OPINION
PER CURIAM:On October 7, 2003, California voters will be asked to cast a ballot on some of the most important issues facing the State, including an unprecedented vote on the recall of a governor. However, forty-four percent of the electorate will be forced to use a voting system so flawed that the Secretary of State has officially deemed it “unacceptable” and banned its use in all future elections. The inherent defects in the system are such that approximately 40,000 voters who travel to the polls and cast their ballot -will not have their vote counted at all. Compounding the problem is the fact that approximately a quarter of the state’s polling places will not be operational because election officials have insufficient time to get them ready for the special election, and that the sheer number of gubernatorial candidates will make the antiquated voting system far more difficult to use.
Plaintiffs allege that the use of the obsolete voting systems in some counties rather than others will deny voters equal protection of the laws in violation of the United States Constitution. They seek to postpone the vote until the next regularly scheduled statewide election six months from now, when the Secretary of State has assured that all counties will be using acceptable voting equipment, and all the polls will be open. We agree that the issuance of a preliminary injunction is warranted and reverse the order of the district court.
I
It is now well over a century since Herman Hollerith won a national design competition and his invention, the punchcard, was used to tabulate the 1890 census. Although now an anachronism in the world of commerce, his creative legacy endures in elective politics. It was nearly a half century ago, around the time when newer information technologies were beginning to supplant its use in business, that Dr. Joseph P. Harris conceived of using the punchcard as a means of recording votes, and the YotoMatic machine was born. It was first used in a few counties in Georgia during the 1964 primary election, then by San Joaquin and Monterey counties in California during the 1964 general election.
VotoMatic punchcard voting systems use a pre-scored heavy stock paper ballot with columns of small, perforated rectangles. These pre-scored rectangles are removed by force to create a space that can be read by a computer or tabulating machine. A voter uses the pre-scored punchcard voting system by first inserting it in a hollow mechanical holder. A hinged booklet is attached to the mechanical holder. The booklet is attached so that it is centered over the inserted punchcard. The pages of the booklet are crimped to a hinge, so that they cannot be removed during the voting process. The attachment permits the voter to see one row of the punchcard at a time. As the pages are turned, a different row of the punchcard is exposed. The candidates and ballot initiatives are not listed on the punchcard, but must be discerned by the voter by cross-referencing the rectangles with the election booklet listing the candidates and other ballot measures that are the subject of the election. To register their vote, voters then find the correct page and punch out the appropriate rectangle with a metal stylus to create a hole in the punchcard. After *889casting a ballot, the punchcard voter does not have an opportunity to inspect the ballot for vote accuracy. All the voter is left to examine is a standard Hollerithian punchcard with holes punched through certain numbers.
The material separated from the punch-card when forming the hole is known as a chad. If the ballot is not positioned correctly in the voting machine, the incorrect rectangles will be removed. If the chad is not removed completely by the stylus, the tabulation machine may not count the vote. Unlike mechanical lever machines, the Vo-toMatic system does not have any built-in protection preventing the voter from casting more than one vote for a candidate or ballot measure. In that event, the software is designed not to count the vote at all. The system is subject to mechanical problems. If the mechanical holder is not constructed properly, if its materials have deteriorated through use, if the punchcard has not been pre-scored properly, if the punchcard’s manufactured dimensions are not within tolerances, the vote will either not be counted or will be counted incorrectly. The data reader must also be functioning properly, and must be free of extraneous objects, like paperclips, staples and chads, that may fall into it during tabulation. Otherwise votes will not be counted, or will be counted incorrectly. Occasionally, like the similar problems with facsimile machines and copiers, the tabulation machine will grab two cards, rather than one, resulting in counting errors. If the data card reader jams, it is left to the operator to decide whether the votes have been tabulated.
The first major study of the efficacy of the VotoMatic system was performed by Ray G. Saltman in 1975 while he was working for the United States National Bureau of Standards, now known as the National Institute of Standards and Technology. As his report later described it, “[s]erious problems in computerized vote-tallying had been experienced in San Francisco in 1968, in Los Angeles and Houston in 1972, and in other places ...” In the wake of those controversies, and “in recognition of concerns expressed by Congress, and by election officials and the public,” the United States General Accounting Office requested the National Bureau of Standards to perform a national study of punchcard voting. Saltman was assigned the project. His study revealed significant problems with the system.
In the ensuing years, additional problems were reported in the use of the pre-scored punchcard system. After press reports of votes improperly being added to a candidate’s tally in San Francisco in 1986 because of electrical fluctuations in the equipment, an elimination of the votes of an entire precinct in San Joaquin County in 1984 because of a loose chad jamming the system, the reported incorrect transfer of 15,000 votes from one candidate to another in Orange County in 1980, and numerous other reported problems, the National Bureau of Standards decided to update its study. Saltman was again charged with the project. He issued an updated report in 1988 that buttressed the earlier analysis critical of the use of pre-scored punchcard voting devices. However, the use of the systems continued.
A competing punchcard voting system, Datavote, uses a mechanical device to create holes in the ballots in the appropriate locations. The ballots are not pre-scored. The Datavote card, unlike the Votomatic card, contains the candidates’ names so that voters may examine the card to make sure their vote has been correctly recorded.
Just as the black and white fava bean voting system of revolutionary times was replaced by paper balloting, and the paper *890ballot replaced by mechanical lever machine, newer technologies have emerged to replace the punchcard, including optical scanning and touch screen voting. Optical scanning systems use what is commonly termed a “marksense” form, in which voters use a pencil to indicate their choices on a pre-printed form. The marksense ballots are then counted by an optical scanner. Touch screen voting machines, also known as “direct recording electronic devices,” allow the voter to touch the name of the candidate on a screen to record his or her vote. Direct recording electronic devices are programmed to prevent a voter from casting more than the allowed number of votes. Once the voter has completed voting, the computer records the vote electronically.
In California, the Secretary of State is charged with the responsibility of establishing “regulations governing the use of voting machines, voting devices, and vote tabulating devices.” Cal. Elec.Code § 19100. Under California law, “[n]o voting system, in whole or in part, shall be used unless it has received the approval of the Secretary of State, prior to any election at which it is to be first used.” Cal. Elec.Code § 19201. Presently, California counties are not uniform in the choice of voting method. All four major voting systems (VotoMatic, Datavote, optical scanning and direct recording electronic device) are used in various parts of the State.
The Secretary of State is also charged with the responsibility to review voting systems and to withdraw certification if the systems are “defective” or “unacceptable.” Cal. Elec.Code § 19222. The governing language provides:
The Secretary of State shall review voting systems periodically to determine if they are defective, obsolete, or otherwise unacceptable. The Secretary of State has the right to withdraw his or her approval .previously granted under this chapter of any voting system or part of a voting system should it be defective or prove unacceptable after such review. Six months’ notice shall be given before withdrawing approval unless the Secretary of State for good cause shown makes a determination that a shorter notice period is necessary. Any withdrawal by the Secretary of State of his or her previous approval of a voting system or part of a voting system shall not be effective as to any election conducted within six months of that withdrawal.
Id.
The origins of the present controversy date to the aftermath of the 2000 presidential election, when national attention was drawn to the eccentricities of voting by pre-scored punchcards. Various groups whose members are voters of color and a number of individuals (collectively referred to as the “Common Cause Plaintiffs”) filed an action seeking injunctive and declaratory relief based on alleged deficiencies in the California electoral process. See First Amended Complaint for Injunctive and Declaratory Relief, Common Cause, et. al. v. Jones, (C.D.Cal.2002) (No. 01-03470) (“Common Cause I ”). In that lawsuit, the Common Cause Plaintiffs alleged that use of pre-scored punchcard ballots violated the rights to equal protection, due process, and the privileges and immunities guaranteed to United States citizens secured by the Fourteenth Amendment and the Voting Rights Act, 42 U.S.C. § 1973. The Common Cause Plaintiffs’ theory was that the Secretary of State’s permission to counties to use pre-scored punchcard voting systems violated their federal statutory and constitutional rights because voters in counties using punchcard voting were substantially less likely to have their votes *891counted because of inherent deficiencies in the system.
Shortly after the district court denied the Secretary of State’s motion for judgment on the pleadings in Common Cause I, Secretary of State Bill Jones issued a proclamation decertifying VotoMatic and Pollstar pre-scored punch-card systems for use in California pursuant to Cal. Gov’t Code § 12172.5 and Cal. Elec.Code § 19222, effective January 1, 2006. In the official decertification proclamation dated September 18, 2001, the Secretary of State stated:
As I order this proclamation, I want to be very clear on two points. First, Vo-tomatic and Pollstar voting systems are old technology and their use today can be seen as analogous to the use of typewriters — they worked well for many years but are now obsolete in the world of the personal computer. One of these systems was initially approved for use in 1965. Voters are entitled to have the infrastructure of democracy upgraded to reflect technological improvements to the voting process. Second, it is critically important that the transition to any new technologies be orderly and well thought out. A poorly planned rush to implement a new voting technology without providing adequate time for counties to purchase these systems, and to train their staff, pollworkers, and voters on the proper use of new equipment could easily result in great harm to the most fundamental right of the people, the right to vote.
In a press release accompanying the proclamation, the Secretary of State stated that “[w]e cannot wait for a Florida-style election debacle to occur in California before we replace archaic voting systems.” The release noted that though the Secretary had previously decertified a number of obsolete voting systems, this marked the first time that California had decerti-fied a system in use by California counties.
Several months prior to the decertification of the pre-scored voting systems, the Secretary of State had certified that two propositions, Propositions 53 and 54, had acquired sufficient signatures to be placed on the ballot at the next general election. He issued a certification proclamation placing the initiatives on the March 2, 2004 primary election ballot. These are the initiatives now rescheduled for the special October 7, 2003 election.
The Secretary’s decertification rendered most of the underlying issues in Common Cause I moot, except the question of remedy. The Common Cause Plaintiffs had requested that the Votomatic and Pollstar pre-scored punchcard systems be replaced prior to the 2004 primary or general election. On that question, the district court held that it was “plainly feasible for the PPC counties to convert to ‘other certified voting equipment’ by March 2004.” Order, Common Cause I at 3. The parties then entered into an agreement under which the decertification of the punchcard voting systems would be advanced to March 1, 2004, in time for the next regularly scheduled statewide election. To effectuate the settlement, the parties submitted to a consent decree embracing the settlement terms, which was approved by the district court. Judgment was entered. No appeal was taken from the Court’s entry of judgment or any order issued in Common Cause I. Since the entry of judgment, primary and general elections were held in 2002, along with numerous local elections.
On March 25, 2003, a petition for the recall of Governor Gray Davis was served on the Secretary of State pursuant to Cal. Const, art. II, § 14(a). On July 23, 2003, Secretary of State Kevin Shelley certified, pursuant to Cal. Elec.Code § 11109, that *892sufficient signatures had been obtained on the recall petition to hold an election. Under the California Constitution, the Lieutenant Governor is charged with setting the date of a gubernatorial recall election. See Cal. Const, art. II, § 17.
The California Constitution requires that the election be held not less than 60 days and not more than 80 days from the date of certification. See Cal. Const, art. II, § 15(a). The only temporal exception exists when a regular election is already scheduled to be held within 180 days of the date of certification. See Cal. Const, art. II, § 15(b). When the Secretary of State certified the gubernatorial recall, the next regular election was scheduled for March 2, 2004. Because that was more than 180 days from the certification date, the Constitution required the special recall election to be held not less than 60 and not more than 80 days from the certification date. If the Secretary of State had issued the certification a month and a half later than he did, as originally planned based on the date by which sufficient valid petition signatures needed to be filed, the recall election would have been held at the next regular election March 2, 2004. However, sufficient signatures were submitted before the deadline, motivating the Secretary of State to issue a certification earlier than originally planned. One day after the Secretary of State’s certification, the California Lieutenant Governor scheduled the recall vote for October 7, 2003.
Two ballot initiatives originally scheduled to be placed on the March 2004 regular election ballot, Propositions 53 and 54, were added to the special recall election by the Secretary of State after the Lieutenant Governor announced the schedule. Proposition 53 is a proposed amendment to the California Constitution that would dedicate part of the state budget each year to state and local infrastructures, such as water, highway, and park projects. Proposition 54 is another proposed amendment to the California Constitution which would prevent the State from collecting or retaining racial and ethnic data about health care, hate crimes, racial profiling, public education, and public safety.
Plaintiffs commenced the instant action seeking to enjoin the proposed election until it can be conducted without the use of any pre-scored punchcard voting systems. The district court denied Plaintiffs’ motion for a preliminary injunction. This appeal followed.
II
“Voting is one of the most fundamental and cherished liberties in our democratic system of government.” Burson v. Freeman, 504 U.S. 191, 214, 112 S.Ct. 1846, 119 L.Ed.2d 5 (1992) (Kennedy, J., concurring). As the Supreme Court put it in Reynolds v. Sims, 377 U.S. 533, 555, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964): “The right to vote freely for the candidate of one’s choice is of the essence of a democratic society, and any restrictions on that right strike at the heart of representative government.”
In this case, Plaintiffs allege that the fundamental right to have votes counted in the special recall election is infringed because the pre-scored punchcard voting systems used in some California counties are intractably afflicted with technologic dys-calculia. They claim that the propensity for error in these voting systems is at least two and a half times greater than for any other voting technology used in California. The effect is not trivial. At least six California counties plan to employ the technology during the special election, namely, Los Angeles, Santa Clara, San Diego, Sacramento, Mendocino, and Solano. These counties comprise 44% of the total electorate. They include the most populous coun*893ty in the State and the county in which the state capitol is located. Plaintiffs tendered evidence showing that 40,000 voters who cast ballots in these counties would not have their votes counted because of technological defects in the pre-scored punch-card voting system. It is perhaps ironic that the sitting governor could well east a vote on his own recall that would not be tallied. Many candidates seeking to replace him would face a similar risk. Plaintiffs also allege that the affected counties contain a significantly higher percentage of minority voters than the other counties, causing a disproportionate disenfranchisement of minority voters.
In order to obtain a preliminary injunction on their claims, the Plaintiffs were required to demonstrate “(1) a strong likelihood of success on the merits, (2) the possibility of irreparable injury to plaintiffs] if preliminary relief is not granted, (3) a balance of hardships favoring the plaintiffs], and (4) advancement of the public interest (in certain cases).” Johnson v. Cal. State Bd. of Accountancy, 72 F.3d 1427, 1430 (9th Cir.1995) (internal quotation marks omitted). Alternatively, injunctive relief could be granted if the Plaintiffs “demonstrate^ either a combination of probable success on the merits and the possibility of irreparable injury or that serious questions are raised and the balance of hardships tips sharply in [their] favor.” Id. (internal citations and quotation marks omitted). “These two alternatives represent ‘extremes of a single continuum,’ rather than two separate tests.... ” Clear Channel Outdoor Inc. v. Los Angeles, 340 F.3d 810, 2003 WL 21947181 (9th Cir. Aug.15, 2003) (internal quotation marks omitted). Thus, the greater the relative hardship to the party seeking the preliminary injunction, the less probability of success must be established by the party. Id. “In cases where the public interest is involved, the district court must also examine whether the public interest favors the plaintiff.” Fund for Animals, Inc. v. Lujan, 962 F.2d 1391, 1400 (9th Cir.1992); see also Caribbean Marine Servs., Co. v. Baldrige, 844 F.2d 668, 674 (9th Cir.1988).
In general, we review the denial of a preliminary injunction for abuse of discretion. Bay Area Addiction Research & Treatment, Inc. v. Antioch, 179 F.3d 725, 730 (9th Cir.1999). The district court “necessarily abuses its discretion when it bases its decision on an erroneous legal standard or on clearly erroneous findings of fact.” Rucker v. Davis, 237 F.3d 1113, 1118 (9th Cir.2001) (en banc), rev’d on other grounds, Dept. of Hous. & Urban Dev. v. Rucker, 535 U.S. 125, 122 S.Ct. 1230, 152 L.Ed.2d 258 (2002). When the district court is alleged to have relied on an erroneous legal premise, we review the underlying issues of law de novo. Does 1-5 v. Chandler, 83 F.3d 1150, 1152 (9th Cir.1996).
The district court assumed that the Plaintiffs would suffer irreparable injury, but concluded that the Plaintiffs were not likely to prevail on the merits. It also concluded that the balance of hardships and consideration of the public interest weighed heavily in favor of allowing the special election to proceed.
Ill
The first factor in a preliminary injunction analysis is the probability that the plaintiff will succeed on the merits. Under the continuum analysis of Clear Channel, the greater the demonstrated harm, the lesser the requirement of probability of success. Here, the district court assumed irreparable harm, and we agree with its assumption. As the district court properly observed, Plaintiffs will have no remedy for their claims following the election. *894The district court concluded that the Plaintiffs had no likelihood of success on the merits of their claims as a matter of substance, and further concluded that the claims were likely barred by the doctrines of res judicata and laches. We respectfully disagree and conclude that the district court erred in its legal analysis.
A
We conclude that the Plaintiffs have satisfied the requirement of establishing a sufficient probability of success on their federal constitutional claims on the merits. As we recently noted, “[v]ot-ing is a fundamental right subject to equal protection guarantees under the Fourteenth Amendment.” Idaho Coalition United for Bears v. Cenarussa, 342 F.3d 1073, 2003 WL 22072191 at *2, (9th Cir. Sept. 8, 2003) (citing Reynolds, 377 U.S. at 561-62, 84 S.Ct. 1362). Additionally, “[t]he ballot initiative, like the election of public officials, is a ‘basic instrument of democratic government,’ and is therefore subject to equal protection guarantees.” Id. (quoting Cuyahoga Falls v. Buckeye Comm. Hope Found., 538 U.S. 188, 123 S.Ct. 1389, 1395, 155 L.Ed.2d 349 (2003)) (citations omitted).
In this case, Plaintiffs’ Equal Protection Clause claim mirrors the one recently analyzed by the Supreme Court in Bush v. Gore, 531 U.S. 98, 121 S.Ct. 525, 148 L.Ed.2d 388 (2000). As the Supreme Court held in that case: “Having once granted the right to vote on equal terms, the State may not, by later arbitrary and disparate treatment, value one person’s vote over that of another.” Id. at 104-05, 121 S.Ct. 525 (citing Harper v. Virginia Bd. of Elections, 383 U.S. 663, 665, 86 S.Ct. 1079, 16 L.Ed.2d 169 (1966)). This holding was consistent with a long line of Supreme Court precedent holding that the right to vote includes the right to have one’s vote counted. As the Supreme Court wrote in Reynolds,
Undeniably the Constitution of the United States protects the right of all qualified citizens to vote, in state as well as in federal elections. A consistent fine of decisions by this Court in cases involving attempts to deny or restrict the right of suffrage has made this indelibly clear. It has been repeatedly recognized that all qualified voters have a constitutionally protected right to vote, and to have their votes counted.
377 U.S. at 554-55, 84 S.Ct. 1362 (citations omitted).
This was not a new view of the Constitution in 1964. More than two decades previously, the Court observed in United States v. Classic, 313 U.S. 299, 315, 61 S.Ct. 1031, 85 L.Ed. 1368 (1941) that: “Obviously included within the right to choose, secured by the Constitution, is the right of qualified voters within a state to cast their ballots and have them counted....” Nor was it a novel concept that the Court expressed in Classic. Almost three decades before Classic, Justice Holmes stated that it is “unquestionable that the right to have one’s vote counted is as open to protection by Congress as the right to put a ballot in a box.” United States v. Mosley, 238 U.S. 383, 386, 35 S.Ct. 904, 59 L.Ed. 1355 (1915).
Plaintiffs argue that the use of defective voting systems creates a substantial risk that votes will not be counted. In addition, they claim that the use of defective voting systems in some counties and the employment of far more accurate voting systems in other counties denies equal protection of the laws by impermissibly diluting voting strength of the voters in counties using defective voting systems. In short, the weight given to votes in non-punchcard counties is greater than the weight given to votes in punchcard coun*895ties because a higher proportion of the votes from punchcard counties are thrown out. Thus, the effect of using punchcard voting systems in some, but not all, counties, is to discriminate on the basis of geographic residence.
This is a classic voting rights equal protection claim. As the Supreme Court explained in Bush, “ ‘the right of suffrage can be denied by a debasement or dilution of the weight of a citizen’s vote just as effectively as by wholly prohibiting the free exercise of the franchise.’ ” 531 U.S. at 105, 121 S.Ct. 525 (quoting Reynolds, 377 U.S. at 555, 84 S.Ct. 1362). Further, the “ ‘idea that one group can be granted greater voting strength than another is hostile to the one man, one vote basis of our representative government.’ ” Id. at 107, 121 S.Ct. 525 (quoting Moore v. Ogilvie, 394 U.S. 814, 819, 89 S.Ct. 1493, 23 L.Ed.2d 1 (1969)). As the Court stated much earlier in Wesberry v. Sanders, 376 U.S. 1, 8, 84 S.Ct. 526, 11 L.Ed.2d 481 (1964), “To say that a vote is worth more in one district than in another would ... run counter to our fundamental ideas of democratic government....”
Plaintiffs’ equal protection claim is much the same as the one in Gray v. Sanders, 372 U.S. 368, 83 S.Ct. 801, 9 L.Ed.2d 821 (1963). That case involved a Georgia county unit voting system that weighted rural county votes more heavily than urban county votes and weighted the votes from some small rural counties more heavily than larger rural counties. The Supreme Court held that this constituted a violation of the Equal Protection Clause, writing that “once the class of voters is chosen and their qualifications specified, we see no constitutional way by which equality of voting power may be evaded.” Id. at 381, 83 S.Ct. 801. As the Court put it: “Every voter’s vote is entitled to be counted once. It must be correctly counted and reported.” Id. at 380, 83 S.Ct. 801. Gray echoed Reynolds’ admission that “the basic principle of representative government remains, and must remain, unchanged — the weight of a citizen’s vote cannot be made to depend on where he lives.” Reynolds, 377 U.S. at 567, 84 S.Ct. 1362. As the Court noted: “A citizen, a qualified voter, is no more nor no less so because he lives in the city or on the farm. This is the clear and strong command of our Constitution’s Equal Protection Clause.” Id. at 568, 84 S.Ct. 1362. In short, the Equal Protection Clause requires “the opportunity for equal participation by all voters in the election....” Id. at 566, 84 S.Ct. 1362.
Plaintiffs’ claim presents almost precisely the same issue as the Court considered in Bush, that is, whether unequal methods of counting votes among counties constitutes a violation of the Equal Protection Clause. In Bush, the Supreme Court held that using different standards for counting votes in different counties across Florida violated the Equal Protection Clause. 531 U.S. at 104-07, 121 S.Ct. 525. The Plaintiffs’ theory is the same, that using error-prone voting equipment in some counties, but not in others will result in votes being counted differently among the counties. In short, they contend that a vote cast in Los Angeles or San Diego is entitled to the same weight as a vote cast in San Francisco.
No voting system is foolproof, of course, and the Constitution does not demand the use of the best available technology. However, what the Constitution does require is equal treatment of votes cast in a manner that comports with the Equal Protection Clause. Like the Supreme Court in Bush, “[t]he question before [us] is not whether local entities, in the exercise of their expertise, may develop different systems for implementing elections.” *896531 U.S. at 109, 121 S.Ct. 525. Rather, like the Supreme Court in Bush, we face a situation in which the United States Constitution requires “some assurance that the rudimentary requirements of equal treatment and fundamental fairness are satisfied.” Id.
It is virtually undisputed that pre-scored punchcard voting systems are significantly more prone to errors that result in a voter’s ballot not being counted than the other voting systems used in California. As the Supreme Court observed in Bush: “This case has shown that punchcard balloting machines can produce an unfortunate number of ballots which are not punched in a clean, complete way by the voter.” 531 U.S. at 104, 121 S.Ct. 525.
In addition to difficulties with punching the card, there are mechanical and software anomalies that compound the error rate. In California, the Secretary of State’s decertification is almost dispositive of the question of the viability of pre-scored punchcard voting systems. In banning the systems from use in California, the Secretary was required by state law to find that a voting system was “defective” and “unacceptable” before issuing an official proclamation of decertification. The Secretary of State conceded the fallacies in the system in the opening fines of its brief to this Court, stating:
Punch-card voting systems are old technology more prone to voter error than are newer voting systems. Both the present and the prior Secretary of State have been acutely aware of this reality, and have taken aggressive steps to eliminate the use of punch-card machines statewide.
In addition, Plaintiffs tendered evidence that statistically significant disparities exist between the systems. They offered affidavits from Roy G. Saltman, referenced earlier, who had produced the seminal work in this field in the 1970s and 1980s for the United States National Bureau of Standards, and Dr. Henry Brady, a Professor of Political Science at the University of California, Berkeley, and Director of the University of California’s Survey Research Center.
Mr. Saltman identified several problems inherent in the use of pre-scored punch-card systems, including:
• The fact that voters rather than machines are responsible for removing the chad, leading to chads being incompletely separated from the punch-card, leaving the chad attached to the punchcard. During the counting process, these hanging chads may be pressed back into the card, altering the voters’ intent.
• Chads intended not to be removed may be removed during the counting process due to excessive handling, action of the counting machine, or manipulation, altering the voter’s intent.
• Unlike other systems, there is no mechanism to prevent overvoting (i.e., voting for more than one candidate or more than the allotted number of candidates).
• Voters are unable to examine the ballot for accuracy before leaving the polling place.
• Manual examination of pre-scored punchcard ballots to determine the voter’s intent is highly subjective. For example, manual counters are forced to determine whether a pinprick point on a chad demonstrated an intent to register a vote.
• The defects in the pre-scored punch-card voting system are fundamental and cannot be fixed by engineering or management alterations.
The experience of the presidential election in Florida in 2000 supports Mr. Salt-*897man’s conclusions. There, counters were asked to ascertain the voters’ intent by examining “hanging chads” (in which one corner of the chad is attached to the ballot card), “tri chads” (in which three corners of the chad are hanging but the hole has been punched), “pregnant chads” (in which the hole is punched through the chad but it still hangs from all four sides), and “dimpled chads” (in which there is an indentation in the chad, but no clean hole has been punched). In addition, the use of some mechanical guides, such as the one designed for the now infamous “butterfly” ballot, have proven confusing to the voter. Disabled and visually impaired voters have particular difficulty in using the technology. All of this demonstrates that pre-scored punchcard systems are more prone to human error than other certified systems. And errors may not be treated equally. The Supreme Court commented on one aspect of this problem in Bush, observing that under the pre-scored punchcard voting system:
[t]he citizen whose ballot was not read by a machine because he failed to vote for a candidate in a way readable by a machine may still have his vote counted in a manual recount; on the other hand, the citizen who marks two candidates in a way discernible by the machine will not have the same opportunity to have his vote count, even if a manual examination of the ballot would reveal the requisite indicia of intent. Furthermore, the citizen who marks two candidates, only one of which is discernible by the machine, will have his vote counted even though it should have been read as an invalid ballot.
531 U.S. at 108, 121 S.Ct. 525.
In addition, there are mechanical problems inherent in the use of the technology. The punchcard itself is somewhat fragile. Although now just a generational memory, Americans living in the era of widespread punchcard use in business and government can hardly forget the ubiquitous admonition not to “fold, spindle, or mutilate” the card. Tabulating machines, which cannot be admonished, have a tendency to damage the cards in the mechanical counting process, causing the vote not to be counted or to be counted contrary to the voter’s intent. This creates recount problems. As Mr. Saltman observed in his 1988 study: “It is generally not possible to exactly duplicate a count obtained on pre-scored punch cards, given the inherent physical characteristics of these ballots and the variability in ballot-punching performance of real voters.”
Dr. Brady analyzed the differences in residual voting rates (i.e., uncounted votes) in counties using pre-scored punchcard voting systems and those that did not, and concluded that the probability the differences occurred by chance was less than one in a billion. He concluded that the error rate for punchcard systems was at least two and a half times greater than those of other systems.
Dr. Brady also concluded that the use of pre-scored punchcard voting systems discriminated against minorities in several respects. First, the six punchcard counties have a larger percentage of minorities (46%) than non-punchcard counties (32%). Second, the analysis indicated that when pre-scored punchcard systems were used, minority voters had significantly higher residual vote rates than non-minorities.
Independent research confirms the error difference between pre-scored punch-card systems and others in use. The July 2001 Report of the Caltech-MIT Voting Technology Project (“Caltech-MIT Report”) studied the residual vote rates of different voting systems from 1988-2000 in the entire country, and found that punch-cards lose significantly more votes than *898optically scanned paper ballots. More significantly, the report found that:
[t]hese patterns hold up to closer statistical scrutiny, holding constant turnout, income, racial composition of counties, age distributions of counties, literacy rates, the year of a shift in technology, the number of offices and candidates on the ballot, and other factors that operate in a county or in a particular year.” Caltech-MIT Report 22 (emphasis added).
Given the vast amount of data the authors of this study were able to incorporate, and given the many control variables to which the pattern of disparity is resistant, it is likely that Plaintiffs would prevail in rebutting any allegations that residual rate disparities in voting systems are completely attributable to different rates of voter intent to abstain or overvote, rather than to the use of different voting machines.
The district court discounted the impact of voting systems on the special election, relying in part on the Secretary of State’s attestation that he would “be undertaking extensive voter education efforts that could have the effect of lowering the residual rate in the upcoming election.” However, Plaintiffs effectively countered this unsupported assertion with statistical evidence showing that voter education was ineffective in counteracting the error rates inherent in the use of pre-scored punchcard voting systems. Indeed, the error rate in the use of punchcard systems after the national attention given the problems of punchcard voting in the 2000 election actually increased in the 2002 California gubernatorial rate when compared with the 1992 gubernatorial election. Further, as we shall discuss later, the Secretary of State has already missed statutory deadlines for submitting educational information to voters concerning the initiatives on the ballot and faces the additional impediment of having a quarter of the state’s polling places non-operational for the special election.
The district court also noted that inter-venor Costa had tendered expert testimony showing that affirmative choices of voters and other factors contribute to a residual vote rate. Plaintiffs rejoined with statistical evidence taking these factors into account, and also pointing out that affirmative choice cannot explain the differences in residual voting rates between punchcard voting and other systems.
Intervenor Costa argues that every voting system is prone to some errors and that we need to allow some “play in the joints.” If there were equal “play in the joints,” this argument would have more force. However, a long line of studies establishes that the difference of error rates between pre-scored punchcard voting systems is of statistical significance at the highest level, even accounting for other factors. Thus, according to the best scientific studies analyzing voting techniques over the past thirty years, the significantly increased number of errors in punchcard voting as opposed to other methods cannot be explained by other factors, or chance. Dr. Brady, for example, concluded that the probability that these differences in error rate occurred by chance was less than one in a billion. In short, the vast weight of the evidence shows that voters in counties using pre-scored punchcard balloting will have a statistically more probable chance that their vote will not be counted than voters in other counties. It is in this intrinsically unequal treatment that the constitutional problem lies.
More importantly, the Secretary of State has already concluded that use of this technology is unacceptable and must be prohibited from future use in California. His decision to postpone the decertification *899was not because the technology was reliable; rather, his decision was based on what he considered a reasonable time for California counties to implement the change in systems.
Obviously, these are issues that each party disputes. However, as we have noted, to prevail on a motion for a preliminary injunction, a plaintiff need not demonstrate that he will prevail at trial, or that no other reading of the evidence possibly could be “conjured up,” as the district court put it. A plaintiff must only show that the likelihood is such that, when considered with the demonstrated hardship, a preliminary injunction should issue to preserve the respective rights of the parties. Here, the Plaintiffs have tendered more than sufficient proof to satisfy that preliminary burden.
Despite the evidentiary doubts expressed by the district court, its conclusion as to the Plaintiffs’ chances of success appeared to be largely founded on public policy, the notion that the voting process “will invariably impose some burden upon individual voters,” Burdick, v. Takushi, 504 U.S. 428, 433, 112 S.Ct. 2059, 119 L.Ed.2d 245 (1992), and that the court “must weigh the character and magnitude of the asserted injury ... against the precise interests put forward by the State as justifications for the burden imposed by its rule.” Burdick, 504 U.S. at 434, 112 S.Ct. 2059 (internal quotation marks omitted). The precise contours of the district court’s rebanee on Burdick are unclear. To the extent that the district court relied on Burdick for its analysis of the likelihood of success, the district court improperly conflated balance of hardship and pubbe interest factors, see Sammartano v. First Judicial Dist. Court, 303 F.3d 959, 974 (9th Cir.2002), then further conflated this analysis into an examination of the Plaintiffs’ bkelihood of success on the equal protection claims.
To the extent that the district court intended to employ Burdick as a template for a rational basis review of this case, we respectfully disagree with its legal analysis. We recently held in a similar context that strict scrutiny review should be ap-pbed. See Idaho Coalition United for Bears, 342 F.3d 1073, 2003 WL 22072191 at *2; see also Bullock v. Carter, 405 U.S. 134, 144, 92 S.Ct. 849, 31 L.Ed.2d 92 (1972) (holding that a case involving impact related to the resources of the voters supporting a particular candidate should be “closely scrutinized and found reasonably necessary to the accomplishment of legitimate state objectives in order to pass constitutional muster.”) (internal quotation marks omitted). Even in Burdick, the Court did not mandate the use of rational basis scrutiny; it only held that not every burden on the right to vote must be subject to strict scrutiny. 504 U.S. at 432, 112 S.Ct. 2059. Burdick clearly states that the “rigorousness” of the “inquiry into the propriety of a state election law depends upon the extent to which a challenged regulation burdens First and Fourteenth Amendment rights.... [W]hen those rights are subjected to severe restrictions, the regulation must be ‘narrowly drawn to advance a state interest of compelling importance.’ ” Id. at 434, 112 S.Ct. 2059 (quoting Norman v. Reed, 502 U.S. 279, 289, 112 S.Ct. 698, 116 L.Ed.2d 711 (1992)). Thus, Burdick did not mandate rational basis scrutiny; rather, it merely described the continuum of review appropriate in a particular circumstance. In Burdick, the plaintiffs’ First Amendment rights to ballot access were at issue. The instant action alleges deprivation of a right to have one’s vote counted, which traditionaby has been examined under strict scrutiny.
In addition, plaintiffs are likely to show a much more severe restriction on their *900rights than the restrictions present in Bur-dick. The punchcard voting system eliminates some voters’ ballots entirely. In contrast, the majority viewed the write-in ban at issue in Burdick as simply a burden on the interest in waiting until “the eleventh hour,” id. at 439, 112 S.Ct. 2059, to make a decision about whom to vote for, or the interest in having the state record a “protest vote,” id. at 438, 112 S.Ct. 2059. In the instant case, Plaintiffs are not seeking to have the State record protest votes or to wait until the last minute to make decisions. Rather, they are seeking to have the State count their votes in the first instance. Under a strict scrutiny standard of review, the State must show a compelling reason for using outdated equipment, and that using the equipment is a narrowly tailored way to meet the compelling purpose. They have not asserted any such reason. They have asserted the need to conduct the recall election within the dictates of the California Constitution, but this is more properly weighed in the manner courts normally weigh it — as a public interest argument against enjoining a specific election even in the face of an equal protection violation, rather than as an interest justifying the use of poor equipment as a general matter.
However, we need not reach the question of whether strict scrutiny should be applied in this context, because Plaintiffs have demonstrated a sufficient likelihood of success on the merits regardless of the standard of review. Plaintiffs have tendered sufficient evidence to demonstrate a likelihood of success in establishing that there is no rational basis for using voting systems that have been decertified as “unacceptable” in some counties and not others. The State is not leaving pre-scored punchcard systems in place indefinitely because it cannot afford to update them, nor does it deem pre-seored punch-card voting systems acceptable for use in California elections. Rather, the State has conceded the deficiencies in the systems and agreed to remedy the deficiencies by the next statewide election.
The only potential justification is that the California Constitution requires that a recall election be held within sixty days of certification by the Secretary of State. However, this justification has no application to placement of the initiatives on the ballot because there is no similar time constraint applicable to them, and they were originally scheduled to be placed on the March 2004 ballot. As to the gubernatorial recall vote, this rationale is also weak. Indeed, had the recall petition been certified just a month and a half later than it was, the recall election would have been scheduled to take place not within sixty to eighty days as provided in the California Constitution, art. II, § 15(a), but instead in March 2004 under the California Constitution, art. II, § 15(b). That exception provides for the efficient consolidation of a recall election with an upcoming regularly scheduled election: “A recall election may be conducted within 180 days from the date of certification ... in order that the election may be consolidated with the next regularly scheduled election.... ” The operation of this exception produces arbitrary results; because the signatures were certified seven and a half — instead of six— months in advance of the March 2004 election, this exception does not apply, and the deadline falls in early October. In essence, granting a preliminary injunction would put the election only one and a half months after the longer six-month time period provided for by the California Constitution. Regardless, the salient question for us is not whether the Plaintiffs will ultimately prevail on this question, but rather whether they have established a sufficient likelihood of success to satisfy this element of the preliminary injunction *901analysis. We conclude that they have and that the district court erred in its legal analysis. Because we conclude that Plaintiffs have established a sufficient likelihood of success on their equal protection claims, we need not examine the Plaintiffs’ likelihood of success on their Voting Act claims.
B
In the context of assessing the Plaintiffs’ likelihood of success on the merits, the district court stated that res judi-cata posed a significant obstacle to recovery, although it did not rule definitively on the question. We conclude that the district court erred in the legal analysis of this claim. For a res judicata defense to be successful, three elements must be shown: “(1) an identity of claims, (2) a final judgment on the merits, and (3) privity between parties.” Tahoe-Sierra Preservation Council, Inc. v. Tahoe Reg’l Planning Agency, 322 F.3d 1064, 1077 (9th Cir.2003); Stratosphere Litig. L.L.C. v. Grand Casinos, Inc., 298 F.3d 1137, 1143 n. 3 (9th Cir.2002); see also Owens v. Kaiser Found. Health Plan, Inc., 244 F.3d 708, 713 (9th Cir.2001).
1
The following four criteria are used to determine whether an identity of claims exists for res judicata purposes:
(1) whether rights or interests established in the prior judgment would be destroyed or impaired by prosecution of the second action; (2) whether substantially the same evidence is presented in the two actions; (3) whether the two suits involve infringement of the same right; and (4) whether the two suits arise out of the same transactional nucleus of facts.
Nordhorn v. Ladish Co., 9 F.3d 1402, 1405 (9th Cir.1993).
The first criteria for our consideration is whether the rights or interests established in Common Cause I would be destroyed or impaired by prosecution of the instant action. Rather than destroy the rights obtained in Common Cause I, the instant action would vindicate them. The consent decree obtained in Common Cause I assured the Common Cause Plaintiffs, some of whom are also plaintiffs in this action, that no pre-scored puncheard balloting would occur on initiatives presented to the electorate prior to the March 2004 elections. Setting a special election because of the unprecedented recall election altered that equation. The Plaintiffs’ action attempts to secure those rights that they thought they had obtained in Common Cause I.
The second factor is the identity of the evidence. While Plaintiffs’ evidence making out the Voting Rights Act and Equal Protection Clause violations is substantially the same as in Common Cause I, the decision to enjoin an election depends on more than successfully showing a violation of federal law, since it is well established that the public interest in going forward with the election must be part of the calculus. “In awarding or withholding immediate relief, a court is entitled to and should consider the proximity of a forthcoming election and the mechanics and complexities of state election laws, and should act and rely upon general equitable principles.” Reynolds, 377 U.S. at 585, 84 S.Ct. 1362. The evidence a court must hear to determine whether or not to enjoin this special election is very different from the evidence a court would hear in deciding whether or not to enjoin regularly scheduled elections. Thus, for the purposes of res judicata analysis, there are significant differences in critical aspects of the evidence presented.
*902As to the third factor, the two suits do involve infringement of the same right: the fundamental right to vote and have one’s vote counted. Thus, this factor favors application of res judicata.
The fourth factor is transactional identity. Traditionally, we have considered that “[t]he fourth of the factors, whether the two suits arise out of the same transactional nucleus of facts, is the most important.” Central Delta Water Agency v. United States, 306 F.3d 938, 952 (9th Cir.2002) (internal quotation marks omitted); see also Frank v. United Airlines, Inc., 216 F.3d 845, 859 (9th Cir.2000); Costantini v. Trans World Airlines, 681 F.2d 1199, 1202 (9th Cir.1982).
The two suits do not arise out of the same transactional nucleus of facts because the relevant nucleus of facts is specific to this particular election. Often, the transactional nucleus of facts differs from that in a prior suit when an as-applied challenge is brought after a facial challenge. Tahoe-Sierra Preservation Council, 322 F.3d at 1080. Analogously, Plaintiffs bring a claim that this special election should be enjoined until the punchcard systems that they allege violate federal law can be replaced. Thus, although the basic constitutional and federal challenge to use of the pre-scored punchcard system in some counties but not others remains the same, its application to this particular election requires examining a new set of evidence. This is especially so given the unique nature of this election, involving evidence that was both hypothetical and unavailable during the Common Cause I litigation.
For the same reasons, this claim could not have been brought in the Common Cause I litigation. Determining whether or not to enjoin an election based on violations of federal law requires analyzing the effects on the public interest, and this factor will vary with the particular circumstances of the election. For example, a court should consider the amount of public and private money expended and anticipated to be expended, whether any public offices will be unfilled or held beyond their term if the election is postponed, whether the items on the ballot are the sorts of questions in which there is a strong public interest in speedy resolution, and the extent to which voter confusion would be created by any particular course of action.
Additionally, the evidence that should be examined in answering these questions with respect to the two initiatives is unique, and clearly could not have been examined in the prior litigation.
Thus, even if the Common Cause Plaintiffs had attempted to litigate this claim, a court could not have resolved it because the election-specific public interest factors would have been purely speculative. The date of a hypothetical recall election, that numerous candidates would run in the election, whether it would be conducted as a special election rather than consolidated with a general election, and whether other questions would be placed on the same ballot would all be unknown. Thus, the Common Cause Plaintiffs could not have litigated the claim that equity required delaying any hypothetical recall election until the punchcard system could be replaced.
Plaintiffs here claim that, even if it is impossible to replace the punchcard systems in time, this special election should be delayed. They raise election-specific arguments with respect to the public interest, such as the fact that the two initiatives on the special ballot were originally scheduled for determination in March 2004, at the next general statewide election. Given the different evidence that must be presented to resolve this claim, the different *903transactional nucleus of facts out of which the claim arises, and the fact that it would have been impossible to litigate the claim previously, Defendants are unlikely to meet their burden of showing an identity of claims for purposes of the defense of res judicata. The district court accounted for neither the election-specific nature of the evidence in a claim to enjoin the election, nor the hypothetical nature of this particular election at the time of the Common Cause I litigation.
2
The second traditional res judica-ta analysis factor is whether a final judgment was entered in the prior action. This factor is not in dispute. A final judgment is present when parties enter a consent decree, Rein v. Providian Financial Corp., 270 F.3d 895, 903 (9th Cir.2001), and the district court specifically entered a final judgment in Common Cause I.
3
The third factor in a res judi-cata analysis is privity of the parties. Generally speaking, we consider parties to be in privity when their interests “are so closely aligned as to be virtually representative.” Tahoe-Sierra Preservation Council, 322 F.3d at 1082 (internal quotation marks omitted). However, the district court applied this standard erroneously to find privity between the NAACP, a plaintiff in this action that was not a party to the Common Cause I litigation, and the Common Cause Plaintiffs.
The district court erred in finding that privity was likely present because it did not consider recent developments narrowing the apparent scope of the virtual representation or commonality of interests theories of privity. We have affirmed that “the mere existence of litigation brought by other parties with similar interests does not bar a plaintiff from pursuing his own litigation.” Green v. City of Tucson, 255 F.3d 1086, 1100 (9th Cir.2001) (citing Richards v. Jefferson County, 517 U.S. 793, 116 S.Ct. 1761, 135 L.Ed.2d 76 (1996)). “[E]arlier litigation brought by parties with similar interests [cannot] preclude subsequent plaintiffs from bringing their own lawsuit even though they were aware of the prior litigation and shared a lawyer with the earlier plaintiffs.”- Id. at 1101 (citing S. Cent. Bell Tel. Co. v. Alabama, 526 U.S. 160, 167-68, 119 S.Ct. 1180, 143 L.Ed.2d 258 (1999)). In other words, one cannot be a representative for another for purposes of binding them to a final judgment unless there was privity or some other special relationship between the two sets of plaintiffs. Indeed, we have held that when a plaintiff himself was the counsel (but not a party) in a prior FOIA case, he could not be bound by the prior judgment. Favish v. Office of Independent Counsel, 217 F.3d 1168, 1171 (9th Cir.2000) (holding that “an abstract interest in enforcement of FOIA”, was “insufficient to create privity”).
These limitations on the virtual representation doctrine are grounded in due process. In Richards, the Supreme Court reaffirmed the holding of Hansberry v. Lee, 311 U.S. 32, 61 S.Ct. 115, 85 L.Ed. 22 (1940) at length, rejecting the notion that merely claiming to represent a broad class of persons is sufficient to bind those parties: The plaintiff in the earlier case at issue in Hansberry “had alleged that she was proceeding ‘on behalf of herself and on behalf of all other property owners in the district.’ ” Richards, 517 U.S. at 800, 116 S.Ct. 1761 (quoting Lee v. Hansberry, 372 Ill. 369, 372, 24 N.E.2d 37 (1939)).
Richards circumscribed “extreme” applications of res judicata that conflicted with due process. Id. at 797, 116 S.Ct. 1761. Ordinary litigation does not automatically *904become a mandatory class action for purposes of issue preclusion, binding all people of color in California and the many groups of which they are a part merely by virtue of the fact that public interest organizations who share some of their abstract interests in ending voting discrimination are involved. Indeed, the plaintiffs in Common Cause I did not even purport to represent all people of color in California. They only purported to represent their members. There is no showing that all members of the NAACP who are voters in California are also members of other Common Cause Plaintiff groups. Yet the rule as applied by the district court would bind them to the Common Cause Plaintiffs’ litigation strategies and choices, implying a burden of intervention on all people of color in California.
Indeed, the district court applied the virtual representation standard in an extreme manner even when compared to pre-Richards cases. The district court cited Nordhom for its “virtual representative” theory of privity, but in Nordhom we actually declined to find privity between the two relevant parties, 9 F.3d at 1405, and applied a very bounded concept of virtual representation. We provided, as an example of a situation in which privity was present, the relationship between a corporate party wholly owned by a prior litigant, id., and found there was no “similar identity of interest or control” between current and prior litigants. Id.
There is no evidence in the record that the NAACP participated in or influenced the Common Cause I litigation and negotiations in any way, or that there is any relationship between the NAACP and any of the Common Cause Plaintiffs that rises to the level of a relationship in which legal interests, as distinguished from broad social/politieal interests, are shared. The district court appeared to believe that since the Common Cause Plaintiffs and NAACP would be on the same side of the issue, and because the NAACP and some of the Common Cause Plaintiff groups have similar mission statements, they share the same “interests.” But sharing the general aim or mission of equality for people of color is very different from sharing the same legal interests. The fact that the Common Cause I Plaintiffs were willing to give up rights in exchange for a promise to replace punchcard systems by March 2004 does not indicate that the NAACP or its members were willing to do the same. The district court also highlighted the fact that the Common Cause Plaintiffs share some of the same counsel with the NAACP. Yet this factor has explicitly been deemed irrelevant to privity by recent holdings. S. Cent. Bell Tel. Co., 526 U.S. at 168, 119 S.Ct. 1180; see also Favish, 217 F.3d at 1171.
Finally, the Supreme Court and this Circuit have examined, and found the absence of, privity in voting rights cases before. In Lockport, a voting rights challenge by county residents was not barred by the county’s prior suit because the voters were not in privity with the county. Lockport v. Citizens for Cmty. Action at Local Level, Inc., 430 U.S. 259, 263 n. 7, 97 S.Ct. 1047, 51 L.Ed.2d 313 (1977). Similarly here, none of the Common Cause Plaintiffs should be able to bind the NAACP.
In sum, the record does not support a finding of privity.
4
The policy underlying the doctrine of res judicata also speaks against finding preclusion in this case. “Res judicata ensures the finality of decisions ... encouraging] reliance on judicial decisions, bar[ring] vexatious litigation, and freeing] the courts to resolve other disputes.” Brown v. Felsen, 442 U.S. 127, 131, 99 *905S.Ct. 2205, 60 L.Ed.2d 767 (1979). However, application of res judicata should, by the same token, avoid encouraging advisory adjudication. In Brown, the Court refused to apply res judicata in a bankruptcy proceeding that had been preceded by a state collection suit, noting that
[i]n the collection suit, the debtor’s bankruptcy is still hypothetical. The rule proposed by respondent would force an otherwise unwilling party to try § 17 questions to the hilt in order to protect himself against the mere possibility that a debtor might take bankruptcy in the future. In many cases, such litigation would prove, in the end, to have been entirely unnecessary, and it is not surprising that at least one state court has expressly refused to embroil itself in an advisory adjudication of this kind.
Id. at 135, 99 S.Ct. 2205.
This sort of ex ante reasoning can be easily analogized to the present case. During litigation to change state voting practices, a special recall election is entirely hypothetical. In fact, it is probably more hypothetical than many debtors’ bankruptcies. Application of res judicata to bar a motion to enjoin the special election would force plaintiffs to litigate the issue of when practices could feasibly be ' changed “to the hilt,” rather than settling for a reasonable compromise. Moreover, entertaining these questions, even if it were possible, would be advisory in the same sense that entertaining § 17 bankruptcy questions in a state collection judgment would be advisory.
For all these reasons, we conclude that the district court’s reliance on res judicata as justification for concluding that the Plaintiffs were unlikely to prevail on the merits was misplaced.
c
In the context of assessing the Plaintiffs’ likelihood of prevailing on the merits, the district court also concluded, but did not hold, that Plaintiffs’ action was likely barred by laches. The district court raised this issue sua sponte; the Defendants had not pleaded it as an affirmative defense, as required under Federal Rules of Civil Procedure 8(c). We conclude that the district court erred in its legal analysis of this claim.
“Laches is an equitable time limitation on a party’s right to bring suit.” Boone v. Mechanical Specialties Co., 609 F.2d 956, 958 (9th Cir.1979). “The affirmative defense of laches requires proof of (1) lack of diligence by the party against whom the defense is asserted, and (2) prejudice to the party asserting the defense.” In re Beaty, 306 F.3d 914, 926 (9th Cir.2002) (internal quotation marks omitted). “[L]aches is generally not a bar to prospective injunctive relief.” Jarrow Formulas, Inc. v. Nutrition Now, Inc., 304 F.3d 829, 840 (9th Cir.2002) (citing Danjaq, LLC v. Sony Corp., 263 F.3d 942, 959-60 (9th Cir.2001) and Lyons Partnership, L.P. v. Morris Costumes, Inc., 243 F.3d 789, 799 (4th Cir.2001)).
A determination of whether a party exercised unreasonable delay in filing suit requires (1) an assessment of the length of delay “measured from the time the plaintiff knew or should have known about its potential cause of action” and (2) a determination of whether the delay was reasonable. Jarrow Formulas, 304 F.3d at 838. “The reasonableness of the plaintiffs delay is considered in light of the time allotted by the analogous limitations period.” Id. (citations omitted). “We also consider whether the plaintiff has proffered a legitimate excuse for its delay.” Id.
“An indispensable element of lack of diligence is knowledge, or reason to know, of *906the legal right, assertion of which is ‘delayed.’ ” Portland Audubon Soc’y v. Lujan, 884 F.2d 1233, 1241 (9th Cir.1989) (quoting City of Davis v. Coleman, 521 F.2d 661, 667 (9th Cir.1975)). “There must, of course, have been knowledge on the part of the plaintiff of the existence of the rights, for there can be no laches in failing to assert rights of which a party is wholly ignorant, and whose existence he had no reason to apprehend.” Halstead v. Grinnan, 152 U.S. 412, 417, 14 S.Ct. 641, 38 L.Ed. 495 (1894); see also Advanced Cardiovascular Sys., Inc. v. SciMed Life Sys., Inc., 988 F.2d 1157, 1161 (Fed.Cir.1993) (“When applying the equitable doctrine of laches in order to bar a claim, the period of delay is measured from when the claimant had actual notice of the claim or would have reasonably been expected to inquire about the subject matter.”).
The instant action was not ripe until the Lieutenant Governor scheduled the special election and the Secretary of State decided to place the propositions on the ballot, less than two months ago. When the recall election was scheduled and the decision made to include the propositions on the ballot, Plaintiffs promptly filed suit. When prospective injunctive relief is sought based on new actions of a defendant, laches ordinarily does not apply. As we have observed:
[T]he general rule that laches does not bar future injunctive relief stems from a practical recognition of the interaction between the temporal components of those two doctrines. Laches stems from prejudice to the defendant occasioned by the plaintiffs past delay, but almost by definition, the plaintiffs past dilatoriness is unrelated to a defendant’s ongoing behavior that threatens future harm.
Danjaq, 263 F.3d at 959-60.
As the Fourth Circuit has similarly observed: “A prospective injunction is entered only on the basis of current, ongoing conduct that threatens future harm. Inherently, such conduct cannot be so remote in time as to justify the application of the doctrine of laches.” Lyons Partnership, 243 F.3d at 799.
The Secretary of State contends that the Plaintiffs should have been aware of the possibility of recall because recall provisions are included in the California Constitution. However, no gubernatorial recall has previously been certified for election in the history of California. As for the initiatives, the election had already been set for March 2004, so Plaintiffs had no reason to believe that the propositions would be put to the electorate prior to that time. To have anticipated the events at hand, the Plaintiffs would have necessarily had to have impressive prophetic powers. Even so, if Plaintiffs had filed an action to enjoin the special election based on their hunch that such an election would be held, it is doubtful that any court would have countenanced it. It plainly was not ripe. “A claim is not ripe for adjudication if it rests upon contingent future events that may not occur as anticipated, or indeed may not occur at all.” Texas v. United States, 523 U.S. 296, 300, 118 S.Ct. 1257, 140 L.Ed.2d 406 (1998) (internal quotation marks omitted).
In assessing whether the Secretary of State would be prejudiced by the delay, one must consider that it was the Secretary of State who initiated the recall election by certifying the sufficiency of the number of petitioners and who made the affirmative decision to alter the original schedule and place the initiatives on the special election ballot. Of course, we do not suggest anything improper in these actions. However, in assessing prejudice it would be an odd result for a party successfully to be able to claim prejudice resulting from its own initiation of events. *907In any case, the prejudice to the Secretary of State is posed by the existence of the lawsuit, not delay in asserting it; therefore, no prejudice lies within the meaning of the doctrine of laches. See Shouse, 559 F.2d at 1147 (“Difficulties caused by the pendency of a lawsuit, and not by delay in bringing the suit do not constitute prejudice within the meaning of the laches doctrine.”).
In sum, we conclude that the district court’s legal analysis concerning this claim was incorrect.
D
Because the Plaintiffs are likely to succeed on the merits of their equal protection claims and because their action is not likely to be barred by res judicata or laches, Plaintiffs have satisfied the first factor in a preliminary injunction analysis.
IV
The second factor for our preliminary injunction analysis is the degree of harm that the Plaintiffs will suffer if preliminary injunction relief is not granted. Johnson, 72 F.3d at 1430. As we have noted, the district court assumed that the Plaintiffs would suffer irreparable harm, and we agree with its analysis. Plaintiffs have tendered credible evidence that if the election is held in October, a substantial number of voters will be disenfranchised by voting in counties utilizing pre-scored puncheard technology. Thus, their votes will be diluted in comparison to the votes cast in counties with more modern technology. Once the election occurs, the harm will be irreparable because Plaintiffs are without an adequate post-election remedy. “Abridgement or dilution of a right so fundamental as the right to vote constitutes irreparable injury.” Cardona v. Oakland Unified School Dist., 785 F.Supp. 837, 840 (N.D.Cal.1992). As the Plaintiffs have noted, those whose votes are not counted by the puncheard machines are irreparably denied their vote and to have an equal say in the choices facing the electorate on Oct 7. There is no possible post-election remedy that could remedy this violation. Thus, this factor also favors the issuance of a preliminary injunction.
V
The third factor analyzed by the district court was the balance of hardships. However, it concluded that because the examination of the balance of hardships involved matters of public concern, it turned to the public interest component of the preliminary examination for analysis of balance of hardships. However, it erred as a matter of law in doing so. In a preliminary injunction analysis, the public interest factor is examined separately, not simply as a part of hardship balancing. Sammartano, 303 F.3d at 974. Further, the public interest analysis in preliminary injunction cases is focused on the impact on non-parties rather than parties. Bernhardt v. Los Angeles County, 339 F.3d 920, 931 (9th Cir.2003). Thus, the relative burdens on the parties should be examined separately as part of the balance of hardships factor.
As we have observed, the harm faced by the Plaintiffs is irreparable. The question then is the relative burden on the Secretary of State. To this, there are different answers. The burden of canceling the election entirely at this late hour will involve expense to the State. However, this is the inverse of the usual election situation. Normally, enjoining an election would require that a special election be held later, at great financial cost. But here, the election Plaintiffs seek to enjoin is itself a special election, and if enjoined, voting would occur at a regularly sched*908uled election. Thus, the great difference in cost between regularly scheduled and special elections is not as significant a factor as in the usual election case. Nevertheless, there is undoubtedly a burden and expense to the State in canceling the election, although the Secretary of State chose not to quantify this cost in his submissions.
However, simply postponing a vote on the initiatives does not pose a significant extra burden on the Secretary of State. As we have noted, the Secretary of State originally set March 2, 2004 as the date for inclusion of the propositions on a statewide ballot. That is a regularly scheduled statewide election, so the extra burden of adding the initiative to the ballot would be insignificant. Further, the Secretary of State has already mailed information to the voters on the initiatives, so no additional expense would be incurred. ,
Thus, as to the question of voting on the initiatives, the balance of hardships tips sharply in favor of granting relief to the Plaintiffs. As to the gubernatorial recall, the balance of hardships is a closer question, but, in our judgment, slightly favors the Plaintiffs.
VI
The fourth traditional factor in preliminary injunction analysis in cases such as this is the public interest. As we have recently noted:
The public interest inquiry primarily addresses impact on non-parties rather than parties. It embodies the Supreme Court’s direction that “in exercising their sound discretion, courts of equity should pay particular regard for the public consequences in employing the extraordinary remedy of injunction.” Weinberger v. Romero-Barcelo, 456 U.S. 305, 312, 102 S.Ct. 1798, 72 L.Ed.2d 91 (1982).
Bernhardt, 339 F.3d at 931-32 (internal quotation marks and citations omitted).
A
The district court placed disposi-tive weight on the public interest in complying with state election law. We agree that the Secretary of State has an interest in complying with state election law, and that this interest must be accounted for in the balance of hardships. However, the district court erred in treating this state interest as if it were a large part of the public interest. An abstract interest in strict compliance with the letter of state law is a strong state interest, but it is a less important public interest in the context of challenges to state law under the equal protection clause of the Fourteenth Amendment. Of course, the public has an interest in lively public debate, being informed of political issues, orderly elections, speed in resolving challenges to officials, confidence in fair elections, and the like, and many state election laws are designed to promote these interests. To the extent compliance with those laws promotes these important public interests, they deserve great weight in assessing which way the public interest factor points. But it is the principles and spirit of these state laws, not the necessarily the letter, that deserve weight in examining the public interest.
Indeed, if an abstract interest in compliance with state law, divorced of the democratic principles that motivate those laws, were counted as a significant part of the public interest calculus, the supremacy of the equal protection clause over state law would be undermined in a manner especially inconsistent with its purpose. Strict compliance with state law deserves weight in contexts where state interests deserve deference. See, e.g., Lee v. Kemna, 534 U.S. 362, 375, 122 S.Ct. 877, 151 L.Ed.2d 820 (2002) (stating that the practice of *909declining jurisdiction when a state court decision rests on independent and adequate state law grounds “applies with equal force whether the state-law ground is substantive or procedural”). But this is not such a context. In this context, it is the public, not the State, whose interests must be protected.
Federal law is supreme over state law. The Fourteenth Amendment was enacted during post-CM War Reconstruction, a special time in our history when the federal government took on a much greater role in protecting citizens from unjust state infringements of their rights. See, e.g., Ex parte Virginia, 100 U.S. 339, 345, 25 L.Ed. 676 (1879) (“[The Reconstruction Amendments] were intended to be, what they really are, limitations of the powers of the States and enlargements of the power of Congress.”). The public interest is an essential factor in determining whether to enjoin an election due to a likely violation of the equal protection clause, ratified to secure important rights of citizens. Thus, placing dispositive weight on compliance with state law as part of the public interest calculus would introduce a deference to state law that is entirely inappropriate in the context of the equal protection clause. It would erroneously equate state interests with public interests. The district court erred as a matter of law in so doing.
Indeed, had the view of the Secretary of State that state law is the only relevant consideration been the rule, the Supreme Court could not have reached the conclusions it did in Reynolds, Gray, Harper, and Bush. Reynolds required almost all states to redraw their electoral boundaries for electing members of state legislatures in compliance with one-man, one-vote principles. Gray struck down a state’s established voting system. Harper found a state-established poll tax unconstitutional. Most recently, in Bush, the Supreme Court found that a state’s method of counting votes offended the Constitution. In all these cases, and in many more, the public interest was measured in broader terms that were not confined to resort to state law alone.
The appropriate examination of the public interest in this context will instead place heavy weight on the principles underlying state law. Those principles of fair and efficient self governance belong in a court’s assessment of the public interest regardless of the presence of state election laws motivated by them. State election law can merely highlight for a court which of those democratic principles the people of a State hold in particularly high regard.
B
The vote on the gubernatorial recall was scheduled by the Lieutenant Governor in compliance with the requirements of the California Constitution. There is a strong public interest in holding elections as scheduled. To enjoin the election of candidates for office has the potential of disrupting government. It could well result in unfilled essential government positions. In the case of election to national office, it could result in a state not having representation in Congress. These are serious considerations. In the case of a vote on a recall petition, these concerns are considerably lessened because governmental functions will continue. California’s very short schedule for recall voting exemplifies the significant public interest in a rapid resolution to a challenge to existing leadership. Candidates have already invested considerable sums in the election, as well as time and energy, in reliance on the recall date.
But there are substantial competing public interest considerations. The public has a substantial interest in taking all reasonable steps to avoid violating its *910citizens’ rights to equal protection. We recently recognized G & V Lounge, Inc. v. Mich. Liquor Control Comm’n, 23 F.3d 1071, 1079 (6th Cir.1994), which, in reversing the denial of a preliminary injunction, noted that “it is always in the public interest to prevent the violation of a party’s constitutional rights.” Sammartano, 303 F.3d at 974. Thus, public interest strongly favors holding the recall election during the general election in March 2004 to avoid an equal protection violation. The California Constitution already permits up to a six month delay to advance the State’s interest in efficiency and convenience; the requested injunction would result in only a seven and a half month delay to cure a substantial constitutional violation. As the Supreme Court put it in Bush: “The press of time does not dimmish the constitutional concern. A desire for speed is not a general excuse for ignoring equal protection guarantees.” 531 U.S. at 108, 121 S.Ct. 525.
The State has an interest in holding a fair election — one trusted by the candidates and the voters to yield an accurate and unbiased result. The high error rate associated with the decertified machines to be used by 44 percent of the voters in October would undermine the public’s confidence in the outcome of the election. The margin of victory could well be less than the margin of error in the use of punchcard technology. This would not be the case in an election held in March 2004, when all the obsolete machines will have been totally withdrawn from use. Avoiding an election that promises to dilute the votes of any particular community — let alone communities with a disproportionately high concentration of minority voters— firmly promotes the public interest in a fair election.
In addition, although the State and counties have spent money to organize the October election, it may conserve resources by consolidating the recall with the regularly-scheduled statewide March 2004 election. Such savings undoubtedly would favor the public interest.
There are also some unique pragmatic problems associated with this election that may be alleviated by a short postponement. For example, because of the short timetable established for this election, approximately a quarter of California’s polling places — 5,000 of 20,000 — will not be ready for use and voters will be forced to vote at a different polling place. This has the potential of creating substantial voter confusion on election day. Further, the sheer number of gubernatorial candidates — there are currently 135 names on the October 2003 ballot — will make operation of the plastic guide substantially more cumbersome to use, potentially compounding the inherent problems in its use.
Further, many members of the armed forces and California National Guard members did not fill out absentee ballot requests because they did not expect to be overseas for this length of time and did not anticipate a special election. A short postponement of the recall election will serve the public interest by permitting California men and women who are serving our country overseas and who did not anticipate an October election more time to request and submit absentee ballots, thus allowing them to enjoy one of the fundamental rights for which they put themselves in harm’s way — the right to vote.
On balance, as to the gubernatorial recall vote, public interest considerations favor postponing the election.
C
The district court analyzed the public interest with respect to the gubernatorial recall, but did not assess the public in*911terest with respect to voting on the propositions. However, the questions of the gubernatorial recall and the vote on the initiatives present distinct issues of public interest. The case for postponing the election is even stronger with respect to the votes on Propositions 53 and 54. The two propositions on the special election ballot were originally scheduled to be placed on the ballot of the March 2004 election. Indeed, on July 15, 2002, more than a year ago, then Secretary of State Bill Jones issued and signed a certification placing the initiatives on the March 2, 2004 primary election ballot. The Lieutenant Governor’s formal proclamation and order of July 24, 2003 called for a special election “to determine whether Gray Davis, Governor of the State of California, shall be recalled, and if the majority vote on the question is to recall, to elect a successor.” No mention was made of the initiatives; the proclamation was limited to holding an election on only two questions: recall and the election of a successor. On July 25, 2003, the Secretary of State decided on his own to place the two initiatives on the special election ballot.
There is no urgency in obtaining a public vote on the propositions. Proposition 53, to establish the “California Twenty-First Century Infrastructure Investment Fund,” would require specified percentages of the California general fund revenues to be set aside for acquisition, construction, rehabilitation, modernization or renovation of local infrastructure. The first year affected by the proposition would be 2006.
Proposition 54 would amend the California Constitution to prohibit state and local governments from using race, ethnicity, color, or national origin to classify current or prospective students, contractors, or employees in public education, contracting, or employment operations. The initiative has an effective date of January 1, 2005.
In short, there is no justification based on the substance of the propositions for accelerating the vote on the initiatives from the originally scheduled date of March 2, 2004.
Further, scheduling a vote on the initiatives on the special election ballot would not vindicate California’s statutory election procedure. On the contrary, it would impair it and its underlying principles. In contrast to the short timeline established for a vote on gubernatorial recall, the California Constitution contemplates an extended period prior to a vote on initiatives. Indeed, the California Constitution specifies that a vote be held on initiatives “at the next general election at least 131 days after it qualifies.” Cal. Const, art 2, § 8(c). The Secretary of State has claimed that he was forced to set the vote on the initiatives at the time of the special election, but the constitutional language is disjunctive, requiring that the vote be set at the general election or a special election in the intervening period. See id. If the drafters had wished to establish this a mandatory requirement, this could have been easily accomplished by using appropriate language.
In addition to the Constitutional requirement that establishes a long time interval before a vote on qualified initiatives, the legislature has also ensured that sufficient time exists for voters to become informed about the initiatives through the distribution of explanatory ballot pamphlets. Under California election law, elections officials are required to begin mailing ballot pamphlets no less than 40 days prior to the election. Cal. Elec. Code § 9094(a). The elections officials are required to provide a copy of the final ballot pamphlet to the State printer at least 40 days before mailing is to commence, in other words, at *912least 80 days prior to the election. Cal. Elec.Code § 9082. Elections officials are also required to make the ballot pamphlet available for public inspection no less than 20 days before the copy is given to the printer, in other words, at least 100 days prior to the election. Cal. Elec.Code § 9092. One of the purposes of these laws is to allow the public sufficient time to become informed about the proposed initiatives. Another purpose is to allow time for any legal challenges to the explanatory language proposed for distribution by election officials. Indeed, the California Election Code specifically allows any elector, subject to some restrictions, to seek a writ of mandate requiring an amendment to the explanatory pamphlet if it is “false, misleading, or inconsistent with the requirements of this code.” Id.
The Secretary of State’s action in placing the propositions on the special election ballot caused elections officials to violate these provisions of state law. Elections officials began mailing ballot pamphlets 33 days before the election, which was clearly not within the period prescribed by the Election Code. The Secretary of State provided a copy of the ballot pamphlet to the printer only 37 days prior to the election rather than the required 80 days. The public was permitted to examine the pamphlet only 57 days prior to the election, rather than the required 100 days. If the effect of voter education is as significant as the Secretary of State claims, this delay could have a profound effect on the outcome of the initiative votes.
Thus, the election laws of California not only do not require that the initiatives be placed on the special election ballot; the laws will be violated if they are placed on the ballot. The significance of this for our purposes is not to analyze the statutory violation for purposes of remedying it; that is an issue of state law. Rather, the relevant question is public interest in the present context. Given the aims of the statutory framework, the public interest is clearly served by adherence to state election laws.
Thus, in considering the public interest in avoiding disenfranchisement and in complying with California law, the overwhelming factors favor granting the preliminary injunction as to the propositions. Indeed, requiring a vote on the initiatives at the special election would be antithetical to the public interest.
There is an additional public interest consideration with respect to Proposition 54. Plaintiffs have described the proposition as “racially charged.” Whether or not that description is accurate, it is clear that the proposition is of significant interest to minority voters. As we have noted, the Plaintiffs have tendered significant evidence that the use of pre-scored punehcard voting systems will disproportionately affect the minority population of California. Thus, there is a significant public interest in avoiding disproportionate disenfranchisement of the population most affected by the proposition. Given all of these factors, it is clear that the public interest is best served by holding the vote on the initiatives at its originally scheduled date, rather than on the accelerated schedule established by the Secretary of State less than two months ago.
D
In sum, in assessing the public interest, the balance falls heavily in favor of postponing the election for a few months. The choice between holding a hurried, constitutionally infirm election and one held a short time later that assures voters that the “rudimentary requirements of equal treatment and fundamental fairness are satisfied” is clear. See Bush, 531 U.S. at 109, 121 S.Ct. 525. These issues are better resolved prophylactieally than by bitter, post-election litigation over the legiti*913macy of the election, particularly where the margin of voting machine error may well exceed the margin of victory. The Supreme Court’s admonition in Bush bears re-quoting:
The press of time does not diminish the constitutional concern. A desire for speed is not a general excuse for ignoring equal protection guarantees.
Id. at 108, 121 S.Ct. 525.
In addition to the public interest factors we have discussed, we would be remiss if we did not observe that this is a critical time in our nation’s history when we are attempting to persuade the people of other nations of the value of free and open elections. Thus, we are especially mindful of the need to demonstrate our commitment to elections held fairly, free of chaos, with each citizen assured that his or her vote will be counted, and with each vote entitled to equal weight. A short postponement of the election will accomplish those aims and reinforce our national commitment to democracy.
VII
In considering all the relevant factors, we conclude that the district court erred as a matter of law in denying the preliminary injunction with respect to the vote on Propositions 53 and 54 and the gubernatorial recall. Therefore, we reverse the order of the district court. The Secretary of State is enjoined from conducting an election on any issue on October 7, 2003. In view of the pendency of the election, we direct the Clerk of Court to issue the mandate forthwith, but stay our order for seven (7) days to allow the parties to seek further relief from this decision, if they so desire.
REVERSED.