TJOFLAT, Circuit Judge, dissenting, in which BIRCH and DUBINA, Circuit Judges, join and in which CARNES, Circuit Judge, joins as to Part V. of Judge Tjoflat's dissent in Touchston v. McDermott:

I dissent. The Florida election scheme at issue is unconstitutional for the reasons set forth in my dissenting opinion in Touchston v. McDermott, No. 00-15985 (11th Cir. Dec. 6, 2000) and by Judge Carnes in his dissenting opinion.

BIRCH, Circuit Judge, dissenting, in which TJOFLAT and DUBINA, Circuit Judges, join:

While I concur in the dissenting opinions by my colleagues, Judges Tjoflat, Dubina and Carnes, my concern about the constitutional deprivations alleged in these cases is focused on the lack of standards or guiding principles in the Florida manual recount statute. Florida's statutory election scheme envisions hand recounts to be an integral part of the process, providing a check when there are "error[s] in the vote tabulation which could affect the outcome of the election." See Fla. Stat. Ann. § 102.166(5). The 1989 Florida legislature, however, abdicated its responsibility to prescribe meaningful guidelines for ensuring that any such manual recount would be conducted fairly, accurately, and uniformly. While Florida's legislature was unquestionably vested with the power under Article II, Section One of the United States Constitution to devise its own procedures for selecting the state's electors, it was also required to ensure that whatever process it established comported with the equal protection and due process requirements of the Fourteenth Amendment to that same Constitution. Other states, such as Indiana, have provided clear and definitive standards under which manual recounts are to be conducted. See Ind. Code § 3-12-1-9.5 (providing in part that chads that have been pierced count as valid votes, but those with indentations that are not separated from the ballot card do not). Absent similar clear and certain standards, Florida's manual recount scheme cannot pass constitutional muster.

Moreover, Congress, to which the electors from Florida will be ultimately certified, has established a safe harbor, 3 U.S.C. § 5, that requires that such rules and standards be established before the election. Because the 1989 Florida legislature has, in my view, abdicated its responsibility to formulate constitutionally clear and objective statutory rules and standards for the election process in Florida, it has disenfranchised voters throughout the state. The well-intended and responsible county canvassing boards across the state have been given, in legislative terms, an unfunded mandate — discern the voter's intent without any objective statutory instructions to accomplish that laudable goal. The effect of such an unguided, standardless, subjective evaluation of ballots to ascertain voter intent is to cause votes to be counted (or not to be counted) based only upon the disparate and unguided subjective opinion of a partisan (two members are elected in partisan voting) canvassing board. Since their opinions as to voter intent are standardless no meaningful judicial review is possible by a Florida court. Accordingly, by finding an abridgement to the voters' constitutional right to vote, irreparable harm is presumed and no further showing of injury need be made.

It has been said that to err is human — and humans vote. Thus, it should not be surprising that the voting process is subject to error. However, as demonstrated in the recent Presidential election, the frequency, magnitude and variety of error associated with the exercise of this sacred right of citizenship is at once astounding and deeply troubling. Moreover, the media's focus on the campaign preceding November 7, having been eclipsed by its subsequent frenzy, has left the average citizen at the least skeptical, and at the worst cynical, about our democratic institutions. Moreover, in its present incarnation, the post-election debacle that brings these cases to us for resolution may be cynically viewed by some as depicted by Congresswoman Shirley Chisholm:

[P]olitics is a beautiful fraud that has been imposed on the people for years, whose practitioners exchange gilded promises for the most valuable thing their victims own: their votes. And who benefits the most? The lawyers.

Shirley Anita Chisholm, Unbought and Unbossed, 1970. To respond in that way would be a mistake.

While our nation's citizens have every right to be concerned, exasperated, fatigued and even cynical, it is my fervent hope that from these events they will come to understand, if not appreciate, the role of government's Third Branch in the life of our precious democracy. Our basic function in this society is to provide a forum in which disputes — both great and small (although to those involved, a dispute is never "small") — can be decided in an orderly, peaceful manner; and with a high level of confidence in the outcome. Lawyers, as officers of the court, are integral to that process in our adversarial system.

The right to vote — particularly for the office of President of the United States, our Commander-In-Chief, — is one of the most central of our fundamental rights in a democracy. Accordingly, any dispute that has at its core the legitimacy of a presidential election and impacts upon every citizen's right to vote, deserves the most careful study, thought and wisdom that we can humanly bring to bear on the issues entrusted to us. Thus, I feel compelled to attest to the fact that my brother and sister judges have embraced this case with a sense of duty, concern, and conscientious hard work that is worthy of the issues before us.

Aware of the importance of these cases and the urgency attendant to the issues presented, we decided to take these disputes en banc — that is, before the entire court of twelve judges. Moreover, utilizing a procedure that we normally employ in death penalty cases, we arranged through the clerks of the district courts involved to have copies of all filings there "lodged" (i.e., copies provided) with us contemporaneously. Hence, we have been able to review and study the progress of the factual and legal matters presented in these cases from their inception. Accordingly, long before the anticipated notices of appeal were filed, formally bringing them to us, we were about the study and review of the legal issues to be resolved. Thus, the reader of our opinions in this case should understand that our time for consideration has been considerably longer than it might appear at first blush.

Just as the electorate was divided in their good faith effort to cast their votes for our nation's chief executive, the members of this court have discharged their duty to interpret the law in the context of this case in an unbiased and sincere effort. Inevitably the pundits will opine that a judge's decision is somehow linked to the political affiliation of the President that appointed the judge. While we at all levels of the judiciary have come to expect this observation we continue to regret that some "think" that is so. It may be true that a judge's judicial philosophy may reflect, to some degree, the philosophy of the appointing President — not a surprising circumstance — but to assume some sort of blind, mindless, knee-jerk response based on the politics of a judge's appointer does us and the rule of law a grave injustice. More importantly it is just wrong.

I would hope that a careful and thoughtful review of the opinions of my brothers and sisters would dispel any suggestion that their views on the important issues before us are anything but the result of days of careful study and thoughtful analysis — because these opinions are nothing less. We have done our duty. I am proud to be associated with my judicial colleagues that have been called upon to discharge their respective constitutional obligations, albeit reluctantly — both on this court and the many other state and federal courts involved. Indeed these recent events have been a civics lesson for some — particularly the young; but they have also been a reminder that our nation's system of governance has weathered the test of time and tumult; the old three-legged stool still stands erect and with sufficient strength to support the hopes and dreams of our nation's citizens.

The revered and quotable jurist, Learned Hand, once observed: "The spirit of liberty is the spirit which is not too sure that it is right . . ." While not "right" about many things, I am confident that we have given these matters the attention they justly deserve and trust that, at least, we have laid the groundwork for an informed decision by the justices of the United States Supreme Court should they exercise their judgment to hear this case. It is my hope that they do. We have done our best so that they can do their best.

DUBINA, Circuit Judge, dissenting, in which TJOFLAT and BIRCH, Circuit Judges, join:

I agree with the majority's disposition of the issues of abstention, res judicata, collateral estoppel, and mootness. I also join and concur fully in the dissenting opinions filed by Judges Tjoflat, Birch, and Carnes. I dissent from the disposition of the remaining issues discussed in the majority's opinion. Specifically, I disagree with the notion that we cannot convert the preliminary injunction and reach the merits of this case. See Thornburgh v. American College of Obstetricians Gynecologists, 476 U.S. 747, 106 S.Ct. 2169, 90 L.Ed.2d 779 (1986).

As to the merits of this case, the legal principles set forth in the cases of Moore v. Ogilvie, 394 U.S. 814, 89 S.Ct. 1493, 23 L.Ed.2d 1 (1969), and Roe v. Alabama, 43 F.3d 574 (11th Cir. 1995), govern. Based on these principles, I would reverse the judgment of the district court in this case.

CARNES, Circuit Judge, dissenting, in which TJOFLAT, BIRCH and DUBINA, Circuit Judges, join:

I agree with the Court that the lawsuits in this case and in Touchston v. McDermott, 234 F.3d 1161, are not barred by the Rooker-Feldman doctrine or by the doctrines of res judicata, collateral estoppel, or mootness, and that there is no basis for this Court to abstain. I disagree, however, with the Court's conclusion that irreparable injury has not been shown in these two cases. My disagreement with that conclusion stems from my belief that the selective manual recounts in some of the Florida counties that use the punch card system of voting violate the equal protection rights of the voters in the other punch card system counties. The harm from that violation exists and will continue so long as the results of any of those selective manual recounts are included in Florida's certified election results. Because the existence and nature of the constitutional violation is inextricably linked to the question of irreparable injury, I turn first to a discussion of the selective manual recounts in this case, and how those recounts violated the constitutional rights of the similarly situated voters who did not receive the benefit of them.

Of course, not every election dispute implicates the Constitution and justifies federal court intervention, and "[g]enerally, federal courts do not involve themselves in 'garden variety' election disputes." Roe v. Alabama, 43 F.3d 574, 580 (11th Cir. 1995) ( Roe I) (quoting Curry v. Baker, 802 F.2d 1302, 1315 (11th Cir. 1986)). But this case is more than a garden variety election dispute. It concerns more than the validity of individual ballots or the administrative details of an election. This case involves part of a state's election law designed in a way that permits or even encourages infringement of the federal constitutional rights of a large category of voters, and a claim that the law was actually applied in a way that violated those rights. Federal courts have the authority and duty to address and decide such claims. That is what the Supreme Court did in Moore v. Ogilvie, 394 U.S. 814, 89 S.Ct. 1493, 23 L.Ed.2d 1 (1969) (striking down as unconstitutional part of Illinois' method for selecting Presidential electors). That is what we did in the Roe cases. See Roe I, 43 F.3d at 580 (affirming preliminary injunction against counting votes that state trial court had ordered to be counted); Roe v. Alabama, 52 F.3d 300 (11th Cir. 1995) ( Roe II) (same); Roe v. Alabama, 68 F.3d 404 (11th Cir. 1995) ( Roe III) (same as to permanent injunction). That is what we should do in this case.

The record in this case is not replete with factual detail, but there are sufficient undisputed facts to establish a constitutional violation based upon the selective manual recounts that were undertaken in only a few punch card counties and the resulting discriminatory treatment or weighting of the votes of similarly situated voters. For present purposes, I accept as fact everything represented as fact in the affidavits filed by the Democratic Party, which is the party that requested the selective manual recounts at issue in this case, and the chief party in interest on the defendants' side, and will add to them only those facts which neither party disputes. Proceeding in that manner makes it appropriate to decide the merits and whether permanent relief should be granted in these two appeals from the denials of preliminary injunctions. See Thornburgh v. Am. Coll. of Obstetricians Gynecologists, 476 U.S. 747, 755-57, 106 S.Ct. 2169, 2176, 90 L.Ed.2d 779 (1986), overruled on unrelated grounds, Planned Parenthood v. Casey, 505 U.S. 833, 112 S.Ct. 2791, 120 L.Ed.2d 674 (1992). The Thornburgh decision establishes that a court of appeals may decide the final merits of a case in an appeal from the grant or denial of a preliminary injunction if "the facts are established or of no controlling relevance," and it is not a situation "when there is no disagreement as to the law, but the probability of success on the merits depends on facts that are likely to be developed at trial." Id. at 757 n. 8, 106 S.Ct. at 2177 n. 8. The facts that are established or undisputed in these two cases entitle the plaintiffs to relief for reasons I will explain, and thus all disputed or undeveloped facts are of "no controlling relevance."

Proceeding in this manner, the Florida Democratic Party's factual position plus the undisputed facts are these. Twenty-four of Florida's 67 counties use a vote system in which the voter's preference is expressed by punching a stylus through a card that is later passed through a tabulating machine. See Siegel, Aff. of William F. Galvin, Appendix to Brief of Florida Democratic Party ("Fla.Dem.App.") at tab 10; Chart A. There are different models of punch card tabulating machines, but all of them work by directing light at the punch card being fed through the machine and reading the beam that results from the light passing through the hole that has been punched in the card by the voter. See Siegel, Aff. of William F. Galvin, Fla. Dem.App. at tab 10. If the hole punched in the card is not clear of any chad, there is a possibility, perhaps a likelihood, that the tabulating machine will not count the vote. Id.

The failure of the punch card system to count all of the intended votes is a problem inherent in that voting system. See, e.g., Siegel, Aff. of Ion V. Sancho, Fla. Dem. App. at tab 9; Siegel, Aff. of William F. Galvin, Fla. Dem.App. at tab 10; Siegel, Aff. of Rebecca T. Mercuri, Fla. Dem.App. at tab 16. It is a serious problem that results in a significant number of intended votes not being counted; and those intended votes will remain uncounted unless there is a manual recount involving visual inspection of the punch cards by human beings. See Siegel, Aff. of Jackie Winchester, Fla. Dem.App. at tab 8; Siegel, Aff. of Ion V. Sancho, Fla. Dem.App. at tab 9; Siegel, Aff. of William F. Galvin, Fla. Dem. App. at tab 10; Siegel, Aff. of Jon Ausman, Fla. Dem.App. at tab 13; Siegel, Aff. of Rebecca T. Mercuri, Fla. Dem.App. at tab 16. While plaintiffs question whether human beings can accurately ascertain the intent of a voter by inspecting a punch card with an indented, pregnant, swinging, or otherwise not fully removed chad, the theory of the selected manual recounts undertaken in this case is that it can be done, and that as a result intended votes which would otherwise have been disregarded can and will be counted in a manual recount.

Indeed, the unwavering refrain of the Florida Democratic Party underlying its requests for manual recounts in 3 punch card counties, and throughout all of the state and federal litigation related to this case, has been that punch card systems necessarily and invariably undercount votes which can only be recaptured and considered by manual recounts. In justifying its request for manual recounts in the 3 counties, the Party told the Florida Supreme Court in a related state court case that, "It is well established that machine tabulation of votes fail (sic) to capture votes cast by a large number of voters, particularly when the number of votes cast is substantial — almost six million in the case of Florida's Presidential election. Machine tabulation of these votes, without some additional process for counting votes that the machines fail to tabulate, results in the disenfranchisement of countless voters." Answer Brief of Petitioners/Appellants Al Gore, Jr. and Florida Democratic Party at 20, Palm Beach County Canvassing Bd. v. Harris, 772 So.2d 1220 (filed in the Fla. Supreme Court Nov. 19, 2000); see also id. at 15 ("Underlying the addition of a provision for a manual recount is an understanding that the process is more accurate than machine counts, not less.") (emphasis in original); id. at 16 ("[M]any studies indicate that machine counts of punch card ballots produce significant inaccuracies.").

In the briefs the Democratic Party filed in our court in these two cases, it has told us that:

The optical scanner voting system used by most Florida counties provides good results, including a "non-vote" percentage for the Presidency (where one would expect "non-votes" to be very low) of only 0.40%. Punch card voting, by contrast, which is in effect in the three larger counties that have undertaken considerable manual recounts . . . is much less reliable, yielding an improbable "non-vote" percentage for the Presidency of over 3%.

Brief of Intervenor/Appellee Florida Democratic Party at 23-24, Touchston v. McDermott, 234 F.3d 1161 (11th Cir. 2000); see also id. at 10 ("Punch card ballots generate a consistently greater level of undervotes — approximately 3.2% — due to imperfect perforations and still-appended chads.").

The Democratic Party told the United States Supreme Court essentially the same thing: "Because of the high percentage of undervotes created by punch card voting systems, the vast majority of counties in Florida do not use them." Brief of Respondents Al Gore, Jr., and Florida Democratic Party at 4 n. 2, Bush v. Palm Beach County Canvassing Bd., No. 00-836 (filed in the United States Supreme Court Nov. 28, 2000).

Summarizing its theory of the case, the Democratic Party has said: "the evidence in this case suggests that some Florida voters could potentially be disenfranchised because the automated systems utilized in some Florida counties caused thousands of votes to go uncounted. The only means whereby those uncounted votes can be examined is to discern the intent Florida's voters is (sic) through a manual recount auditing process." Response of Intervenor/Appellee the Florida Democratic Party In Opposition to Appellants' Emergency Motion for Injunction Pending Appeal at 7, Touchston, No. 00-15985 (filed in the 11th Cir. Nov. 16, 2000). In any punch card county where manual recounts are not undertaken, the Party says, "outright disenfranchisement" occurs. See id. at 40 ("Each of the county standards employed [in the Palm Beach and Broward County manual recounts] was, thus, a vast improvement over the outright disenfranchisement that results from machine undercounts caused by hanging and dimpled chads.").

If the Florida Democratic Party's theory is not valid, then the manual recounts it requested and any change in votes resulting from those manual recounts would amount to stuffing the ballot boxes in the selected counties with illegal or non-existent votes, and counting those bogus votes would be unconstitutional for that reason. See Baker v. Carr, 369 U.S. 186, 208, 82 S.Ct. 691, 705, 7 L.Ed.2d 663 (1962) (recognizing that the right to vote is infringed by false tally or by stuffing the ballot box); Roe I, 43 F.3d at 581. But, as I have explained, the Democratic Party insists that a manual recount actually results in the counting of intended votes that would not be detected by machine, and it has put in the record numerous affidavits supporting that view. The Florida Supreme Court seems to have embraced the theory as well by interpreting "error in the vote tabulation" in Fla. Stat. § 102.166(5) to include a discrepancy between a machine count and a sample manual recount in a punch card county. See generally Palm Beach County Canvassing Bd. v. Harris, 772 So.2d 1220, 1228-30 (Fla. 2000), vacated, Bush v. Palm Beach County Canvassing Bd., 531 U.S. ___, 121 S.Ct. 471, ___ L.Ed.2d ___ (2000) (per curiam). Because the state high court did so, and because the theory is a necessary premise of the manual recounts the Party requested in the selected counties, I accept as a fact for present purposes the proposition that manual recounts of punch card ballots will result in intended votes being counted that otherwise would not have been if the process had stopped with machine tabulation.

If manual recounting had been conducted in all the counties using the punch card voting system so that all voters who were at risk of having their intended votes disregarded were protected to generally the same extent by the corrective process, there would be no federal constitutional violation, at least if we assume (as I will for purposes of this analysis) that the standards applied in the recount were accurate, consistent, and fair enough to satisfy due process. But manual recounts did not occur in all of the punch card counties. Not by a long shot. Instead, the Florida Democratic Party requested and, in conjunction with state officials and using administrative processes sanctioned by state law, brought about a selective manual recount. The process which the Party insists corrects machine errors and ensures that the will of voters is ascertained, that voters are not disenfranchised by defective technology, was requested in only 3 of Florida's 24 counties that suffer from the punch card malady, the 3 being Broward, Palm Beach, and Miami-Dade. No recount was requested or undertaken in the remaining 21 Florida punch card counties: Collier, Desoto, Dixie, Duval, Gilchrist, Glades, Hardee, Highlands, Hillsborough, Indian River, Jefferson, Lee, Madison, Marion, Nassau, Osceola, Pasco, Pinellas, Sarasota, Sumter, and Wakulla.

The manual recounts have been completed in Broward and Palm Beach counties, and the resulting additional votes from Broward County have been added to the statewide totals. Whether part or all of any corrections brought about by the manual recounts in Palm Beach and Miami-Dade Counties will be added to the statewide totals as a result of other ongoing litigation in state court remains to be seen. Given the fluidity of events, I will assume for the remainder of this opinion that the manual recount results from all 3 of the selected counties will be added to the statewide totals. However, irrespective of what is decided in the state litigation involving Palm Beach and Miami-Dade Counties, my conclusion remains the same because any difference in degree of selectivity between one, two, or three counties being manually recounted and the remainder of the 24 punch card counties not being recounted is immaterial under the applicable constitutional principles. The difference between one, two, or three manual recounts being conducted may affect the result of the election, but the Constitution forbids violations of voters' equal protection rights even when those violations do not change the outcome of the election. See infra at 812.

The voters who for whatever reason did not succeed in dislodging the chad next to their choice for President had their votes counted in Broward County and may eventually have their votes counted in the 2 other selected counties, but the voters in all of the other 21 punch card counties who applied the same pressure on the stylus and brought about the same effect, or lack of intended effect, on the chad connected with their choice for President did not have their votes counted. Under the Florida Democratic Party's theory of punch card undercounting, thousands of similarly situated Florida citizens who intended to vote for President were thwarted in their efforts by defective technology, perhaps combined with a bit of personal carelessness, and whether their intended votes count has been made to depend solely upon the county in which they live. If they live in Broward County (or maybe in Palm Beach or Miami-Dade Counties, too), their votes count; but if they live in any of the other punch card counties, they do not. The one and only difference is in which of the 24 punch card counties they live.

"A citizen's right to a vote free of arbitrary impairment by state action has been judicially recognized as a right secured by the Constitution, when such impairment resulted from dilution from a false tally, or by a refusal to count votes from arbitrarily selected precincts, or by a stuffing of the ballot box." Baker v. Carr, 369 U.S. 186, 208, 82 S.Ct. 691, 705, 7 L.Ed.2d 663 (1962) (internal citations omitted); accord Reynolds v. Sims, 377 U.S. 533, 555, 84 S.Ct. 1362, 1378, 12 L.Ed.2d 506 (1964) ("And 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.") (footnote omitted).

For at least a quarter of a century, it has been established that "[d]iluting the weight of votes because of place of residence impairs basic constitutional rights under the Fourteenth Amendment just as much as invidious discriminations based upon factors such as race, or economic status." Reynolds, 377 U.S. at 566, 84 S.Ct. at 1384 (internal citations omitted). As the Supreme Court explained in Reynolds, "Overweighting and overvaluation of the votes of those living here has the certain effect of dilution and undervaluation of the votes of those living there." Id. at 563, 84 S.Ct. at 1382. The Constitution prohibits states from weighting votes differently based on the voters' place of residence. The Supreme Court enforced this prohibition in Gray v. Sanders, 372 U.S. 368, 83 S.Ct. 801, 9 L.Ed.2d 821 (1963), when it struck down the county unit system the Georgia Democratic Party used in its primary elections. Under that complicated system every citizen got one vote, but in the final analysis some votes mattered more than others — they counted more — and the difference was based upon the counties in which the voter lived. Id. at 370-72, 83 S.Ct. at 803-04. The Court held that the Constitution prohibits such selectivity. Id. at 380-82, 83 S.Ct. at 808-09.

Another variation on selective weighting of franchise by county of residence was presented to the Court in Moore v. Ogilvie, 394 U.S. 814, 89 S.Ct. 1493, 23 L.Ed.2d 1 (1969). That case involved an Illinois law that required a candidate seeking a place on the statewide ballot to present a nominating petition containing the signatures of at least 25,000 voters. That basic requirement was not a constitutional problem, but a proviso that also required the nominating petition to include the signatures of 200 or more voters from each of at least 50 counties was a problem. Id. at 815, 89 S.Ct. at 1494. Illinois adopted that proviso in order to ensure that any candidate who got on its statewide ballot had at least minimal state-wide support, because "[a]n elected official on the state level represents all the people in the state," and "[s]uch representatives should be aware of and concerned with the problems of the whole state and not just certain portions thereof." Moore v. Shapiro, 293 F.Supp. 411, 414 (N.D.Ill. 1968) (three-judge court), rev'd sub nom. Moore v. Ogilvie, 394 U.S. 814, 89 S.Ct. 1493, 23 L.Ed.2d 1 (1969). The geographic-spread proviso in Illinois' nominating petition requirement was unquestionably "an expression of rational state policy," Moore v. Shapiro, 293 F.Supp. at 414, but that did not save it from being struck down.

The problem with the Illinois proviso, the Supreme Court explained in Moore, was that it discriminated against voters residing in the more populous counties of the state in favor of those residing in the less populous counties. The constitutional math went like this:

Under this Illinois law the electorate in 49 of the counties which contain 93.4% of the registered voters may not form a new political party and place its candidates on the ballot. Yet 25,000 of the remaining 6.6% of registered voters properly distributed among the 53 remaining counties may form a new party to elect candidates to office. . . . It, therefore, lacks the equality to which the exercise of political rights is entitled under the Fourteenth Amendment.

Id. at 819, 89 S.Ct. at 1496. Although the selective weighting of the franchise accomplished by the proviso involved in Moore was more sophisticated and less direct, and as a result less obvious, than the laws struck down in Reynolds v. Sims, it still failed to "pass muster against the charges of discrimination or of abridgement of the right to vote." Moore, 394 U.S. at 818, 89 S.Ct. at 1496.

Given the fertility of the human mind when focused upon political objectives, denial or debasement of the franchise can be accomplished in myriad ways. But whatever the method or means used to count, weigh, or value some votes differently from others, however sophisticated or indirect the device, the Constitution is up to the task. See Reynolds, 377 U.S. at 563, 84 S.Ct. at 1382 ("One must be ever aware that the Constitution forbids 'sophisticated as well as simpleminded modes of discrimination.'") (citation omitted). Because of the central importance of the right to vote in our system of representative democracy, "any alleged infringement of the right of citizens to vote must be carefully and meticulously scrutinized," id. at 562, 84 S.Ct. at 1381, and that is the duty of the courts.

Of course, many cases dealing with sophisticated debasements of the right to vote have political overtones, and that is no less true here than usual. The Supreme Court was presented in Reynolds with the argument that it ought to stay its hand and keep out of the political thickets involved in that case. To that suggestion the Court responded: "Our answer is this: a denial of constitutionally protected rights demands judicial protection; our oath and our office require no less of us." Id. at 566, 84 S.Ct. at 1384. That is a good answer.

In order to apply the principles of these decisions to the facts of the cases before us, I turn now to a closer examination of the selection of the 3 counties in which a manual recount was requested. Acting pursuant to Fla. Stat. § 102.166(4), the Democratic Party filed written requests for manual recounts in Broward, Palm Beach, and Miami-Dade Counties, and no other punch card counties. Siegel, Fla. Dem.App. at tabs 1, 3 5. There were two common grounds stated in each of those 3 written requests. One ground given in all 3 requests was that the punch card system with its chads created a risk that intended votes had not been counted ("undervotes") or actually did result in undervotes, a problem the requests said could be corrected by a manual recount with its attendant visual inspection of the cards. The other stated ground in all 3 requests was that the election results in Florida showed that the race for President was very close. No other grounds were given in the manual recount requests. See id.

The problem with machine tabulating of punch card ballots is common to counties that use the punch card system. The Democratic Party has never contended to the contrary, but instead has insisted that the problem is inherent in punch card technology. For that reason, the existence of a punch card voting system cannot be a basis for differentiating the 3 counties that were selected from the 21 that were not. And, of course, the fact that the statewide totals in the race for President were extremely close was a common fact, and therefore could have served as grounds for a recount in any of the other 21 punch card counties. There is nothing in the reasons that the Party gave for requesting a manual recount in the 3 selected counties that explains, let alone justifies, the discrimination in favor of those 3 punch card counties and against the other 21. In order to give the Party the benefit of the doubt and to consider all the possibilities, I will now look elsewhere for an explanation.

Charts A-F, which are attached as appendices to this opinion, contain population and other demographic data, as well as relevant vote data on a county-by-county basis. That vote data represents things as they stood on November 9, 2000, after the automatic machine recount required under Fla. Stat. § 102.141(4) had been conducted. That is the relevant vote data for our purposes, because it reflects the facts as of the time the Florida Democratic Party filed its manual recount requests in Broward, Miami-Dade, and Palm Beach Counties on November 9, 2000.

Chart A shows that the 3 counties selected by the Democratic Party for a manual recount share these characteristics: 1) they are the 3 most populous counties in the State of Florida; 2) they are the 3 counties in which the Party's nominee, Vice-President Al Gore, received the largest number of votes; and 3) in each of them he received substantially more votes than his opponent, Governor Bush.

The theory underlying the manual recount, as I have already explained, has always been that the punch card system of voting necessarily and inevitably results in some intended votes not being picked up by the tabulating machine. The Florida Democratic Party has never suggested that its selection of counties for manual recounts was based upon any county-by-county variation in either the way the punch card system operates or in its rate of accuracy. Instead, the consistent position of the Party, which is generally supported by the affidavits it submitted in the district court, is that every time the punch card system is used there will be intended votes that are not counted by the tabulating machine. See supra at 1195-97. Given the stated justification that the manual recounts were necessary in Broward, Miami-Dade, and Palm Beach Counties because those counties used the punch card system, the more relevant focus is on the population and voting data from all of Florida's 24 punch card counties. Chart B shows that data. Of course, because the 24 punch card counties are a subset of all of Florida's 67 counties, the characteristics that distinguish the 3 counties chosen by the Party on a statewide basis also distinguish them in relation to the other 21 punch card counties: those 3 are the most populous and vote-rich of all the punch-card counties, and in each of them the Party's nominee received substantially more votes than his principal opponent.

Not only that, but we also see from the data contained in Chart B another conspicuous fact. The 3 counties the Florida Democratic Party selected for manual recounts are 3 of the 4 punch-card counties that gave its nominee the highest percentage of the vote cast among the two opposing Presidential candidates. Those percentages are as follows: Broward (68.55%); Palm Beach (63.81%); and Miami-Dade (53.18%). No other punch card county gave the Party's nominee a greater percentage of its vote than Broward and Palm Beach Counties, and only one punch card county gave the Party's nominee a greater percentage of its vote than Miami-Dade County did. That lone exception is sparsely populated Jefferson County which, although favoring the Party's nominee with 55.10% of its vote, cast a total of only 5,519 votes for the nominees of both major parties (compared, for example, to the 618,335 votes cast for them in Miami-Dade County). Because so few votes were cast in Jefferson County, that county offered little prospect for finding enough uncounted votes to make a difference. In effect, the voters of Jefferson County were too few in number to matter in view of the Party's objective, which was to change the election result that had been reported to that date.

Given the theory of the recount — finding intended votes that were not counted by the punch card system — the most relevant data of all would be the percentage of votes that were intended but not counted. We do not have that, but neither did the Florida Democratic Party when it selected the punch card counties in which to request recounts. We do have the "no vote" data, which shows the difference between the total number of voters who cast a ballot and the total votes cast for any Presidential candidate. In other words, the no vote data shows the number of ballots in which no vote for President was counted either because the tabulating machine did not pick up from the punch card any vote for President, or because it picked up two or more votes for President on the same card resulting in no vote for President being counted.

Chart C ranks the punch card counties by percentage of no votes in the Presidential race. If Palm Beach, Miami-Dade, and Broward Counties had been selected for manual recounts because of problems resulting in no vote for President being picked up by the tabulating machines, those 3 counties would have the highest no vote rates. They do not. Chart C shows that there are 7 punch card counties with a higher percentage of no votes in the Presidential race than Palm Beach County, yet none of them was selected for manual recounts. The chart also shows that 10 punch card counties have a higher percentage of no votes than Miami-Dade County, but none of them was selected for a manual recount. And as for Broward County, there were 17 punch card counties with a higher no vote rate that were not selected for manual recounts. In fact, Broward is tied for the fourth smallest percentage of no votes for President among all of the 24 punch card counties, yet the Florida Democratic Party still selected it for a manual recount.

One of the many affidavits the Florida Democratic Party submitted in the district court stated that "two groups of citizens, the elderly and minorities, are more prone to have problems on this system than the rest of the population." Siegel, Aff. of Ion V. Sancho, Fla. Dem.App. at tab 9. Perhaps that opinion rests upon derogatory stereotypes that federal courts should not countenance. Even assuming, however, that there is some factual basis for that opinion and that we should consider the possibility, the problems that any group, including the elderly and minorities, have with punch card voting should be captured to some extent in the no vote data contained in Chart C. But as we have seen, the Party's selection of the 3 counties cannot be justified on the basis of that data.

Moreover, Chart D, which ranks the punch card counties by percentage of population over the age of 65, shows that 7 of those counties that were not selected for manual recounts have a greater percentage of their population in that age category than Palm Beach County does; 11 not selected for manual recounts have a greater percentage in that age category than Broward County does; and 13 of them have a greater percentage in that age category than Miami-Dade County does. The Florida Democratic Party's selection of punch card counties for manual recounts could not have been based upon the percentage of elderly in each county's population.

As for "minorities" having more problems with punch card voting, it is unclear exactly what the Florida Democratic Party's affiant meant by "minorities." Chart E shows that if he meant to include both blacks and Hispanics in that grouping, Miami-Dade County's population does have a higher percentage of minorities than any other punch card county. But the chart also shows that 6 punch card counties that were not selected for manual recounts have a higher percentage of minorities in their populations than Broward County, which was selected. And it shows that 8 punch card counties that were not selected for manual recounts have a higher percentage of minorities in their population than Palm Beach County which was also selected.

So, the facts we have about the Florida Democratic Party's selection of the counties in which a manual recount would be undertaken in order to ensure that voters were not disenfranchised by systemic problems with punch card technology or by carelessness, are these. The selection was not based upon the rate of punch card error — the no vote rate — nor was it based upon the relative percentage of senior citizens or minorities in each county's population. Instead, the defining characteristic of the 3 punch card counties chosen to undertake a manual recount is that they are the 3 most populous counties in the state, all of which gave the Party's Presidential nominee a higher percentage of the vote than his opponent.

Of course, none of this is surprising. We expect political parties to act in their own best political interest, and the 3 most populous counties that had voted for its nominee presented the Florida Democratic Party with its best prospects for turning the election around. It would not have served the Party's goal of electing its nominee for President for it to have sought the intended but unsuccessful votes in those punch card counties that went for the other party's nominee, Governor George W. Bush. The voters in 17 of the 24 punch card counties favored Governor Bush. See Chart B. Examples include Hillsborough County (51.6% of its 350,317 Bush/Gore votes went for Bush) and Collier County (66.89% of its 90,351 Bush/Gore votes went for Bush). Id. Making sure that every intended vote was counted in those 17 counties that favored Bush over Gore, over two-thirds of the total number of punch card counties, was not the way for the Florida Democratic Party to get its candidate elected.

Nor would it have been efficient for the Florida Democratic Party to expend its manual recount efforts in vote-poor counties like Jefferson, whose voters did express a pronounced preference for the Party's nominee. Loyal Democrats though they may be, the citizens of Jefferson County suffered from the misfortune of living in a county whose population was so small that the total votes it cast for the two principal candidates for President were only 1.31% of those cast in Palm Beach County, only .98% of those cast in Broward County, and only .89% of those cast in Miami-Dade County. That is too few to have mattered when it came to the Party's goal of changing the results of the statewide election.

There may have been another factor at work in the Florida Democratic Party's selection of the 3 most populous counties as the ones in which to request a manual recount. State law encourages, if not requires, manual recount choices to favor counties with greater vote totals over those with lesser vote totals. Under the statute, once a sample recount of at least 3 precincts and 1 percent of the votes cast in the county has been conducted, the county canvassing board can manually recount all the ballots only "[i]f the manual recount [sample] indicates an error in the vote tabulation which could affect the outcome of the election." Fla. Stat. § 102.166(5). Of course, the larger the number of votes in a county the greater the likelihood that a complete manual recount in that county alone will affect the election, and under § 102.166(5) that appears to be the measuring rod for undertaking a complete manual recount. Because the number of votes obviously varies in relation to a county's population, there is a greater likelihood that a complete manual recount in a more populous county will change the election result. Since the possibility of a different statewide result appears to be a prerequisite for a complete manual recount in a county, the statute encourages and, in some cases — where the pre-manual recount statewide difference in votes is larger than the votes that could be picked up by a full manual recount in a less populated county — may require discrimination against less-populous counties.

Consider the present case. After the statewide machine recount mandated by Florida law, the statewide difference between the two Presidential candidates was 300 votes. It would be far easier for the Florida Democratic Party to show that that margin could be erased by a manual recount in heavily populated Miami-Dade County, which had reported a total of 618,335 votes for the two candidates, than it would be for the Party to show the same thing in sparsely populated Jefferson County, where only 5,519 votes were cast for the two candidates. In fact, depending upon the initial margin of victory, it could well be impossible to get a complete manual recount in many of the punch card counties, regardless of which candidate the voters in that county favored.

It may be that the Florida Democratic Party would have chosen the 3 punch card counties it did even without the requirement in Fla. Stat. § 102.166(5) that the sample recount conducted in each county show that the outcome of the election could be changed by continuing the recount in that county. Somewhat to its credit, the Party has never denied (at least not in federal court during litigation of these two cases) that it chose for manual recounts the 3 counties that it did, and not others, because those counties are populous, i.e., vote rich, and their voters had expressed a preference for its Presidential nominee. In our Court alone, the Party filed over 180 pages of briefs and used more than 40 minutes of oral argument time to explain its position. Never once in its briefs or in its oral arguments did the Party suggest that its selection of the 3 punch card counties out of 24 for a manual recount was based on anything other than partisan self-interest. That the Democratic Party predictably acted in its own best interests in using the state law recount machinery to ensure that intended votes which would otherwise be disregarded would only be counted in counties favoring its candidate does not end the inquiry. There is the matter of the Constitution.

When a political party uses state machinery and exercises prerogatives it is given under state law to influence the counting or alter the effect of votes, it is a state actor subject to the same constitutional constraints that protect citizens from the state and its officials. See Terry v. Adams, 345 U.S. 461, 481, 73 S.Ct. 809, 819, 97 L.Ed. 1152 (1953) (white primary case) ("[A]ny part of the machinery for choosing officials becomes subject to the Constitution's restraints.") (citations and quotations omitted). The manual recount provision contained in Fla. Stat. § 102.166(4), and the selectivity it encourages or permits political parties to exercise in bringing about recounts, is an integral part of the election process in Florida, as we have seen in recent days, and the Supreme Court has held that "[a]ll procedures used by a State as an integral part of the election process must pass muster against the charges of discrimination or of abridgement of the right to vote." Moore v. Ogilvie, 394 U.S. 814, 818, 89 S.Ct. 1493, 1495-96, 23 L.Ed.2d 1 (1969).

The Florida manual recount statute gives government officials some discretion over whether to conduct a manual recount, see Fla. Stat. § 102.166(4)(c) ("The county canvassing board may authorize a manual recount"), and government officials are intimately involved in the actual recount procedure itself. Those two facts reinforce the conclusion that the Florida Democratic Party's selection of the counties in which manual recounts could occur is state action subject to constitutional scrutiny. See Dennis v. Sparks, 449 U.S. 24, 27-28, 101 S.Ct. 183, 186, 66 L.Ed.2d 185 (1980) ("[T]o act 'under color of state law' for § 1983 purposes does not require that the defendant be an officer of the State. It is enough that he is a willful participant in joint action with the State or its agents. Private persons, jointly engaged with state officials in the challenged action are acting . . . 'under color' of law for purposes of § 1983 actions.") (citation omitted); Gray v. Sanders, 372 U.S. 368, 374-75, 83 S.Ct. 801, 805, 9 L.Ed.2d 821 (1963) ("We agree with that result and conclude that state regulation of this preliminary phase of the election process makes it state action.") (citation omitted). What the State of Florida and its officials cannot constitutionally do alone, the State and the Florida Democratic Party acting jointly cannot do either.

If Florida enacted a statute that provided a manual recount procedure for correcting the undervote caused by the use of the punch card voting system, but provided that the corrective procedure could be invoked only in the 3 most populous counties of the state, no one would question that such a provision would be unconstitutional. And it would be unconstitutional no matter how rational the purpose of the statute. Suppose, for example, that the state thought it was more efficient to conduct manual recounts in the really big punch card counties, and not worth the effort to do it in any little, sparsely populated, or vote-poor punch card counties. I hope that no judge on this Court would suggest such a law would be constitutionally permissible.

The reason we would or should be unanimous in holding such a law unconstitutional is that states cannot treat votes differently depending upon the counties in which the voters live. The constitutional wrong in that hypothetical case and in the present case is the mirror image of the one in Moore v. Ogilvie, 394 U.S. 814, 89 S.Ct. 1493, 23 L.Ed.2d 1 (1969). Just as the Constitution forbids a state from counting or weighting votes less because they come from more populated counties, it also forbids a state from counting or weighting votes less because they come from more sparsely populated counties. Yet that is precisely what the manual recounts in the 3 selected Florida counties does.

Recall that the central fact underlying the theory behind the manual recounts in Broward, Palm Beach, and Miami-Dade Counties is that the punch card system of voting necessarily and inevitably results in some intended votes not being counted unless there is a manual recount. See supra at 1195-97. With the selective manual recounts that the Florida Democratic Party and government officials have jointly brought about, voters are treated differently depending upon where they live. There are two sets of punch card voters whose efforts to vote are not picked up by the tabulating machines. One set, the favored one, lives in Broward, Palm Beach, and Miami-Dade Counties. The second set has the misfortune to live in the other 21 punch card counties. The votes of the first set count; the votes of the second set do not. Two voters using the same effort to press an identical stylus against a punch card and bringing about the identical effect on a chad next to a Presidential candidate are treated differently. See O'Brien v. Skinner, 414 U.S. 524, 529, 94 S.Ct. 740, 743, 38 L.Ed.2d 702 (1974) (holding unconstitutional a statute under which two citizens "sitting side by side in the same cell may receive different treatment as to voting rights"). One vote is counted, the other not. The sole reason is that the Florida Democratic Party, acting with the authority given to it by the state, and pursuing its own political interests, chose to have one vote counted and the other not.

The matter was aptly put in a letter Florida Attorney General Robert Butterworth wrote to the Chair of the Palm Beach County Canvassing Board on November 14, 2000. The letter referred to the "extremely serious" legal issues that would arise if manual recounts were conducted in some counties but not others. He said that "a two-tier system for reporting votes would result," and:

A two-tier system would have the effect of treating voters differently, depending upon what county they voted in. A voter in a county where a manual count was conducted would benefit from having a better chance of having his or her vote actually counted than a voter in a county where a hand count was halted.

Touchston, Hearing, Ex., Trans. at 9-16, 44-45 48. That is exactly the situation resulting from the Florida Democratic Party and Florida's state or local officials acting jointly to manually recount votes in only 3 of the 24 punch card counties. In that letter, Attorney General Butterworth went on to say that he felt "a duty to warn that if the final certified total for balloting in the State of Florida includes figures generated from this two-tier system of differing behavior by official canvassing boards, the State will incur a legal jeopardy, under both the U.S. and state constitutions." That "legal jeopardy" under the United States Constitution is what this litigation is about.

If we accept what the Florida Democratic Party has told us, we can even put an estimate on the number of affected voters who are being discriminated against in the manual recount: the number who tried to vote for a Presidential candidate but were prevented from doing so by the punch card system and for whom no effort is being made to ascertain their true intent. The Party says that the optical scanner system used in most Florida counties provides good results and the undervote in counties using that system is only .40%, which the Party says is about what we should expect to occur naturally, i.e., by virtue of voter intent, in a Presidential election in Florida. Brief of Intervenor/Appellee Florida Democratic Party at 23-24, Touchston v. McDermott, No. 00-15985 (filed in the 11th Cir. Nov. 28, 2000). Yet the undervote in punch card counties, the Party says, is approximately 3.2%. Id. at 10. Thus, the difference in the undervote rate caused by the punch card system, if we accept the Party's figures, is approximately 2.8%. The total number of ballots cast in the 21 punch card counties in which no manual recount is being conducted is 2,013,666. See Chart C.

Applying the Party-supplied machine-caused-undervote rate of 2.8% to that figure gives us an estimated 56,382 voters in the non-selected punch card counties who tried to cast their votes but were thwarted by chad problems of one kind or another. It is those more than 56,000 voters whom the Florida Democratic Party, in conjunction with the state, is discriminating against in its selective manual recount. Unlike their similarly situated fellow citizens in the 3 most populous counties, no effort is being made to ascertain their true intent — thereby re-enfranchising those whose attempts to vote were thwarted by defects in the technology — by manually inspecting their punch card ballots. As the Supreme Court held in Reynolds v. Sims, "[w]eighting the votes of citizens differently, by any method or means, merely because of where they happen to reside, hardly seems justifiable. One must be ever aware that the Constitution forbids 'sophisticated as well as simpleminded modes of discrimination.'" 377 U.S. at 563, 84 S.Ct. at 1382 (citations omitted).

The same analysis applies and the same conclusion is reached, of course, if one views the selection factor as being not the population of the counties but instead the number or percentage of votes cast for the Florida Democratic Party's nominee in the counties (both factors coincided here). Just as a state, and a political party acting in conjunction with the state, cannot discriminate among voters based upon the population of their counties, so also they may not discriminate among voters based upon political opinions and beliefs as expressed by the candidates for whom those voters cast their ballots. Shifting the focus of the selection from population to political preference simply adds the weight of the First Amendment to that of the Equal Protection Clause in prohibiting the selectivity. Either way there is unconstitutional discrimination against the voters in the punch card counties not selected for manual recounts. "Their right to vote is simply not the same right to vote as that of those living in a favored part of the State." Reynolds, 377 U.S. at 563, 84 S.Ct. at 1382.

In the face of the constitutional command that votes be treated and weighted the same regardless of where the voter lives within a state, various of the defendants respond with several arguments. One thing they argue is that states are due deference in the way they run elections and, in light of Article II, § 1, cl. 2 of the Constitution, and 3 U.S.C. § 5, states are due special deference when it comes to the selection of electors. But states are due no deference if they go about selecting electors in a way that violates specific provisions of the Constitution, including the Equal Protection Clause. The Supreme Court has expressly held that the power that Article II gives the states to select electors cannot be exercised in a way that violates the Equal Protection Clause. See Williams v. Rhodes, 393 U.S. 23, 29, 89 S.Ct. 5, 9-10, 21 L.Ed.2d 24 (1968) ("Nor can it be thought that the power to select electors could be exercised in such a way as to violate express constitutional commands that specifically bar States from passing certain kinds of laws. . . . We therefore hold that no State can pass a law regulating elections that violates the Fourteenth Amendment's command that 'No State shall . . . deny to any person . . . the equal protection of the laws.'"); accord, Anderson v. Celebrezze, 460 U.S. 780, 795 n. 18, 103 S.Ct. 1564, 1573 n. 18, 75 L.Ed.2d 547 (1983). After all, Moore v. Ogilvie, 394 U.S. 814, 89 S.Ct. 1493, 23 L.Ed.2d 1 (1969), applied the one person, one vote doctrine to strike down an Illinois statute in a case involving the selection of electors. The issue is not about Article II or 3 U.S.C. § 5; it is about whether the selective manual recounts in question violate the Constitution. Because they do, nothing in Article II and certainly nothing in any federal statute insulate that unconstitutional action from remedy.

Getting closer to the merits issue, the defendants also argue that Florida law permits any political party with a candidate on the ballot, or any candidate whose name appears on the ballot, to file a written request with the county canvassing board for a manual recount. See Fla. Stat. § 102.166(4)(a). There is no equal protection problem, they say, because the Republican Party or its candidate could have requested that manual recounts be conducted in each of the punch card counties. This argument is not at all persuasive.

As I have already explained, although the Republican Party or its candidate could have requested a manual recount in any of Florida's counties, the statute permits full manual recounts in only those counties in which a sample manual recount indicates "an error in the vote tabulation which could affect the outcome of the election." Fla. Stat. § 102.166(5). Some of the punch card counties are so sparsely populated, so vote poor, that even if a manual recount had been requested and a sample recount conducted as provided in Fla. Stat. § 102.166(4)(d), the result of that sample recount would not have indicated that a full manual recount in the county could affect the outcome of the election. So, even if the Republican Party or its candidate had requested manual recounts in every punch card county, the process would still have ended up treating some punch card voters differently based upon the counties in which they lived. The Constitution forbids that.

There is a another, more fundamental flaw in the argument that treating punch card voters differently depending upon the county of their residence is permissible because the Republican Party or its candidate could have, but did not, prevent that difference in treatment. The constitutional rights involved are those of the voters in the other punch card counties. It is their votes and their constitutional rights at stake. The voters whose constitutional rights are being violated are not permitted to request a manual recount. See Fla. Stat. § 102.166(4)(a). There is no loophole in the Constitution that permits what would otherwise be an unconstitutional action to occur simply because a third party could have, but did not, prevent it from occurring. Therefore, the fact that both parties were permitted to request manual recounts does not shield the selective recounts from constitutional attack.

Another argument the defendants put forward responds to the criticism of the previous one. Florida Attorney General Butterworth, who was so concerned in his November 14, 2000 letter about the "legal jeopardy" that his state would be in if there was a "two-tier" system in which manual recounts occurred in some counties but not others, a fortnight later filed a brief in this Court telling us there is nothing to worry about after all. According to Attorney General Butterworth's latest position on the subject, manual recounts can be requested or granted under Fla. Stat. 102.166(4)(a)-(c) in as selective or discriminatory a way as the human mind can imagine without running afoul of the Constitution. The reason, he says, is that although a voter cannot request a manual recount at that stage of the election process, a voter can later file an election contest and try to get the court to conduct a manual recount as part of that contest.

That argument is unpersuasive. Even assuming that Florida law provides a mechanism for individual voters to request manual recounts as part of an election contest, the practical and legal burdens imposed upon an individual who seeks to contest an election are entirely different, and far more burdensome, than those that a party or candidate must meet in order to obtain a manual recount. A request filed by a political party or candidate before the results are certified merely has to set out grounds for a manual recount, and the county canvassing board can grant it. Fla. Stat. § 102.166(4). An election contest, on the other hand, cannot be filed until after the last county canvassing board certifies results, see Fla. Stat. § 102.168(2), and once it does, a presumption kicks in and weighs against granting any relief in the contest. Under Florida law, "elected officials are presumed to perform their duties in a proper and lawful manner in the absence of a sufficient showing to the contrary," and "there is a presumption that returns certified by election officials are presumed to be correct." Boardman v. Esteva, 323 So.2d 259, 268 (Fla. 1976) (citation omitted).

Besides, there is the problem of time. Election contests cannot be instituted until "after midnight of the date the last county canvassing board empowered to canvass the returns certifies the results of the election being contested." Fla. Stat. § 102.168(2). That might be enough time in ordinary circumstances to file a contest, have it litigated through the trial and appellate stages of the state court system, win the right to a manual recount, have any issues arising in that manual recount be litigated to conclusion, and have the new result accepted. Maybe, but the circumstances giving rise to these cases are not ordinary. To begin with, the effective deadline in this case is not some time next year as it might be with most elections, but instead is December 12, and the drop-dead deadline is December 18, 2000. Not only that, but the Florida Supreme Court extended the time for the last county canvassing board to certify its results to the Secretary of State from 7 days after the election, the time specified in Fla. Stat. §§ 102.111 and 102.112, until November 26, 2000, which is 19 days after the election. See Harris, 772 So.2d at 1240, vacated, Bush v. Palm Beach County Canvassing Bd., 531 U.S. ___, 121 S.Ct. 471, ___ L.Ed.2d ___ (2000) (per curiam). That cut 12 days out of the period that would otherwise have been provided for conducting an election contest through to conclusion.

We know from the inability of Miami-Dade and Palm Beach Counties to finish the actual manual recounts in even the extended time the Florida Supreme Court allotted them, that it would have been impossible as a practical matter for a voter in, for example, Hillsborough County, a punch card county in which 369,467 ballots were cast in the Presidential election, see Chart C, to file an election contest demanding a manual recount in that county, try the case before the trial court, succeed on appeal in time for the canvassing board to conduct and complete a full manual recount, and then have any issues arising in that recount decided. An election contest under Florida law is not a practical remedy for voters who have been discriminated against in the Florida Democratic Party's selection of punch card counties in which to request a manual recount.

Even if there were enough time for such manual recounts after the extended period for the county canvassing boards to report, there is another serious obstacle to a voter using the Florida election contest procedures to secure a manual recount in that voter's county. Except in cases of outright fraud, bribery, or other corruption, or the ineligibility for office of the successful candidate, Florida law requires that anyone filing an election contest show that correction of the problem complained about would change the results of the election. See Fla. Stat. § 102.168(3)(c) ("Receipt of a number of illegal votes or rejection of a number of legal votes sufficient to change or place in doubt the result of the election.") (3)(e) ("Any other cause or allegation which, if sustained, would show that a person other than the successful candidate was the person duly nominated or elected to the office."). If the voter cannot show that the constitutional violation he suffered changed the result of the election, he has no grounds for contesting the election under the Florida election statute.

While Florida's interest in bottom line election results is certainly expedient, the Constitution demands more than expediency. It is concerned with values other than the outcome of elections. To say that it is sufficient to remedy only those constitutional violations that matter to the political parties and their candidates is to say the rights of voters themselves do not matter. Can anyone seriously suggest that the Reynolds v. Sims, Gray v. Sanders, and Moore v. Ogilvie doctrines apply only when election results would be changed? When the Supreme Court in Reynolds said, "[t]o the extent that a citizen's right to vote is debased, he is that much less a citizen," 377 U.S. at 567, 84 S.Ct. at 1384, the Court did not add "unless it makes no difference in the election results." When the Court said 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," id., surely the Court did not mean for that basic principle to be inapplicable except where it was outcome determinative for a candidate.

In Moore there was "absolutely no indication in the record that the appellants could not, if they had made the effort, have easily satisfied Illinois' 50-county, 200-signature requirement," see 394 U.S. at 820-21, 89 S.Ct. at 1497 (Stewart, J., dissenting). In other words, there was absolutely no indication that the differential treatment of citizens based upon the counties in which they lived affected whether any would-be candidate could get on the ballot. Nonetheless, the Supreme Court did not hesitate to strike down the discrimination among voters, explaining that "[t]he 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 819, 89 S.Ct. at 1496. The one person, one vote principle is not so fickle as to depend upon the closeness of an election.

One last argument relating to the merits which is put forward by several of the defendants is that there is no constitutional violation in selective manual recounts based upon county of residence, because there are variations among the counties in election systems and different systems give rise to different error rates. In other words, unless the Constitution mandates that every county use the same voting system, it logically cannot prohibit selective correction of error rates in counties that use the same system. But why not? Why are differences in the number of vote errors that occur as a result of local variations in choice of vote systems before an election the constitutional equivalent of selective correction of errors based upon county of residence after the election?

There is no reason to believe that any county would attempt to choose for itself a voting system with a high error rate in order to disadvantage its citizens compared to those of other counties. There is every reason to believe that political parties or candidates will selectively choose the counties in which to initiate the process of manual recounts based upon how those counties voted and their population. The intent behind the two actions is different. To understand the importance of that difference, consider this hypothetical. Suppose a state legislature mandated the type of voting systems to be used in each county, and deliberately favored urban counties with low-error systems that would keep down the undervote, while sticking rural counties with high-error systems that would increase the undervote in those counties thereby reducing their influence in statewide elections. Maybe the legislature, dominated by members from the more populous counties, just wanted to keep the country folks in their place. Is there any doubt that such legislation would be unconstitutional under Reynolds and related cases? It would be unconstitutional even though the discriminatory choice occurred on the front end, before the election, and even though it involved variations in the vote systems used in different counties.

How then can it be constitutionally permissible to make a materially similar, discriminatory choice on the back end after the election: to favor the voters of more populous counties who went for one candidate with a process that ameliorates their undervote, while not applying that process to ameliorate the same or worse undervote problems in less populous counties that went for the other candidate? The answer is that it is not constitutionally permissible to discriminate in favor of the voters of Broward, Palm Beach, and Miami-Dade, or any combination of those counties, and against the voters in the other 21 punch card counties when it comes to a post-election remedy of the undervote problem caused by the voting system technology.

The Florida Supreme Court reminded us that: "Courts must not lose sight of the fundamental purpose of election laws: The laws are intended to facilitate and safeguard the right of each voter to express his or her will in the context of our representative democracy." Harris, 772 So.2d at 1237 (footnote omitted). But we also must not lose sight of the constitutional guarantee of equal protection, which prohibits states from selectively facilitating and safeguarding the rights of voters based upon where they live in the state. Florida's election laws, as applied in this case, run afoul of that prohibition.

Finally, the defendants contend that we need not even decide the merits of the constitutional claims in this case because the plaintiffs have not suffered an irreparable injury. They base that assertion on two premises. First, the defendants maintain that it is inappropriate at this juncture to decide whether permanent injunctive relief should be issued. I disagree for the reasons I have already stated. See supra at 1194-95, discussing Thornburgh v. Am. Coll. of Obstetricians Gynecologists, 476 U.S. 747, 755-57, 106 S.Ct. 2169, 2176-77, 90 L.Ed.2d 779 (1986). Second, the defendants maintain that there is no equal protection violation unless and until the outcome of the election is altered by the inclusion of the manually recounted ballots in Florida's certified results. But, as I have already explained, the constitutional harm is inflicted when the ballots of similarly situated voters are counted and weighted differently, and that harm exists regardless of the outcome of the election.

The standard for a permanent injunction is essentially the same as for a preliminary injunction except that the plaintiff must show actual success on the merits instead of a likelihood of success. Amoco Prod. Co. v. Village of Gambell, 480 U.S. 531, 546 n. 12, 107 S.Ct. 1396, 1404 n. 12, 94 L.Ed.2d 542 (1987). In addition to succeeding on the merits, a plaintiff must "demonstrate the presence of two elements: continuing irreparable injury if the injunction does not issue, and the lack of an adequate remedy at law." Newman v. State of Ala., 683 F.2d 1312, 1319 (11th Cir. 1982). Explaining the distinction between "irreparable injury" and "adequate remedy at law," our predecessor circuit said:

[T]he essential prerequisite to a permanent injunction is the unavailability of an adequate remedy at law. Irreparable injury is, however, one basis, and probably the major one, for showing the inadequacy of any legal remedy. . . . Often times the concepts of "irreparable injury" and "no adequate remedy at law" are indistinguishable. . . . "[T]he irreparable injury rubric is intended to describe the quality or severity of the harm necessary to trigger equitable intervention. In contrast, the inadequate remedy test looks to the possibilities of alternative modes of relief, however serious the initial injury."

Lewis v. S.S. Baune, 534 F.2d 1115, 1124 (5th Cir. 1976) (citations omitted).

Here, I believe that the plaintiffs in these two cases have succeeded on the merits by establishing that the disparate treatment of similarly situated voters violates the Equal Protection Clause. That constitutional injury to their right to vote is irreparable, since it "cannot be undone through monetary remedies." Cunningham v. Adams, 808 F.2d 815, 821 (11th Cir. 1987), both because of the unquantifiable nature of the right to vote as well as its fundamental importance in our system of representative democracy. See Reynolds v. Sims, 377 U.S. 533, 562, 84 S.Ct. 1362, 1381, 12 L.Ed.2d 506(1964) (the right to vote is "a fundamental political right, because [it is] preservative of all rights") (citation and quotations omitted). See also Northeastern Fla. Chapter of the Assoc. of Gen. Contractors v. City of Jacksonville, 896 F.2d 1283, 1285 (11th Cir. 1990) (discussing cases in which this Court has recognized that an on-going violation of the First Amendment or privacy rights constitutes irreparable injury, and stating that "[t]he rationale behind these decisions was that chilled free speech and invasions of privacy, because of their intangible nature, could not be compensated by monetary damages; in other words, plaintiffs could not be made whole").

Not surprisingly, there is no suggestion by the defendants that there is an adequate remedy at law to address the voting-rights injury presented in this case. See Dillard v. Crenshaw County, 640 F.Supp. 1347, 1363 (M.D.Ala. 1986) ("Given the fundamental nature of the right to vote, monetary remedies would obviously be inadequate in this case; it is simply not possible to pay someone for having been denied a right of this importance."). There is an irreparable injury to the right to vote for which there is no adequate remedy at law. Accordingly, granting the requested injunctive relief is the only appropriate remedy.

Appendices CHART A All Counties Ranked by Population Total Bush/ Gore County Population Vote System Bush Vote Gore Vote Gore Vote Percentage Miami-Dade 2175634 punch 289533 328802 618335 53.18% Broward 1535468 punch 177323 386561 563884 68.55% Palm Beach 1049420 punch 152951 269732 422683 63.81% Total 15111244 2910492 2910192 5820684 50.00%

CHART B Punch Card Counties Ranked by Population Total Bush/ Gore County Population Bush Vote Gore Vote Gore Vote Percentage Miami-Dade 2175634 289533 328802 618335 53.18% Broward 1535468 177323 386561 563884 68.55% Palm Beach 1049420 152951 269732 422683 63.81% Total 9365502 1640232 1853067 3493299 53.05% CHART C Punch Card Counties Ranked by Percentage of "No Vote" Ballots Presidential No Percent Gore County Cast Votes Counted Votes No Votes Share Palm Beach 462588 433186 29402 6.36% 63.81% Miami-Dade 654044 625443 28601 4.37% 53.18% Broward 588007 573396 14611 2.48% 68.55% Total 3718305 3572705 145600 3.92% 53.05% CHART D Punch Card Counties Ranked by Percentage of Citizens 65+ Population Percent Gore County Population 65+ 65+ Share Palm Beach 1049420 253796 24.18% 63.81% Broward 1535468 258033 16.80% 68.55% Miami-Dade 2175634 310642 14.28% 53.18% Total 9365502 1737707 18.55% 53.05% CHART E Punch Card Counties Ranked by Percentage of Black/Hispanic Total Total Total Black/ Percentage Gore County Population Black Hispanic Hispanic Black/Hispanic Share Miami-Dade 2175634 443024 1249358 1692382 77.79% 53.18% Broward 1535468 285918 196581 482499 31.42% 68.55% Palm Beach 1049420 155763 117114 272877 26.00% 63.81% Total 9365502 1503601 1953026 1204826 12.86% 53.05% CHART F Marksense (Optical Scan) Counties Ranked by Percentage of "No Votes" Ballots Presidential No Percent Gore County Population Cast Votes Counted Votes No Votes Share Total 5614905 2353811 2320099 33712 1.43% 45.47%