Full opinion text
STRAUB, Circuit Judge. This case requires us to peer inside New York State’s political clubhouses and determine whether party leaders have arrogated to themselves a choice that belongs to the people. The task falls to us by way of interlocutory appeal. Specifically, defendants-appellants appeal from the grant of a preliminary injunction by the District Court for the Eastern District of New York (John Gleeson, Judge). In its opinion and order, the District Court found a clear likelihood that New York State’s process for nominating Supreme Court Justices violates the First Amendment rights of plaintiffs-appellees, who consist of judicial candidates, Republican and Democratic voters from across the state, and the non-profit group Common Cause/NY. Accordingly, the Court preliminarily enjoined defendants-appellants New York State Board of Elections and its commissioners from enforcing the statutory provisions that regulate the nominating process. Although the Court declined to order the State Legislature to enact a new nominating system, it required that nominations for the office of Supreme Court Justice proceed by primary election until the Legislature enacts a new scheme of its own accord. The District Court then stayed its order until after this year’s election cycle, scheduled to conclude in November. The precise issues presented are whether the District Court exceeded its discretion in (1) finding a clear likelihood that New York State’s system for nominating its Supreme Court Justices violates the First Amendment, and (2) remedying that violation by (a) facially enjoining the relevant statutory provisions and (b) requiring that Supreme Court Justice nominations be settled through primary elections until the State Legislature enacts corrective legislation. We hold that the District Court acted within its allowable discretion on all scores. Given the number of issues involved, we set out the following table of contents: BACKGROUND I. New York State’s Electoral Scheme In the country’s other 49 states, the term “Supreme Court Justice” might signify a member of the highest appellate court. In New York, however, that term denotes a judge of the State’s general jurisdiction trial court. Apart from peculiar terminology, New York employs a method of selecting its Supreme Court Justices that is unique in the nation. In 1846, New York amended its constitution to require that Supreme Court Justices be popularly elected. In particular, Article VI, section 6 of the state constitution provides, “The justices of the supreme court shall be chosen by the electors of the judicial district in which they are to serve.” N.Y. Const, art. VI, § 6(c). Since adopting that constitutional provision, New York has experimented with different methods of effecting its guarantee of public choice. At first, the State’s political parties employed conventions to designate their Supreme Court nominees. That all changed in 1911, when the State Legislature, buoyed by a wave of progressive politics, provided for nominations by primary election. However, over the next decade, concern grew that bare-knuckled primary elections dissuaded qualified candidates from seeking these significant judicial positions. As to those brave enough to enter the contest, observers worried that the need to raise large sums of money might compromise their independence, or at least appear to do so, and lodge effective control of the nominating process in the hands of political bosses who directed their party’s large election apparatus. As a result of those concerns, in 1921 New York recast the electoral process for Supreme Court Justices. The Legislature did not entirely dispense with primary elections. Instead, it enacted a three-part scheme that combines a primary election, a nominating convention, and a general election. During the first phase, the State holds a primary election at which rank- and-file party members elect judicial delegates. N.Y. Elec. L. §§ 6-106, -124. Next, those delegates attend a convention at which they select their party’s nominees. N.Y. Elec. L. §§ 6-106, -124, -158. The individual so chosen automatically receives a place on the general election ballot. N.Y. Elec. L. § 7-116(1). Last, the State holds a general election at which Justices are elected. N.Y. Elec. L. § 8-100(l)(c). Having described the basic form of the judicial election process, we next examine in greater detail the manner in which that process actually functions. In so doing, we rely on the facts found by the District Court during a comprehensive preliminary injunction hearing, and which we conclude are not clearly erroneous. That hearing spanned 13 days and included testimony from 24 witnesses. The District Court admitted 10,000 pages of documentary evidence and nearly 500 pages of proposed fact findings and legal conclusions. With the benefit of that robust record and the District Court’s findings, we next describe the reality of Supreme Court elections in present-day New York. A. The Primary Election A network of district lines defines the primary election battlefield. Each judicial candidate stands for election in a particular judicial district. New York currently is divided into 12 judicial districts. N.Y. Const, art. YI, § 6(a), (b). In turn, each judicial district encompasses several other smaller political subdivisions known as assembly districts. See generally N.Y. Const, art. Ill, § 5. Because judicial districts are so large, each one comprises at least nine assembly districts and as many as 24. The appendix to this opinion contains a map of New York State showing all county borders and assembly and judicial district lines, as well as a similar map of New York City. The appendix also contains individual maps of the First, Second, Tenth, Eleventh, and Twelfth Judicial Districts. As noted above, judicial candidates do not run in the primary election themselves. Instead, they have the option of assembling a slate of delegates to run on their behalf, with an eye toward placing those delegates at the judicial nominating convention so that they can east their votes in favor of the candidate with whom they are affiliated. N.Y. Elec. L. §§ 6-106, -124. This entire slate of delegates, however, does not run for election as one group across the entire judicial district. Instead, small subgroups of delegates stand for election within each assembly district comprising the larger judicial district. N.Y. Elec. L. § 6-124. In this fashion, the primary election consists of a series of contests between groups of delegates within each assembly district. New York allows each political party to determine how many delegate slots to allot per assembly district, but requires that this number “be substantially in accordance with the ratio, which the number of votes cast for the party candidate for the office of governor, on the line or column of the party at the last preceding election for such office, in any unit of representation, bears to the total vote cast at such election for such candidate on such line or column in the entire state.” N.Y. Elec. L. § 6-124. State law also allows each political party to provide for an equal number of “alternate delegates” who may attend the convention and vote in the place of delegates who do not attend, which, as described, below, is a common event. Id. The political parties use various formulae to set these numbers. For example, the state Democratic Party begins with a baseline of one delegate and one alternate for each assembly district. It then adds one delegate and one alternate for every 2,500 votes cast on the party’s ballot line in the previous gubernatorial election. See Rules of Democratic Party of the State of New York, art. II, § 5. The Republican Party uses a similar formula. See Rules of the New York Republican State Committee, § 18. Although these formulae allot a handful of delegates to each assembly district, the aggregate allotment across an entire judicial district is quite large. In 2004, for example, the parties allotted each judicial district at least 64 total delegates and as many as 248, including alternates. To appear on the primary ballot, the delegates must circulate designating petitions within the assembly district in which they are running. Within a span of 37 days, each slate of delegates must gather 500 valid signatures from party members residing in that assembly district. N.Y. Elec. L. §§ 6-134(4), -136(2)(i), (3). Each party member may sign only one petition. N.Y. Elec. L. § 6-134(3). Consequently, the number of available signatories shrinks each time a party member signs a designating petition. Further, because petition signatures are routinely and successfully challenged pursuant to the one-petition signature rule, among others, each delegate slate must realistically gather between 1,000 and 1,500 signatures to gain a primary ballot position. Taking the lower figure of 1,000 signatures per assembly district, in order to run a full complement of delegates, a judicial candidate must gather at least 9,000 signatures (in the judicial district with only nine assembly districts) and as many as 24,000 signatures (in the judicial district with 24 assembly districts). In addition, because each group of delegates runs in a different assembly district, the signatures must be gathered from a variety of particular subdivisions spread throughout the entire judicial district. At the close of the petitioning period, the State Board of Elections determines which assembly districts present contested delegate races. If only one group of delegates has filed designating petitions in an assembly district, then those delegates are “deemed elected.” N.Y. Elec! Law § 6-160(2). Delegates who are “deemed elected” do not appear on the primary ballot. Id. Once the delegates achieve ballot status in contested races, they may not indicate on the ballot the judicial candidate with whom they are affiliated. Accordingly, in order to run delegate slates in any useful fashion, the judicial candidate must inform the primary electorate in each assembly district of which delegates are pledged to her in that specific locale. In the Second Judicial District, for example, which encompasses Brooklyn and Staten Island, a judicial candidate who ran a slate in each assembly district would have to mount 24 different voter education campaigns. In the Fourth Judicial District, which encompasses roughly one quarter of the State’s land, a judicial candidate seeking to run a slate in each assembly district must conduct 10 different voter education campaigns across 11 different counties. Three current or former judges, including plaintiff Margarita López Torres, averred that the process described above often shuts out candidates lacking either great wealth or the benefit of a political party’s county-wide apparatus. According to those judges, the requirements of the process — recruiting large numbers of delegates and alternates, assembling different delegate slates in each assembly district, recruiting petition circulators, collecting several thousand signatures, and conducting a host of localized voter education campaigns — effectively foreclosed their ability to access the primary election phase. For example, when a former City Court Judge ran for Supreme Court Justice in the Seventh Judicial District, he found that recruiting and running slates of delegates was “not a realistic option.” To compete in the primary election, he would have had to “recruit over 55 people to represent the 11 Assembly Districts that were within [his] Judicial District ... each of whom would have had to be willing to contribute significant energy, time, and money.” Finding that he would not be able to recruit this many delegate candidates, he pared down the list to the absolute minimum of delegates he thought he needed to elect to be competitive. Even if he were able to recruit enough delegate candidates, the signature requirements were daunting: he would have to recruit “several dozen people to work full-time at gathering” at least 9,000 signatures over the course of 37 days. Although an experienced campaigner who was active in county politics for several years prior to running for the Supreme Court, the lower court judge concluded that “there was simply no way [he] could overcome these organizational and financial burdens.” An Albany City Court Judge, also an experienced campaigner, agreed. After attempting to recruit, organize, and run slates of delegates who were independent of the party leadership in the Third Judicial District, he concluded that the process presented “insurmountable structural and practical barriers.” Like those two judges, Civil Court Judge Margarita López Torres found that the various delegate and petitioning requirements created impossibly high entry barriers for candidates lacking institutional support — even for those who possessed significant public support. López Torres was an experienced and successful campaigner who twice won countywide election to the Civil Court in Brooklyn — no simple feat considering that Brooklyn includes several million people of myriad racial, ethnic, religious, and socioeconomic backgrounds. In one recent election for Civil Court Judge, she received over 200,000 votes — more than any Democratic candidate for Supreme Court Justice received in Brooklyn that year. Yet López Torres, who sought the Democratic nomination for Supreme Court Justice on several occasions, found that the “burdens and barriers to organizing such [delegate] campaigns are truly insurmountable” for those candidates who lack the “county party leaders’ backing.” Faced with those barriers in her 1998 campaign for Supreme Court Justice, she realized “that [she] had no realistic chance” to fulfill the primary election balloting requirements despite her substantial public support. Accordingly, she “ended [her] bid for the [Democratic] nomination” for Supreme Court Justice. While these regulations effectively prohibit candidates who lack the support of the party leadership from putting their slates of delegates on the primary election ballot, candidates backed by the local party leadership easily navigate the primary system with the benefit of the party’s preexisting apparatus. Within each county, the party leadership consists of the county leader, who chairs the party’s county committee, and a group of assembly district leaders, who serve as members of the party’s state committee and may also serve as members of the party’s executive county committee. Although there is some variation between parties and counties, typically two district leaders — one male, one female — are elected from each assembly district within the county, so the party’s network is spread throughout the region. See N.Y. Elec. L. §§ 2-102, -104; Rules of Democratic Party of the State of New York, art. II, §§ 1-2; Rules of the New York Republican State Committee § 1; Rules and Regs, of the Democratic Party of the County of New York, art. II, §§ 1-4; art. Ill, § 1. We refer to candidates backed by a party’s county leaders as “party-backed candidates.” This party leadership recruits judicial delegate candidates, alternate candidates, and petition circulators. The party then simply includes its delegate candidates on the omnibus designating petitions it circulates during every primary election cycle, which may include candidates for the State Legislature, lower courts, and even Congress. In that way, the party leadership ensures that its group of loyal, handpicked delegate candidates achieves ballot status as a matter of course. Defendants’ own expert witness, New York City Board of Elections Commissioner Douglas Kellner, testified that the primary system is designed to produce this remarkable disparity between “individual” candidates’ and party-backed candidates’ ability to compete. Kellner certainly is in a position to know. He served as Law Chair of the New York County Democratic Party, has attended “every judicial nominating convention of the Democratic Party in the First Judicial District since 1976,” and often operated as the county leader’s right-hand man at those conventions. According to Kellner, regardless of the fact that the electoral scheme ostensibly provides for an open primary election, “the idea that an individual candidate would go out and recruit delegate candidates and run delegates pledged to that candidate in the primary is not the system and it twists the design of the system on its head.” The process of running a slate of delegates on the primary election ballot is so beset with obstacles that nearly all candidates recognize the attempt as a fool’s errand and do not even try. In the normal course, only one slate of delegates — that supported by local party leadership — even files a designating petition. The uncontested slate is then “deemed elected” by operation of law and does not appear on the primary ballot. This kind of invisible, automatic “election” is the norm rather than the exception. Between the years 1999 and 2002, four of the State’s counties — -Albany, Nassau, Suffolk, and Tompkins — did not field one single contested delegate race in any of their assembly districts. In New York City, the situation is only slightly improved. Between 1999 and 2003 in the First, Second, Eleventh, and Twelfth Judicial Districts, only 12.7 percent of delegate elections were contested, and these contested races occurred only in portions of the judicial districts. Accordingly, in 87.3 percent of delegate races over that four-year period, voters did not even see the delegates’ names on the ballot, much less have the opportunity to vote them up or down. B. The Delegate Lobbying Period Political parties hold their judicial nominating conventions one to two weeks after the judicial delegates are elected. N.Y. Elec. L. §§ 6-124, -126, -158(5). During that interim period, any Supreme Court Justice aspirant, whether or not she sought to have delegates elected at the primary election, theoretically may lobby the delegates for support. However, the evidence established that for two reasons, a candidate who lacks the support of her party’s leadership has no actual opportunity to lobby delegates. First, the time frame for lobbying delegates is unrealistically brief. In contested delegate elections, candidates have only two weeks to lobby at least 64 delegates and as many as 248, depending on the judicial district in which they were running. Where the delegate elections are uncontested, the delegates-elect theoretically may be known as early as July, when the nominating petitions are due. As we will explain, however, at least one candidate who attempted to obtain the names of delegates prior to the primary found her effort thwarted by local party officials. More importantly, delegates do not exercise their own judgment when deciding which candidate to support. Instead, they endorse the choice of the entity with which they are affiliated and to which they are subject. As set forth above, in the case of almost all delegate slates that achieve ballot access, that entity is the local party leadership. Defense expert Kellner admitted as much: “By definition, the convention system is designed [so] that the political leadership of the party is going to designate the party’s candidates. Specifically, judicial delegates are part of the party leadership and responsive to it and make it up, you know, constitute the party leadership.” Henry Berger, the former Chairman of New York’s Commission on Judicial Conduct, as well as a former district leader and judicial delegate, agreed that there is a close organizational affiliation between the county leaders, district leaders, and judicial delegates. Berger described that county leaders exert control over their committee members, i.e., the district leaders, who in turn exert control over the delegates: “In my experience, the district leaders almost always follow the wishes of the county party chairperson when it comes to voting for Supreme Court candidates at the convention. In turn, the delegates follow the wishes of the district leaders who have selected them and support the county party chairperson’s chosen candidates.” Berger testified that this dynamic was so established in the Second and Twelfth Judicial Districts that party officials told him the names of the judicial nominees even before the delegates — who purportedly select the nominees — were even elected. A professor of sociology from the University of Washington who served as a defense expert described a process in the Eleventh and Twelfth Judicial Districts by which, according to one participant, delegates offer and unanimously confirm nominees though “they don’t even know how to pronounce their name[s].” As to the Ninth Judicial District, a longtime judicial delegate averred that because it is “practically impossible” for a candidate to field her own slate of delegates, “county party leaders control the selection of delegates and alternates to the convention.” Thus, a candidate’s “only path [to the nomination] is to obtain the support of his or her county Party chairman and then seek the blessing of the Westchester County chair.” The county leaders’ organizational affiliation with the judicial delegates is not the only reason they are able to control them so precisely. Also important is the fact that the party leadership possesses the power to doom a delegate’s political career if she should reject its choice for Supreme Court Justice. Assemblymember Herman “Denny” Farrell, who is a district leader, chairman of the New York County Democratic Committee, and State Democratic Party chairman, testified of judicial delegates that, “No one wants to get me angry, so they will not go against me until they have nothing to lose.” Similarly, a defense witness who serves as a Manhattan district leader and often served as a judicial delegate testified that “it makes no sense to alienate the [county leader] over a choice of Supreme Court candidate” because “three years down the line when we were seeking something from the county leader ... he might remember that I didn’t support him at a different time.” Asked whether he could remember a single instance of a judicial candidate receiving the nomination over Farrell’s objection, the district leader replied, “No. I can’t remember it ever happening that way.” Similarly, two judges testified that in the Third and Seventh Judicial Districts, respectively, the party leadership tapped “reliable people” to be judicial delegates, people who would “adhere to the instructions of each county chairman.” Otherwise, as the judges averred, they would be “jeopardizing their political future.” The District Court also found that county leaders do not have to issue explicit commands to control the manner in which delegates vote. Defense expert Kellner conceded that at least in some judicial districts, the “leadership of the party ... hold[s] a meeting before the convention” to “work things out,” and then makes nomination “recommendations” to the delegates. Asked whether those “recommendations” were “always followed,” Kellner replied, “Generally, yes.” Another defense expert admitted that, in the Second Judicial District, “[i]n virtually all cases, the delegates support the candidates supported by their district leader.” Similarly, plaintiff and former judicial delegate John W. Carroll testified that delegates are “told” by party leaders “this is the slate we’re supporting at the convention” and then “that delegation supports] that slate,” even though the party leadership did not expressly tell the delegation how it should vote. State Senator Martin Connor’s actions at the 2002 Democratic Convention in the Second Judicial District further illustrate the power of implicit dynamics. Connor, who was then the minority leader of the New York State Senate, chaired that convention. He testified that Assemblymem-ber Clarence Norman, Jr., who at the time was the Brooklyn county leader, had decided to confer a nomination on a person Connor believed was a “horrible” choice— unqualified and temperamentally unfit for the bench. However, Connor needed Norman’s support in order to fend off a challenge to an incumbent Democratic State Senator who was facing a difficult reelection bid in Brooklyn. Accordingly, instead of voicing his objection, Connor simply convened the convention, stepped down as chair, quickly departed, and allowed the nomination to go forward. “I long since learned the value [in] politics of the unspoken thought,” he testified. C.The Nominating Conventions The evidence showed that the conventions were perfunctory affairs at which no debate occurred. Minutes from conventions held statewide between 1990 and 2002 show that over 96 percent of nominations went uncontested. The overwhelming majority of nominations were by unanimous voice vote. From the record, it appears that a convention chair never has been challenged successfully. Not only were conventions devoid of debate and competition, they were fleeting. Over a 12-year span, conventions statewide averaged a mere 55 minutes in length. In 1996, the Second Judicial District’s convention lasted 11 minutes but yielded eight nominees. Given the pro forma nature of conventions, it appears from the record that many delegates do not believe they are worth attending. Even though a delegate’s only duty is to attend the convention and vote, and even though delegates publicly run for that privilege just two weeks before the convention, delegate absentee rates are quite high. Over the course of seven years, delegate absentee rates were generally around 25 to 32 percent, but ranged as high as 69 percent — hence the need for so many alternates. D. The General Election The final phase of the judicial electoral process is the general election. Empirical evidence showed that because one-party rule is the norm in most judicial districts, the general election is little more than ceremony. Over a 12-year period between 1990 and 2002, almost half of the State’s elections for Supreme Court Justice were entirely uncontested, meaning that only one party’s candidate appeared on the ballot. In certain judicial districts, contested elections verged on the non-existent. In the Sixth Judicial District, 91 percent of judicial elections were uncontested during the 12-year time span. In the First Judicial District, 85 percent of judicial elections were uncontested. In eight of the State’s 12 judicial districts, more than half the elections were uncontested, which left 62 percent of the State’s voters with no choice to make on the November ballot. E. The Candidacy of Margarita Ló-pez Torres The experiences of plaintiff López Torres illustrate the manner in which the electoral scheme described above operates to exclude challenger candidates, even those with significant public support. Originally backed by the Kings County Democratic Committee, López Torres won election as a Brooklyn Civil Court Judge in 1992. Soon after her election, two high-ranking committee members directed her to hire a person of their choosing as her law secretary. Given the large caseloads that Civil Court judges must manage, the position of law secretary is a significant one. A law secretary conducts legal research, assists in drafting opinions and orders, schedules cases, and conferences with attorneys. The two party leaders who prevailed upon López Torres were Norman, who was the committee’s chair and thus the county party’s leader, and Assemblymember Vito Lopez, who was a district leader and a member of the county’s executive committee. Through Vito Lopez’s “avid sponsorship,” López Torres had secured the county party’s support for her successful Civil Court candidacy. Norman and Lopez expressed their directive through a letter authored by the Secretary of the Kings County Democratic Committee. That letter set forth that “Clarence and Vito” wished the Secretary to “refer this wonderful gentleman to you as your Law Secretary” and that López Torres should “obtain the necessary paperwork for his employment.” The resume of the favored appointee was attached. López Torres interviewed the prospect and contacted his prior employer, a Brooklyn Supreme Court Justice for whom the prospect had served as law secretary. The Justice told López Torres that the law secretary’s work was “mediocre” and that “he had spent an enormous amount of time on the phone doing political work.” López Torres hired someone else instead. An “extremely upset” Norman later called López Torres and chastised her for refusing the hire. Norman told her that she “did not ‘understand the way it works.’ ” Attorneys such as the one he recommended, he explained, “work hard for the Democratic Party’s political clubs to get candidates elected” and the law secretary “job is a way to reward them.” He demanded that she fire the person whom she had hired on the merits, and hire a person of the party leadership’s choosing. López Torres refused. Some day, Norman warned her, she “would want to become a Supreme Court Justice and ... the party leaders would not forget this.” He told her that “without the ‘County’s’ support,” her Supreme Court nomination “will not happen.” Assemblymember Vito Lopez also confronted her. He told her that her refusal to effect the hire was “an embarrassment to him” because he supported her campaign. He demanded that López Torres “make it right” by “hiring the person referred by ‘County,’ ” but she again refused. Through an intermediary, he later offered her a chance to redeem herself: if she hired his daughter as her law secretary, he would secure her nomination as the Democratic Party’s candidate for Supreme Court Justice. She declined the proposition. As Norman had anticipated, López Torres eventually did aspire to the State Supreme Court bench. She contacted Norman, who requested that they meet at one of Brooklyn’s busiest political salt licks, Junior’s Restaurant. When López Torres informed Norman that she desired the party’s nomination for a vacant Supreme Court Justice position in the Second Judicial District, he predictably declined to support her because her refusal to hire his favored candidate for law secretary was a “serious breach of protocol.” Nonetheless, she entered her name for consideration at the upcoming judicial convention. In an “urgent phone call,” Norman demanded that she withdraw. Her continued candidacy, he claimed, was a political affront to him, and running in an open convention “was not the way it works.” She again rebuffed him. At the convention, not a single delegate proposed her nomination. Undeterred, López Torres sought her party’s nomination the following year. The Judicial Screening Committee — a body subject to the control of the county party that ostensibly examines the qualifications of judicial candidates and makes recommendations to the party — interviewed her. However, when she requested the committee’s report from Norman and its chairman, Jerome Karp, neither would disclose it. She considered running her own slate of judicial delegates, but realized that such an effort was “impossible” as a practical matter. “There is simply no way I could ever overcome the[] organizational and financial burdens to place delegate candidates on the ballot,” she concluded. Without any realistic hope of running her own delegates or receiving Norman’s blessing, she withdrew from consideration. Four years later, López Torres yet again sought her party’s nomination. Karp informed her that the screening committee would interview only those candidates that Norman referred; Norman, in turn, refused to make the reference. Over the next few months, López Torres pressed Norman for a referral to the committee, but he repeatedly told her that he opposed her candidacy because she had been “disloyal.” Since she was unable to run her own contingent of delegates and the screening committee would not even evaluate her, López Torres’ candidacy effectively ended. That same year, López Torres ran for reelection to the Civil Court bench. The Kings County Democratic Committee — by now overtly hostile to one of its own elected officials — ran a candidate against her in the primary election. Facing the party-backed candidate in an open primary, Ló-pez Torres prevailed. In the general election, she received over 200,000 votes— more than any of the Democratic candidates for Supreme Court Justice received in Brooklyn that year. In 2003, the persistent López Torres again sought her party’s nomination for Supreme Court Justice. This time, Norman agreed to forward her name to the screening committee. Regardless, in a meeting with López Torres on June 6 he declined to support her because of her past defiance. He also stated that she lacked sufficient support. López Torres pointed out that the prior year she defeated the county party’s Civil Court candidate in the primary election and garnered more than 200,000 votes to win the general election. Norman responded that “County” would support only those candidates who supported “County” in turn. López Torres nonetheless attempted to lobby convention delegates. Beginning in March, she wrote repeatedly to the Kings County Democratic Committee requesting the date, time, and place of the convention; the names of the delegates; and whether she could address the delegates at the convention. In September, the committee’s Executive Director, Jeffrey C. Feld-man, finally responded. He sneered that “a learned jurist, such as yourself, [should] be well aware that [the delegates] stand for independent election in the Primary Election, yet to be held.” No list therefore existed “anywhere in the world.” As to the prospect of addressing the delegates, Feldman wrote: “While I am neither an attorney nor a graduate of law school, I suffer from the innocent belief that the floor of the Convention is open, only, to elected Delegates and their successors. I am not aware of any Convention in my thirty (30) years of attendance, which permitted a non-accredited member to be accorded the privilege of the floor .... ” He closed his letter by chastising her for the alleged improper use of a fax machine “in violation of Federal Communication Commission regulations.” At the convention, two delegates unsuccessfully attempted to nominate López Torres. After that effort failed, the delegates nominated the slate of candidates that Norman had endorsed. Subsequently, a Brooklyn district leader, who also served as a judicial delegate, penned a letter to the county’s other district leaders. He wrote that although Ló-pez Torres was “highly qualified,” he voted against her nomination because she was “an ingrate” who had offended party leadership. Her sin, he explained, was that she “courted Vito Lopez to support her for Civil Court, but then decided she didn’t need him anymore and denied his daughter a job.” Years of careful study by a number of groups whose reports were in evidence established that López Torres’ experience was no anomalous political mugging. In 2003, New York State’s Chief Judge, Judith S. Kaye, created the Commission to Promote Public Confidence in Judicial Elections and charged it with determining ways to improve voter participation in the judicial election process. Chief Judge Kaye named as Chairman John D. Feer-ick, former Dean of Fordham Law School, and as Commissioners two active state Supreme Court Justices, three active Associate Justices of the Appellate Division, one active state Administrative Judge, and a number of other public servants and private practitioners representing each of the State’s judicial districts. After studying the nominating process statewide for more than a year, the Commission concluded that “the uncontested evidence before [It] is that across the state, the system for selecting candidates for the Supreme Court vests almost total control in the hands of local political leaders.... And in many parts of the State, being on the dominant party’s slate is tantamount to winning the election.” The Commission is hardly the only entity to reach this conclusion — it merely is the latest. Since 1944 New York’s judicial nominating system has been described as exclusionary and boss-dominated; reports and newspaper editorials from that time forward have decried an electoral practice “that mocks choice,” and criticized a system in which “voters can never know the candidates and have to accept party slates,” while the “real choice is ... left to political bosses ... who control nominations.” After studying Supreme Court elections for six years, the Fund for Modern Courts concluded that “the selection of Supreme Court justices in New York, is, by and large, a process controlled not by the voters but by political leaders.” “[T]he nomination, not the election, is the lynchpin of the judicial selection process,” the Fund set forth. “Political leaders, not voters, control judicial conventions and decide who will receive the nomination — and thus who will be the judge.” Therefore, “it is not the decision of the voters as to the relative merits of judicial candidates, but the relative strength of the political parties in the judicial districts which determines the outcome of these elections.” Less than a decade later, a Task Force on Judicial Diversity appointed by then-Governor Mario Cuomo agreed that, “In practice it is the political party leaders who have the decisive power to determine who will be nominated. Most often this nomination is tantamount to election.” “As we all know,” the Governor’s task force remarked, “our system is only nominally one of election.” II. Procedural History In March of 2004, plaintiffs brought suit against the New York State Board of Elections and its commissioners pursuant to 42 U.S.C. § 1983, claiming that New York’s electoral scheme violates the First and Fourteenth Amendments to the U.S. Constitution. In particular, plaintiffs claim that the system described above violates the First Amendment’s guarantee of political association as to Supreme Court Justice candidates and the voters wishing to support them. Plaintiffs also claim that the scheme violates the Equal Protection Clause of the Fourteenth Amendment because it places unequal burdens on the right to vote. In their complaint, plaintiffs sought a declaration that the provisions of state law providing for the electoral scheme described above, see N.Y. Elec. Law §§ 6-106, -124, -158, are unconstitutional. They also sought an injunction requiring the State Legislature to enact a new election scheme, and in the meantime requiring the State to conduct direct primary elections for the office of Supreme Court Justice. In June of 2004, plaintiffs formally moved for that relief. Shortly thereafter, the District Court granted motions to intervene permissively by defendants New York County Democratic Committee, New York Republican State Committee, Associations of New York State Supreme Court Justices in the City and State of New York, and the State Association’s President, Justice David De-marest. The District Court also granted a motion to intervene statutorily by Attorney General Eliot Spitzer. In late 2004 the District Court held a 13-day hearing on plaintiffs’ motion for a preliminary injunction, during which it admitted the evidence set forth above. In January of 2006, the Court issued a Memorandum and Order granting plaintiffs’ motion on First Amendment grounds without considering plaintiffs’ equal protection claim. See López Torres v. N.Y. State Bd. of Elections, 411 F.Supp.2d 212 (E.D.N.Y.2006). In concluding that plaintiffs demonstrated a clear likelihood of success on their First Amendment claim, the District Court made the factual findings described above. In summary, the Court found that (1) it was “virtually impossible” for a candidate lacking the support of the county party to field slates of delegates, id. at 221; (2) county leaders and district leaders “select the delegates and alternate delegates, who, without consultation or deliberation, rubber stamp the county leaders’ choices ... for Supreme Court Justice,” id. at 223; and (3) county leaders and district leaders need not issue express commands to control the manner in which delegates vote, id. at 224. Ultimately, the District Court found that “Democratic and Republican Party leaders select the nominees,” id. at 231, and that the general election is a mere “formality” at which the party leaders’ selections are confirmed, id. at 230-31. In light of those findings, the Court concluded that the electoral system effectively excluded candidates from the nominating process and thus severely burdened voters’ and candidates’- First Amendment right of association. Applying strict scrutiny, the Court next considered the four interests that defendants asserted to justify the burdens, namely, (1) protecting political parties’ right of association, (2) promoting geographic diversity, (3) promoting racial diversity, and (4) protecting judicial independence by insulating judges from the fundraising process and from public backlash due to unpopular decisions. Id. at 250-55. The Court assumed that all of these interests were compelling, but concluded that the electoral system was not narrowly tailored to serve any of them. Id. The Court thus enjoined defendants from enforcing N.Y. Elec. Law §§ 6-106, -124. Id. at 256. Based on a provision of state election law providing for primary elections as the default nominating process, the Court ordered that the State shall conduct primary elections for the office of Supreme Court Justice until the Legislature enacts a new election scheme. Id. at 255-56; see N.Y. Elec. L. § 6-110 (“All other party nominations of candidates for offices to be filled at a general election, except as provided herein, shall be made at the primary election.”). The Court declined, however, to order the Legislature to take any action. López Torres, 411 F.Supp.2d at 255-56. By later order, the District Court stayed its injunction until after the 2006 general election, scheduled to conclude with the general election on November 7. Defendants appealed the District Court’s January 2006 order, and upon the motion of all parties, we set an expedited briefing schedule. DISCUSSION We first address the District Court’s ruling on plaintiffs’ First Amendment claim, and then the District Court’s choice of remedy. I. The District Court Did Not Exceed Its Allowable Discretion in Concluding that Plaintiffs Demonstrated a Clear Likelihood of Success on Their First Amendment Claim A. Standard of Review We review the grant of a preliminary injunction for abuse of discretion, “which usually consists of clearly erroneous findings of fact or the application of an incorrect legal standard.” Nicholson v. Scoppetta, 344 F.3d 154, 165 (2d Cir.2003) (quotation marks and citation omitted). Where, as here, the relief that plaintiffs seek either (1) stays governmental action taken in the public interest pursuant to a statutory scheme, or (2) mandates an affirmative action, plaintiffs must demonstrate a “clear” or “substantial” likelihood of success on the merits of their claim. Tom Doherty Assoc., Inc. v. Saban Entm’t, Inc., 60 F.3d 27, 34 (2d Cir.1995); see also Hoblock v. Albany County Bd. of Elections, 422 F.3d 77, 96-97 (2d Cir.2005). Accordingly, the ultimate question is whether the District Court exceeded its allowable discretion in concluding that plaintiffs demonstrated a clear likelihood of success on the merits of their First Amendment claim. B. Substantive First Amendment Law Nothing in the Constitution requires a state to provide for the popular election of its judges. Republican Party of Minn. v. White, 536 U.S. 765, 788, 122 S.Ct. 2528, 153 L.Ed.2d 694 (2002). Even when a state chooses to do so, the Constitution grants it substantial authority to structure the “Times, Places and Manner of holding Elections.” U.S. Const. art. I, § 4, cl. 1; see also Tashjian v. Republican Party of Conn., 479 U.S. 208, 217, 107 S.Ct. 544, 93 L.Ed.2d 514 (1986). Nevertheless, once a state exercises its authority to provide for judicial elections, it must comport with constitutional requirements and prohibitions. Specifically, “[i]f the State chooses to tap the energy and legitimizing power of the democratic process, it must accord the participants in that process the First Amendment rights that attach to their roles.” Republican Party of Minn., 536 U.S. at 788, 122 S.Ct. 2528 (internal quotation marks omitted and alteration incorporated). The Supreme Court has identified “two different, although overlapping, kinds of rights” that the First Amendment grants: “ ‘the right of individuals to associate for the advancement of political beliefs, and the right of qualified voters, regardless of their political persuasion, to cast their votes effectively. Both of these rights, of course, rank among our most precious freedoms.’ ” Anderson v. Celebrezze, 460 U.S. 780, 787, 103 S.Ct. 1564, 75 L.Ed.2d 547 (1983) (quoting Williams v. Rhodes, 393 U.S. 23, 30-31, 89 S.Ct. 5, 21 L.Ed.2d 24 (1968)). While the role of voter “is of the most fundamental significance under our constitutional structure,” Burdick v. Takushi, 504 U.S. 428, 433, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992) (internal quotation marks omitted), “the rights of voters and the rights of candidates do not lend themselves to neat separation; laws that affect candidates always have at least some theoretical, correlative effect on voters,” Anderson, 460 U.S. at 786, 103 S.Ct. 1564 (internal quotation marks omitted). Accordingly, the Supreme Court has “minimized the extent to which voting rights cases are distinguishable from ballot access cases.” Burdick, 504 U.S. at 438, 112 S.Ct. 2059. In either context, “[o]ur primary concern is with the tendency of ballot access restrictions to limit the field of candidates from which voters might choose.” Anderson, 460 U.S. at 786, 103 S.Ct. 1564 (internal quotation marks omitted). Yet not every regulation that limits the field of candidates is constitutionally suspect, let alone unconstitutional. As noted above, a state possesses significant power to structure its own elections. Moreover, “as a practical matter, there must be a substantial regulation of elections if they are to be fair and honest and if some sort of order, rather than chaos, is to accompany the democratic processes.” Id. at 788, 103 S.Ct. 1564 (internal quotation marks omitted). Accordingly, in resolving a challenge that pits a State’s power to regulate its elections against the rights secured by the First Amendment, we cannot resort to any “litmus-paper test that will separate valid from invalid restrictions.” Id. at 789, 103 S.Ct. 1564 (internal quotation marks omitted). Instead, we must first ascertain “the character and magnitude of the asserted injury to the rights protected by the First and Fourteenth Amendments that the plaintiff seeks to vindicate.” Id. We must make that assessment not “in isolation, but within the context of the state’s overall scheme of election regulations.” Lerman v. Bd. of Elections in the City of New York, 232 F.3d 135, 145 (2d Cir.2000). The Supreme Court has underscored that in assessing the extent to which a given set of candidate restrictions burdens First Amendment rights, our review is neither formalistic nor abstract. Instead, we must turn a keen eye on how the electoral scheme functions in fact; indeed, “it is essential to examine in a realistic light the extent and nature of [the scheme’s] impact on voters.” Bullock v. Carter, 405 U.S. 134, 143, 92 S.Ct. 849, 31 L.Ed.2d 92 (1972). If our realistic assessment yields the conclusion that the electoral scheme lightly or even moderately burdens First Amendment rights, we apply a relaxed standard of review, according to which the restrictions generally are valid so long as they further an important state interest. Lerman, 232 F.3d at 145. On the other hand, if we conclude that a law imposes severe burdens, we apply strict scrutiny, which requires that the law be necessary to serve a compelling state interest. Id.; see also Bullock, 405 U.S. at 147, 92 S.Ct. 849 (“But under the standard of review we consider applicable to this case, there must be a showing of necessity.”). With those legal principles as general background, we turn to defendants’ claims of error. C. The District Court Did Not Find and Apply an Overly Broad First Amendment Right of Association Defendants’ most basic assertion is a unique argument that does not fit easily within the traditional framework set forth above. Instead of focusing on the burdens that this electoral scheme imposes, defendants focus on the affirmative scope of the First Amendment right upon which the District Court based its decision. The District Court erred, defendants maintain, because at bottom its decision rests on an overly broad right of associational freedom, namely, the “right to ... win a major party’s nomination.” Defendants attack two aspects of this purported right. They claim that the First Amendment does not apply to New York’s nominating process because the State has not provided a direct primary election. They next claim that even if the First Amendment does apply, it guarantees only “access to the nominating phase.” Defendants refined their second contention at oral argument by claiming further that the First Amendment prohibits only categorical exclusions, such as those based on race or sex, during the nominating phase. We disagree. Below we explain why the First Amendment applies to the nominating phase, describe the scope of protection that it guarantees to voters and candidates, and conclude that the District Court applied the proper standard on the facts of this case. 1. The First Amendment’s Guarantee of Freedom of Association Applies to New York’s Judicial Nominating Process It is true, as defendants contend, that “the processes by which political parties select their nominees are [not] ... wholly public affairs that States may regulate freely.” Cal. Democratic Party v. Jones, 530 U.S. 567, 572-73, 120 S.Ct. 2402, 147 L.Ed.2d 502 (2000) (holding that state may not compel political party to allow non-members to vote in its primary elections). But the fact that a political party’s nominating process is not “wholly public” does not render the First Amendment inapplicable. Id. Instead, the State must protect First Amendment rights at each “integral part of the procedure of choice,” and especially at those junctures of the electoral process that “effectively control[] the choice.” United States v. Classic, 313 U.S. 299, 318, 61 S.Ct. 1031, 85 L.Ed. 1368 (1941); see also Moore v. Ogilvie, 394 U.S. 814, 818, 89 S.Ct.. 1493, 23 L.Ed.2d 1 (1969) (“All procedures used by a state as an integral part of the election process must pass muster against the charges of discrimination or of the abridgement of the right to vote.”). In Classic, the Supreme Court considered whether Article I, § 2 of the U.S. Constitution secured the right of qualified Louisiana citizens “to vote in the Louisiana [congressional] primary and to have their ballots counted.” Classic, 313 U.S. at 307, 61 S.Ct. 1031. That constitutional provision, much like the state provision at issue here, provides that members of Congress “shall be ... chosen” in elections structured by the states. U.S. Const, art. I, § 2. In concluding that Article I, § 2 embraced the right to vote at a primary election, the Court relied on the fact that Louisiana had “exercis[ed] its privilege” to alter “the mode of choice from a single step, a general election, to two, of which the first is the choice at a primary,” thereby establishing the primary “by law” as “an integral part of the election machinery.” Id. at 316, 318, 61 S.Ct. 1031. The Court also noted that, even if Louisiana had not made the primary an integral part of the election by law, it could not ignore “the fact ... that the practical influence of the choice of candidates at the primary may be so great as to affect profoundly the choice at the general election.” Id. at 319, 61 S.Ct. 1031. Twelve years later, in Terry v. Adams, 345 U.S. 461, 73 S.Ct. 809, 97 L.Ed. 1152 (1953), the Court concluded that the Fifteenth Amendment’s prohibition on race-based voting exclusions applied not just to a primary election, but to a primary election run by a private club called the Jaybird Democratic Association. Id. at 462, 471, 73 S.Ct. 809. Of central importance to the Court’s finding of state action, and hence constitutional protection, was the fact that the only election that has counted in this Texas county for more than fifty years has been that held by the Jaybirds from which Negroes were excluded. The Democratic primary and the general election have become no more than perfunctory ratifiers of the choice that has already been made in Jaybird elections .... The Jaybird primary has become an integral part, indeed the only effective part, of the elective process that determines who shall rule and govern in the county. Id. at 469-70, 73 S.Ct. 809. Defendants claim that these cases apply only where the state has provided for a direct primary, whereas New York has opted for an indirect delegate primary election followed by a convention at which those delegates vote on the nomination. This variation in the process of public choice does not release the State from its First Amendment obligations. The animating principle in Classic and Terry is that constitutional protection extends to each State-created or State-endorsed “integral part of the election machinery,” not merely to one particular kind of primary election. Classic, 313 U.S. at 318, 61 S.Ct. 1031; see also Terry, 345 U.S. at 469, 73 S.Ct. 809; Bullock, 405 U.S. at 146, 92 S.Ct. 849 (holding that primary election ballot filing fee violated Fourteenth Amendment and noting that “the primary election may be more crucial than the general election in certain parts of Texas”). Consistent with that view, we previously held that constitutional protection attaches to all integral phases of the nominating process, regardless of whether the nomination is conferred directly by public ballot or indirectly by the votes of elected party officials. In Seergy v. Kings County Republican County Committee, 459 F.2d 308 (2d Cir.1972), we held that the Equal Protection Clause of the Fourteenth Amendment applied to those activities of the Kings County Republican County Committee that involved “public electoral functions,” id. at 314, in particular, those activities directly related to “the selection of party nominees,” id. at 315. We specifically included nominations effected entirely by the committee members themselves, namely, “the nomination of candidates to fill vacancies or to run in special elections, or the giving of consent to candidacies [of] non-members of the party” because “in such cases [the committee] is unquestionably playing an integral part in the state scheme of public elections.” Id. at 314; see also Ripon Soc’y v. Nat’l Republican Party, 525 F.2d 567, 586 (D.C.Cir.1975) (m banc) (noting in the context of nominating convention that “[i]f the right to vote is a right to true participation in the elective process, then it is heavily implicated in the nomination process”). In this case, the delegate primary election and subsequent nominating convention are State-ci’eated and legally required aspects of the process of public choice in New York. See N.Y. Elec. L. §§ 6-106, -124. Further, as a practical matter, the evidence established that the result of that nominating process profoundly affects the choice at the general election, a fact to which “we cannot close our eyes” because it means that exclusionary nominating-phase regulations may well “operate to deprive the voter of his constitutional right of choice.” Classic, 313 U.S. at 319, 61 S.Ct. 1031; see also Terry, 345 U.S. at 469, 73 S.Ct. 809; Bullock, 405 U.S. at 146, 92 S.Ct. 849. Accordingly, both segments of the nominating procedure are integral parts of the State’s election machinery, and therefore subject to the requirements and prohibitions of the First Amendment. Seergy, 459 F.2d at 313 (holding that constitutional protection attaches to nominating process regardless of whether voting occurs “directly [by voters] or indirectly through their committeemen”). 2. The First Amendment Guarantees Voters and Candidates a Realistic Opportunity to Participate in the Nominating Phase Free From Severe and Unnecessary Burdens Having concluded that New York must afford voters and candidates the right to associate through and in the judicial nominating process, we now examine the scope of that right. Defendants maintain that the First Amendment grants voters and candidates only the right to “access” the nominating process. The term “access” is vague, and the only definition that defendants advance to flesh it out is that a candidate must “have a chance to enter the judicial convention.” At oral argument, defendants supplemented this definition by asserting that the Constitution prohibits only those regulations that exclude voters or candidates on the basis of race, sex, or some other categorical demarcation. As we explain below, the First Amendment affords candidates and voters a realistic opportunity to participate in the nominating process, and to do so free from burdens that are both severe and unnecessary to further a compelling state interest. Further, while categorical race and sex-based exclusions undoubtedly violate the associational rights of voters and candidates, exclusions that result from a complex of otherwise facially valid regulations also may offend the First Amendment. Although no case has passed on a scheme identical to New York’s unique judicial election process, the above principles derive' directly from a line of Supreme Court cases limiting a State’s power to structure its elections and regulate access to its ballot. In Williams v. Rhodes, 393 U.S. 23, 89 S.Ct. 5, 21 L.Ed.2d 24 (1968), for example, the Court confronted an “entangling web of [Ohio] election laws” that “effectively foreclosed [the state’s] presidential ballot to all but Republicans and Democrats.” Id. at 35, 89 S.Ct. 5 (Douglas, J., concurring). The regulations required new parties to “erect elaborate political machinery,” id. at 37, 89 S.Ct. 5, in order to collect an onerous amount of signatures from a restricted pool of signatories, elect a state central committee comprised of two members from every congressional district and county in the state, and elect a slate of delegates and alternates to the national convention. Id. at 25 n. 1, 89 S.Ct. 5. It was “virtually impossible” for a new party to satisfy those ballot access requirements. Id. at 24, 89 S.Ct. 5 (Black, J., for the Court). Given the severe burdens that the “totality of the ... restrictive laws taken as a whole impose[d] on voting and associational rights,” and the lack of a compelling state interest to justify them, the Court struck the scheme pursuant to the Equal Protection Clause of the Fourteenth Amendment. Id. at 34, 89 S.Ct. 5. Similarly, in Bullock v. Carter, 405 U.S. 134, 92 S.Ct. 849, 31 L.Ed.2d 92 (1972), the Court focused on the “patently exclusionary character” of the regulation to find a constitutional violation. Id. at 143, 92 S.Ct. 849. There, a Texas filing fee “in every practical sense precluded” all but affluent candidates “from seeking the nomination of their chosen party, no matter how qualified they might be, and no matter how broad or enthusiastic their popular support.” Id. at 143, 92 S.Ct. 849. “The effect of this exclusionary mechanism on voters” was to “substantially limit[ ] ... their choice of candidates.” Id. at 144, 92 S.Ct. 849. In light of that “real and appreciable impact on the exercise of the franchise,” the Court applied strict scrutiny, determine