however, anyone can "join" a political party merely by asking for the appropriate ballot at the appropriate time or (at most) by registering within a state-defined reasonable period of time before an election; neither past voting history nor the voter's race, religion, or gender can provide a basis for the party's refusal to "associate" with an unwelcome new member. See 169 F.3d, at 655, and n. 20. There is an obvious mismatch between a supposed constitutional right "not to associate" and a rule that turns on nothing more than the state-defined timing of the new associate's application for membership. See La Follette, 450 U.S., at 133 (Powell, J., dissenting) ("As Party affiliation becomes - easy for a voter to change [shortly before a particular primary election] in order to participate in [that] election, the difference between open and closed primaries loses its practical significance").

The Court's reliance on a political party's "right not to associate" as a basis for limiting a State's power to conduct primary elections will inevitably require it either to draw unprincipled distinctions among various primary configurations or to alter voting practices throughout the Nation in fundamental ways. Assuming that a registered Democrat or independent who wants to vote in the Republican gubernatorial primary can do so merely by asking for a Republican ballot, the Republican Party's constitutional right "not to associate" is pretty feeble if the only cost it imposes on that Democrat or independent is a loss of his right to vote for non-Republican candidates for other offices. Cf. ante, at 10, n. 8. Subtle distinctions of this minor import are grist for state legislatures, but they demean the process of constitutional adjudication. Or, as Justice Scalia put the matter in his dissenting opinion in Tashjian:

"The - voter who, while steadfastly refusing to register as a Republican, casts a vote in [a non-closed] Republican primary, forms no more meaningful an `association' with the Party than does the independent or the registered Democrat who responds to questions by a Republican Party pollster. If the concept of freedom of association is extended to such casual contacts, it ceases to be of any analytic use." 479 U.S., at 235.

It is noteworthy that the bylaws of each of the political parties that are petitioners in this case unequivocally state that participation in partisan primary elections is to be limited to registered members of the party only. App. 7, 15, 16, 18. Under the Court's reasoning, it would seem to follow that conducting anything but a closed partisan primary in the face of such bylaws would necessarily burden the parties' " `freedom to identify the people who constitute the association.' " Ante, at 6-7. Given that open primaries are supported by essentially the same state interests that the Court disparages today and are not as "narrow" as nonpartisan primaries, ante, at 14-18, there is surely a danger that open primaries will fare no better against a First Amendment challenge than blanket primaries have.

By the District Court's count, 3 States presently have blanket primaries, while an additional 21 States have open primaries and 8 States have semi-closed primaries in which independents may participate. 169 F.3d, at 650. This Court's willingness to invalidate the primary schemes of 3 States and cast serious constitutional doubt on the schemes of 29 others at the parties' behest is, as the District Court rightly observed, "an extraordinary intrusion into the complex and changing election laws of the States [that] - remove[s] from the American political system a method for candidate selection that many States consider beneficial and which in the uncertain future could take on new appeal and importance." Id., at 654.8

In my view, the First Amendment does not mandate that a putatively private association be granted the power to dictate the organizational structure of state-run, state-financed primary elections. It is not this Court's constitutional function to choose between the competing visions of what makes democracy work-party autonomy and discipline versus progressive inclusion of the entire electorate in the process of selecting their public officials-that are held by the litigants in this case. O'Callaghan v. State, 914 P.2d 1250, 1263 (Alaska 1996); see also Tashjian, 479 U.S., at 222-223; Luther v. Borden, 7 How. 1, 40-42 (1849). That choice belongs to the people. U.S. Term Limits, Inc. v. Thornton, 514 U.S. 779, 795 (1995).

Even if the "right not to associate" did authorize the Court to review the State's policy choice, its evaluation of the competing interests at stake is seriously flawed. For example, the Court's conclusion that a blanket primary severely burdens the parties' associational interests in selecting their standard bearers does not appear to be borne out by experience with blanket primaries in Alaska and Washington. See, e.g., 169 F.3d, at 656-659, and n. 23. Moreover, that conclusion rests substantially upon the Court's claim that "[t]he evidence before the District Court" disclosed a "clear and present danger" that a party's nominee may be determined by adherents of an opposing party. Ante, at 10. This hyperbole is based upon the Court's liberal view of its appellate role, not upon the record and the District Court's factual findings. Following a bench trial and the receipt of expert witness reports, the District Court found that "there is little evidence that raiding [by members of an opposing party] will be a factor under the blanket primary. On this point there is almost unanimity among the political scientists who were called as experts by the plaintiffs and defendants." 169 F.3d, at 656. While the Court is entitled to test this finding by making an independent examination of the record, the evidence it cites-including the results of the June 1998 primaries, ante, at 10-11, which should not be considered because they are not in the record-does not come close to demonstrating that the District Court's factual finding is clearly erroneous. Bose Corp. v. Consumers Union of United States, Inc., 466 U.S. 485, 498-501 (1984).

As to the Court's concern that benevolent crossover voting impinges on party associational interests, ante, at 11, the District Court found that experience with a blanket primary in Washington and other evidence "suggest[ed] that there will be particular elections in which there will be a substantial amount of cross-over voting - although the cross-over vote will rarely change the outcome of any election and in the typical contest will not be at significantly higher levels than in open primary states." 169 F.3d, at 657. In my view, an empirically debatable assumption about the relative number and effect of likely crossover voters in a blanket primary, as opposed to an open primary or a nominally closed primary with only a brief pre-registration requirement, is too thin a reed to support a credible First Amendment distinction. See Tashjian, 479 U.S., at 219 (rejecting State's interest in keeping primary closed to curtail benevolent crossover voting by independents given that independents could easily cross over even under closed primary by simply registering as party members).

On the other side of the balance, I would rank as "substantial, indeed compelling," just as the District Court did, California's interest in fostering democratic government by "[i]ncreasing the representativeness of elected officials, giving voters greater choice, and increasing voter turnout and participation in [electoral processes]." 169 F.3d, at 662;9 cf. Timmons, 520 U.S., at 364 ("[W]e [do not] require elaborate, empirical verification of the weightiness of the State's asserted justifications"). The Court's glib rejection of the State's interest in increasing voter participation, ante, at 17, is particularly regrettable. In an era of dramatically declining voter participation, States should be free to experiment with reforms designed to make the democratic process more robust by involving the entire electorate in the process of selecting those who will serve as government officials. Opening the nominating process to all and encouraging voters to participate in any election that draws their interest is one obvious means of achieving this goal. See Brief for Respondents 46 (noting that study presented to District Court showed higher voter turnout levels in blanket primary states than in open or closed primary states); ante, at 1 (Kennedy, J., concurring). I would also give some weight to the First Amendment associational interests of nonmembers of a party seeking to participate in the primary process,10 to the fundamental right of such nonmembers to cast a meaningful vote for the candidate of their choice, Burdick v. Takushi, 504 U.S. 428, 445 (1992) (Kennedy, J., dissenting), and to the preference of almost 60% of California voters-including a majority of registered Democrats and Republicans-for a blanket primary. 169 F.3d, at 649; see Tashjian, 479 U.S., at 236 (Scalia, J., dissenting) (preferring information on whether majority of rank-and-file party members support a particular proposition than whether state party convention does so). In my view, a State is unquestionably entitled to rely on this combination of interests in deciding who may vote in a primary election conducted by the State. It is indeed strange to find that the First Amendment forecloses this decision.

The Elections Clause of the United States Constitution, Art. I, §4, cl. 1, provides that "[t]he Times, Places and Manner of holding Elections for Senators and Representatives, shall be prescribed in each State by the Legislature thereof" (emphasis added). This broad constitutional grant of power to state legislatures is "matched by state control over the election process for state offices." Tashjian, 479 U.S., at 217. For the reasons given in Part I, supra, I believe it would be a proper exercise of these powers and would not violate the First Amendment for the California Legislature to adopt a blanket primary system. This particular blanket primary system, however, was adopted by popular initiative. Although this distinction is not relevant with respect to elections for state offices, it is unclear whether a state election system not adopted by the legislature is constitutional insofar as it applies to the manner of electing United States Senators and Representatives.

The California Constitution empowers the voters of the State to propose statutes and to adopt or reject them. Art. 2, §8. If approved by a majority vote, such "initiative statutes" generally take effect immediately and may not be amended or repealed by the California Legislature unless the voters consent. Art. 2, §10. The amendments to the California Election Code that changed the state primary from a closed system to the blanket system presently at issue were the result of the voters' March 1996 adoption of Proposition 198, an initiative statute.

The text of the Elections Clause suggests that such an initiative system, in which popular choices regarding the manner of state elections are unreviewable by independent legislative action, may not be a valid method of exercising the power that the Clause vests in state "Legislature[s]." It could be argued that this reasoning does not apply in California, as the California Constitution further provides that "[t]he legislative power of this State is vested in the California Legislature -, but the people reserve to themselves the powers of initiative and referendum." Art. 4, §1. The vicissitudes of state nomenclature, however, do not necessarily control the meaning of the Federal Constitution. Moreover, the United States House of Representatives has determined in an analogous context that the Elections Clause's specific reference to "the Legislature" is not so broad as to encompass the general "legislative power of the State."11 Under that view, California's classification of voter-approved initiatives as an exercise of legislative power would not render such initiatives the act of the California Legislature within the meaning of the Elections Clause. Arguably, therefore, California's blanket primary system for electing United States Senators and Representatives is invalid. Because the point was neither raised by the parties nor discussed by the courts below, I reserve judgment on it. I believe, however, that the importance of the point merits further attention.

* * *

For the reasons stated in Part I of this opinion, as well as those stated more fully in the District Court's excellent opinion, I respectfully dissent.