BLACK, Circuit Judge:

Defendants-Appellants, who are the current members of the Gainesville City Council, the current clerk of the Gainesville City Council, and the current City Manager for Gainesville (Appellants), appeal the judgment of the district court finding Gainesville's at-large method of electing city council members violates Section 2 of the Voting Rights Act of 1965, as amended, 42 U.S.C. § 1973 (Section 2), and enjoining Gainesville from conducting further city council elections under the at-large system. We hold that the district court's findings of fact and conclusions of law are not sufficiently detailed that we can ascertain the bases for the district court's ultimate conclusion. We therefore vacate the district court's judgment and remand for more specific findings of fact and conclusions of law consistent with this opinion.

I. BACKGROUND

Gainesville is located in Hall County, Georgia. According to the 1990 Census, 23.5% of Gainesville's nearly 20,000 residents and 20.2% of its total voting-age population are black. The Gainesville City Council has five members, all of whom are elected at large. Although one council member must reside in each of the city's five wards, all voters in the city may vote on the representative from each ward. While a majority of the voters in Ward 3 are black, a majority of the voters in the other four wards are white. The council members' terms are staggered. There is a majority-vote requirement, so a runoff election is required if no candidate receives more than 50% of the vote in the general election.

Three contested Gainesville City Council elections have featured a black candidate. In 1978, the two candidates for the Ward 3 seat were black: John W. Morrow, Jr., defeated Reverend C.T. Hester. Morrow was reelected without opposition until 1990. In 1990, Morrow ran for the Ward 3 seat against another black candidate, Plaintiff Rose Johnson. Morrow won reelection. As Johnson received nearly two-thirds of the black vote in the 1990 election, the parties do not dispute she was the preferred candidate of black voters. In the last contested Ward 3 election, which took place in 1995, Morrow again ran against Johnson. Once again, although Johnson received well over two-thirds of the black vote, Morrow defeated her.

After losing her first contest against Morrow, Johnson along with four other black registered voters in Gainesville (Appellees) filed a complaint on January 11, 1991, in the United States District Court for the Northern District of Georgia, alleging that Gainesville's at-large method of electing city council members dilutes the voting strength of black citizens, in violation of Section 2, and that Appellants maintained the at-large system for racially discriminatory purposes, in violation of the Fourteenth and Fifteenth Amendments to the United States Constitution. The district court denied cross-motions for summary judgment in January 1993. Following a three-day bench trial in 1994, the district court entered an order finding the at-large system did not violate Section 2, but specifically declining to rule on Appellees' constitutional claim. Appellees appealed that decision. This Court dismissed the appeal for lack of appellate jurisdiction, holding the district court had not issued a final judgment because it had not ruled on Appellees' constitutional claims, which, if successful, would have altered the judgment. Johnson v. Hamrick, 11th Cir., 1996, (No. 94-9203, Jan. 25, 1996).

In March 1997, the district court granted Appellees' motion to reopen the case to take additional evidence of post-trial elections. In so ruling, the court also stated it would consider the constitutional claim. On the morning of July 11, 1997, the district court held an evidentiary hearing to take additional evidence of elections occurring since the fall of 1994. Almost one year later, on June 10, 1998, the district court issued its new opinion on the merits of the case. This time, the court ruled in favor of Appellees, finding the at-large system violated Section 2 and rejecting Appellants' constitutional challenge to Section 2. The district court again declined to rule upon Appellees' constitutional claims. Appellants filed this appeal of the district court's 1997 judgment, asserting the district court erred in its ultimate determination of vote dilution.

II. STANDARD OF REVIEW

We review a district court's factual findings regarding Section 2 violations and its determination of whether vote dilution has occurred for clear error. Fed.R.Civ.P. 52(a); Thornburg v. Gingles, 478 U.S. 30, 79, 106 S.Ct. 2752, 2781 (1986). Rule 52(a) does not, however, hinder our "power to correct errors of law, including those that may infect a so-called mixed finding of law and fact, or a finding of fact that is predicated on a misunderstanding of the governing rule of law." Gingles, 478 U.S. at 79, 106 S.Ct. at 2781 (quotation and citations omitted). In addition, Rule 52(a) requires that a district court's findings of fact and conclusions of law "be sufficiently detailed that [we] can ascertain the factual and legal basis for the district court's ultimate conclusion." Cross v. Baxter, 604 F.2d 875, 879 (5th Cir. 1979) (citation omitted).

III. DISCUSSION

A Section 2 violation occurs when, "based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivisions are not equally open to participation by members of a [protected] class of citizens . . . in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 42 U.S.C. § 1973(b).

At a minimum, Section 2 plaintiffs must prove "the three now-familiar Gingles factors (compactness/numerousness, minority cohesion or bloc voting, and majority bloc voting)." Johnson v. De Grandy, 512 U.S. 997, 1011, 114 S.Ct. 2647, 2657 (1994). Other factors may, however, in the totality of circumstances, be relevant to a claim of racial vote dilution. Derived from the Senate Report accompanying the 1982 amendment to Section 2, those factors include: (1) "the extent of any history of official discrimination in the state or political subdivision that touched the right of the members of the minority group to register, to vote, or otherwise to participate in the democratic process;" (2) "the extent to which voting in the elections of the state or political subdivision is racially polarized;" (3) "the extent to which the state or political subdivision has used unusually large election districts, majority vote requirements, anti-single shot provisions, or other voting practices or procedures that may enhance the opportunity for discrimination against the minority group;" (4) "if there is a candidate slating process, whether the members of the minority group have been denied access to that process;" (5) "the extent to which members of the minority group in the state or political subdivision bear the effects of discrimination in such areas as education, employment and health, which hinder their ability to participate effectively in the political process;" (6) "whether political campaigns have been characterized by overt or subtle racial appeals;" and (7) "the extent to which members of the minority group have been elected to public office in the jurisdiction." Gingles, 478 U.S. at 36-37, 106 S.Ct. at 2759 (quoting S. Rep. No. 97-417, at 28-29, reprinted in 1982 U.S.C.C.A.N. 177, 206-07). Additional factors that may be probative of vote dilution in some cases are: (1) "whether there is a significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority group;" and (2) "whether the policy underlying the state or political subdivision's use of such voting qualification, prerequisite to voting, or standard, practice or procedure is tenuous." Gingles, 478 U.S. at 37, 106 S.Ct. at 2759 (quoting S. Rep. No. 97-417, at 28-29, reprinted in 1982 U.S.C.C.A.N. 177, 206-07). Also, in a Section 2 case involving state legislative single-member districts, the Supreme Court announced another factor, proportionality, that it defined as the relationship between "the number of majority-minority voting districts [and] minority members' share of the relevant population." De Grandy, 512 U.S. at 1013-14 n. 11, 114 S.Ct. at 2658 n. 11. Finally, two judges of this Court consider racial bias in the voting community to be a relevant factor. Nipper v. Smith, 39 F.3d 1494, 1524 (11th Cir. 1994) (en banc) (plurality opinion of Tjoflat, C.J., joined by Anderson, J.) ("The defendant may rebut the plaintiff's evidence by demonstrating the absence of racial bias in the voting community.").

The parties agree that the analytical framework to be applied to Appellees' claim is as follows. First, a plaintiff must demonstrate the three Gingles factors of compactness/numerousness, minority cohesion or bloc voting, and majority bloc voting. If one or more of the Gingles factors is not shown, then the defendants prevail. If all three factors are shown, then the district court must review all relevant evidence under the totality of the circumstances. The defendants may present evidence of the lack of racial bias in the community, proportionate representation, past and present electoral success, as well as proof of the other factors which are indicative of the existence or non-existence of vote dilution. The plaintiffs may respond in kind. Without treating any single factor as dispositive, the district court then determines whether the members of the minority group are denied equal political opportunity with respect to their race or color. If they are, a claim under Section 2 has been established.

The parties do not dispute the legal accuracy of this framework. Rather, they dispute the district court's application of it, as well as its view of the dilution factors. Appellants argue the district court clearly erred in rendering findings adverse to them, including its ultimate determination of vote dilution. Appellees defend the district court's judgment, contending it properly resolved conflicts in the evidence and considered the totality of the circumstances.

In the judgment from which Appellants appeal, the district court found Appellees had established the three Gingles factors. In particular, the district court reconsidered its previous ruling and found Appellees had demonstrated the existence of the third Gingles factor. Next, the district court considered the totality of the circumstances and found black voters enjoyed less political opportunity. Accordingly, the district court found that Gainesville's at-large electoral system violated Section 2.

Appellants do not dispute Appellees proved the first two prongs of Gingles. Rather, Appellants assert the district court erred in finding Appellees had proved the third Gingles factor: that "the white majority votes sufficiently as a bloc to enable it in the absence of special circumstances, such as the minority candidate running unopposed — usually to defeat the minority's preferred candidate." Gingles, 478 U.S. at 51, 106 S.Ct. at 2766-67 (citations omitted).

Appellants assert the district court made three errors in its analysis of the third Gingles factor. First, Appellants argue the district court erred by giving more weight to the three elections involving black candidates than to any other elections. Second, Appellants argue the district court erred by viewing endogenous elections (those involving only city voters; for instance, Gainesville City Council elections) as more probative than exogenous elections (those involving non-city voters; for instance, county-wide elections). Finally, Appellants argue the district court erred by using so-called "split-preference" analysis to examine the elections in which only white candidates ran.

To establish the third Gingles factor, a plaintiff must show not only that whites vote as a bloc, but also that white bloc voting regularly causes the candidate preferred by black voters to lose; in addition, plaintiffs must show not only that blacks and whites sometimes prefer different candidates, but that blacks and whites consistently prefer different candidates. See Gingles, 478 U.S. at 48, 106 S.Ct. at 2765 ("The theoretical basis for this type of impairment is that where minority and majority voters consistently prefer different candidates, the majority, by virtue of its numerical superiority, will regularly defeat the choices of minority voters." (citation and footnote omitted)).

The district court did not clearly err by giving more weight to elections involving black candidates. There is case law from this Circuit indicating elections involving black candidates may be given more weight in determining vote dilution. See, e.g., Davis v. Chiles, 139 F.3d 1414, 1417 n. 5 (11th Cir. 1998), cert. denied, ___ U.S. ___, 119 S.Ct. 1139 (1999) (noting evidence drawn from elections involving black candidates is more probative); Southern Christian Leadership Conference v. Sessions, 56 F.3d 1281, 1293 (11th Cir. 1995) (stating there was no cause to set aside the district court's finding on the ultimate issue of fact (vote dilution) where the district court "concluded that greater weight should be given to judicial elections and white on black judicial elections, but that it would also consider political races generally — white on black and white on white — in seeking to determine the question of whether the political process is open to minority voters" (internal punctuation omitted)). We do not mean to imply district courts should give elections involving black candidates more weight; rather, we merely note that in light of existing case law district courts may do so without committing clear error. Besides, it is not clear in this case that the district court in fact gave more weight to the three elections involving black candidates than to those involving only white candidates. The district court did pay particular attention to the two city council elections pitting Morrow against Johnson, in which Johnson, the black candidate preferred by black voters, lost. See Dist. Ct. Order at 12-13. However, the district court examined all endogenous (involving only city voters) elections, including those that did not involve a black candidate, and found:

It is also important to recognize that white voters have been able to elect the candidate of their choice in all eight (8) of the contested city council elections in which they expressed a preference. Conversely, black voters have been able to elect the black preferred candidate in only three (3) elections which featured only white candidates and the successful candidate was also the white preferred candidate.

District Ct. Order at 11 (internal footnote omitted).

The district court also did not clearly err by viewing endogenous elections as more probative than exogenous elections. Although the district court examined some exogenous elections, it did seem to focus on endogenous elections and found "the endogenous elections to be more probative." Dist. Ct. Order at 11. Not only does that make sense, but it also is the law of this Circuit. In City of Carrollton Branch of the NAACP v. Stallings, 829 F.2d 1547 (11th Cir. 1987), this Court disagreed with the district court's reliance on municipal elections to indicate voting behavior in a Georgia county, the jurisdiction at issue in the case:

The district court's reliance on municipal elections in Carroll County as proof of minority electoral success of a county electoral scheme is obviously misplaced. The political jurisdiction in question here is the county, not the cities of Villa Rica, Whitesburg, or Carrollton.

Id. at 1560. Gainesville is the jurisdiction at issue here, and thus the district court properly assigned more probative value to its elections.

With respect to Appellants' argument that the district court erred in its application of "split-preference" analysis to elections in which only white candidates ran, however, we conclude the district court made insufficient findings for us to review. In conducting "split-preference" analysis, a court gives more weight to split-preference elections: those in which black and white voters preferred different candidates. Davis, 139 F.3d at 1417. In this case, five of the nine endogenous elections, including the two involving black candidates, were split-preference elections. In all five, the black preferred candidate was unsuccessful. The district court concluded that the five split-preference endogenous elections showed legally significant white bloc voting.

Although the law as to split-preference analysis is hardly well-settled, we doubt it would be clear error to pay particular attention to split-preference elections. We do, however, need a sufficiently detailed explanation of the district court's basis for weighing the elections. We have strictly adhered to the Fed.R.Civ.P. 52(a) requirement of a definite record in voting rights cases. As we have explained:

Because the resolution of a voting dilution claim requires close analysis of unusually complex factual patterns, and because the decision of such a case has the potential for serious interference with state functions, we have strictly adhered to the rule 52(a) requirements in voting dilution cases and have required district courts to explain with particularity their reasoning and the subsidiary factual conclusions underlying their reasoning.

Cross, 604 F.2d at 879 (citations omitted).

In conducting its version of split-preference analysis, the district court completely disregarded the four elections in which the preferences of white and black voters were the same, stating that one "add[ed] little to the court's analysis" and the other three "lack[ed] probative value." Dist. Ct. Order at 13-14. In this case, black and white voters preferred different candidates in five of nine elections, or 56% of the time. The district court concluded without any explanation that this ratio suggested black and white voters consistently preferred different candidates. This is not correct split-preference analysis. Indeed, at least as an initial inference, elections in which black and white voters prefer the same winning candidate weigh just as heavily against Appellees' establishing the third factor of Gingles as any evidence on the other side of the scale. Such elections provide evidence that the interests of the black and white communities may be merging. The district court did not tell us what evidence it found to controvert that inference.

Whether or not a consistent pattern of voting emerges from these data is for the district court to answer, but we believe the question is at least close enough for us to require some explanation for the district court's conclusion. Moreover, at least arguably, in two of those five split-preference elections, white voting did not constitute a bloc and black voting did not demonstrate a preference. In the 1986 Ward 1 runoff election, only 50.4% of white voters voted for the winner, who received 45.6% of the black vote. In the 1986 Ward 4 election, only 52.8% of white voters voted for the winner, who received 47.7% of the black vote. Indeed, these numbers were so close the district court in its first opinion did not regard these two elections as evidence of white bloc voting or black-preferred candidates losing:

There is no evidence in the record to suggest that 50.4% constitutes a "bloc," and this court has reservations about deeming 52.8% a "bloc" as that term is used in voting rights cases. This is particularly true in light of the fact that these percentages are estimates, and thus necessarily contain a statistical margin of error. It therefore cannot be said, in regard to endogenous elections, that white voters usually vote as a bloc to defeat the black preferred candidate.

Dist. Ct. (Sept. 27, 1994) Order at 25-26 (emphasis in original). In its second opinion, the district court changed its mind without any explanation, finding not only that the two 1986 elections were evidence of white bloc voting defeating black-preferred candidates, but also that white bloc voting usually defeated black-preferred candidates.

We emphasize, of course, courts are free to change their minds. Irrespective of whether the district court changed its mind, however, to conduct our review we need more specific factual findings and legal conclusions as to why it decided this issue and this case the way it did. The district court's opinion does not elucidate what evidence it had that the two 1986 elections were examples of white bloc voting defeating black preferences. Even more troublesome, the district court does not explain what evidence it had that white bloc voting usually defeated black-preferred candidates. We do not hold the district court clearly erred by finding Appellees established the third Gingles factor, but we do believe the issue is extremely close and warrants additional and more specific findings of fact and conclusions of law, especially considering the district court changed its mind.

The district court need not necessarily include the required findings of fact or conclusions of law in its analysis of the Gingles factors; the findings and conclusions could just as well fall under the rubric of the totality of circumstances analysis. We do not favor form over substance and will defer to findings and conclusions no matter where they are. In this case, however, not only was the district court's analysis of the third Gingles factor lacking, but its analysis of the totality of the circumstances was quite sparse. The district court addressed only three of the eleven possible factors. First, as to the first Senate factor, the district court averred that "[t]he State of Georgia and Gainesville have a history of official discrimination against blacks." Of course, that does little to distinguish Gainesville or Georgia from any other Southern state or city. Second, as to the third Senate factor, the district court found that Gainesville's electoral system contained two features that have enhanced the discriminatory impact of the at-large electoral system: a majority-vote requirement with a run-off and the use of a subdistrict residency requirement with staggered terms. Finally, as to the fifth Senate factor, the district court considered testimony regarding discrimination against blacks in housing, employment, and education. The district court neglected to mention, however, how this discrimination "hinder[ed] their ability to participate effectively in the political process." Gingles, 478 U.S. at 37, 106 S.Ct. at 2759. In sum, the district court did not adequately state what evidence it relied upon to conclude that Gainesville's black voters were denied equal political opportunity.

We must emphasize we do not express any opinion about the merits of Appellees' claims; we do not suggest the record in this case cannot support a finding of vote dilution. Moreover, whether Appellees establish the third Gingles factor and whether the totality of circumstances demonstrates black voters are denied equal political opportunity are unusually complex, fact-intensive questions for the district court to answer. Our task is merely to ensure the district court made its findings as to those questions without committing clear error. Regardless of how we would have weighed the evidence, we must defer to the district court's findings. Precisely because the issues are so complex and fact-intensive, however, we need something to which to defer.

Here, the district court's findings were inadequate to permit any meaningful appellate review. We reach this conclusion reluctantly, for we are loathe to further ritualize the district court's fact-specific inquiry by requiring findings of fact that may appear obvious to a judge so familiar with Gainesville and its electorate. Nevertheless, our review of the district court's judgment is itself a fact-specific inquiry.

IV. CONCLUSION

We conclude the district court's findings were insufficient to permit review. Accordingly, we vacate the judgment of the district court and remand the case for specific findings of fact and conclusions of law consistent with this opinion.

VACATED AND REMANDED.