Ernestine LEWIS; Sylvester J. Lewis, Plaintiffs-Appellants, and Minnetta E. Hill, Plaintiff, v. ALAMANCE COUNTY, NORTH CAROLINA, including its Board of Commissioners and Board of Elections, Defendant-Appellee, 99 F.3d 600

Affirmed by published opinion. Judge LUTTIG wrote the majority opinion, in which Chief Judge WILKINSON joined. Chief Judge WILKINSON wrote a concurring opinion. Judge MICHAEL wrote a dissenting opinion. OPINION LUTTIG, Circuit Judge: Appellants Ernestine and Sylvester Lewis, black voters of appellee Alamance County, North Carolina, challenged the County’s at-large method of electing county commissioners, arguing that black citizens have been denied an equal opportunity to elect representatives of their choice through vote dilution, in violation of

Section 2 of the Voting Rights Act, 42 U.S.C. § 1973. The district court granted

summary judgment for the County, holding that plaintiffs failed to demonstrate that minority-preferred candidates are usually defeated by white bloc voting, as required by Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). For the reasons that follow, we affirm. I. Alamance County is governed by a Board of Commissioners, the five members of which are elected, in at-large partisan.elections, to four-year staggered terms. Voters are allowed to cast votes for as many candidates as there are vacant seats, but they cannot cast more than one vote for any one candidate.

Since the 1965 passage of the Voting Rights Act, black candidates have run for seats on the Board in eight of fourteen election cycles. Only one black candidate, Jack O’Kelley, has been elected, although he was elected three times, in 1974 (after first being appointed to fill a vacancy), 1976, and 1980. Moreover, white candidates supported by a majority (often substantial) of black voters, either in the primary election, the general election, or both, have repeatedly won election.

Section 2(

a) of the Voting Rights Act of 1965 prohibits a State or its political subdivisions from imposing any voting practice “in a manner which results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color.” 42 U.S.C. § 1973(a).

Section 2(

b) of the Act, as amended in 1982, further provides that a violation of § 2(

a) occurs, if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens protected by subsection (

a) of this

section 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) (emphasis added). In Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986), the Supreme Court interpreted the amended Voting Rights Act as it applied to a challenge to multi-member districts in which candidates were elected at large.

Rejecting the claim that an at-large election scheme is per se violative of the Voting Rights Act, id. at 48, 106 S.Ct. at 2765, the Court established three preconditions to proving that such a voting system dilutes minority group voting strength sufficiently to violate the Act: First, the minority group must be able to demonstrate that it is sufficiently large and geographically compact to constitute a majority in a single-member district_ Second, the minority group must be able to show that it is politically cohesive....

Third, the minority must be able to demonstrate that the white majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate. 478 U.S. at 50-51, 106 S.Ct. at 2766-67. If these preconditions are met, the court must then determine under the “totality of circumstances” whether there has been a violation of

Section 2. See Johnson v. De Grandy, 512 U.S. 997,-, 114 S.Ct. 2647, 2657, 129 L.Ed.2d 775 (1994); Collins v. City of Norfolk, Va., 816 F.2d 932, 938 (4th Cir.1987) (“Collins I”) (“[The] ultimate determination [of vote dilution under the Voting Rights Act] still must be made on the basis of the ‘totality of the circumstances.’ ”). In an effort to meet Gingles ’ second and third preconditions, plaintiffs’ expert in this case conducted bivariate ecological regression analyses on the eleven primary and general elections in which a black candidate was on the ballot.

Based on those regression analyses, the expert estimated the level of support among black voters for each candidate. Plaintiffs then proffered the voter preference estimates as proof of both black voter cohesion (the second Gingles element) and white bloc voting sufficient to usually defeat black-preferred candidates (the third Gingles element).

After reviewing the voter preference estimates from plaintiffs’ expert on these selected elections, the district court held that plaintiffs had failed to provide evidence sufficient to satisfy the third Gingles element, because they had not shown that black-preferred candidates were usually defeated.

Based upon the limited data before the court, this conclusion was fully supported even by plaintiffs’ own expert witness, who admitted during his deposition that twenty of the thirty-one candidates “generally preferred” by black voters in the select number of elections he analyzed won election or nomination, thirteen of twenty-two “strongly preferred” candidates won election or nomination, and eleven of the seventeen “strongest preferred” candidates won election or nomination.

If only the five general elections analyzed by plaintiffs’ expert are considered, eight of the eleven candidates most strongly preferred by black voters were elected to seats on the Board. Plaintiffs advance on appeal four arguments as to why the district court’s conclusion was in error. We reject plaintiffs’ claim that those white candidates who received overwhelming support from black voters in general elections, assertedly only because they were Democrats, should not have been considered as black-preferred candidates by the district court.

And we also reject plaintiffs’ argument that the district court erred in not discounting the repeated success of one of the minority-preferred candidates because of the alleged effects of incumbency. We agree with plaintiffs, however, that the district court improperly aggregated primary and general election results, and also that it faded to conduct an individualized determination into whether some candidates should be treated as black-preferred candidates.

Additionally, we conclude that the district court erred in a third respect, by basing its decision exclusively on data from elections in which a black candidate was on the ballot, rather than on a more representative sample of elections. By limiting its consideration to elections in which a candidate of the minority’s race was on the ballot, the district court may have failed to include as black-preferred candidates some white candidates who may very well have been the representatives of choice of the black community. II.

Turning first to what we perceive to be the overarching error, we believe that, by considering only elections in which a black candidate was on the ballot, the district court failed to analyze a sufficient number of elections to enable it to determine whether white bloc voting usually operates to defeat minority-preferred candidates. As noted, the district court considered only election data from eleven of the twenty-eight primary and general elections held since passage of the Voting Bights Act — six of fourteen primary elections and only five of fourteen general elections.

Although we recognize that this election data was the only data proffered by plaintiffs, we believe that, without a larger, more representative sample of elections, the district court simply did not have before it sufficient.evidence to determine whether the black-preferred candidates usually were defeated, unless one is willing to assume (which we are not) that the black community will never have a preferred candidate in an election in which no black candidate is on the ballot, or, more fundamentally, that a black voter can never prefer a candidate who is white.

Although the district court erred in this regard, we do not reverse its judgment because of this error, for it is the plaintiffs’ burden to establish a violation of

Section 2, and therefore their burden to proffer data from a sufficient number of elections to enable the district court to determine whether white bloc voting usually defeats minority-preferred candidates. Where, as here, plaintiffs fail to carry their burden to proffer sufficient evidence, and the district court correctly concludes on the basis of the proffered evidence that no

Section 2 violation has been established, then the plaintiffs cannot be heard to complain. A

Section 2 of the Voting Rights Act prohibits the use of voting procedures, such as at-large elections, that afford minority voters less opportunity than other members of the electorate “to elect representatives of their choice,” 42 U.S.C. § 1973(b) (emphasis added), or, in the language of Gingles, that afford minority voters less opportunity to elect minority-preferred, candidates. Gingles, 478 U.S. at 51, 106 S.Ct. at 2766-67. The only possible

interpretation of this unambiguous language is that

Section 2 prohibits any election procedure which operates to deny to minorities an equal opportunity to elect those candidates whom they prefer, whether or not those candidates are themselves of the minority race. As Justice Brennan observed in Gingles, “it is the status of the candidate as the chosen representative of a particular racial group, not the race of the candidate,” which is relevant in determining whether Gingles ’ third precondition is satisfied.

Gingles, 478 U.S. at 68, 106 S.Ct. at 2775 (plurality opinion of Brennan, J., joined by Blackmun, Marshall, and Stevens, JJ.); see Collins I, 816 F.2d at 937 n. 5 (“The question is not ... so simple as how many blacks versus whites were elected. The court must examine the parties’ studies of voting preferences to determine which were the preferred candidates of the majority and minority communities.”); see also Lani Guinier, The Triumph of Tokenism: The Voting Rights Act and the Theory of Black Electoral Success, 89 Mich. L.

Rev. 1077, 1103 n. 115 (1991) (“Authentic representatives [of the black community] need not be black as long as the source of their authority, legitimacy, and power base is the black community.”)- That is, a minority-preferred candidate may be a non-minority, just as a minority candidate may be the preferred candidate of the voters of the majority’s race. The black community may prefer a white candidate, as the white community may prefer a black candidate.

A Martin Luther King, Jr. or a Colin Powell can represent white Americans, no less than a John Fitzgerald Kennedy or a Hubert Humphrey can represent black Americans. To indulge the contrary presumption, that every black person necessarily prefers a black candidate over a white candidate, or that every white person necessarily prefers a white candidate over a black candidate, would itself constitute invidious discrimination of the kind that the Voting Rights Act was enacted to eradicate, effectively disenfranchising every minority citizen who easts his or her vote for a non-minority candidate.

To acquiesce in such a presumption would be not merely to resign ourselves to, but to place the imprimatur of law behind, a segregated political system — in Judge Cabranes’ words, an “electoral apartheid,” NAACP v. City of Niagara Falls, N.Y., 65 F.3d 1002, 1016 (2d Cir.1995) — in derogation of the most fundamental principles of our representative democracy and in betrayal of our most cherished beliefs about individual autonomy and political self-determination. Our understanding of

Section 2 that the minority-preferred candidate may be either a minority or a non-minority, and therefore that both elections in which the candidates are of the same race and elections in which the candidates are of different races must be considered in order to determine whether white bloc voting usually defeats the minority-preferred candidate, is confirmed within

Section 2 itself, by the express proviso that “[t]he extent to which members of a protected class have been elected to office” is but “one circumstance which may be considered” in assessing whether minority voters have been denied an equal opportunity “to participate in the political process and to elect representatives of their choice.” See 42 U.S.C. § 1973(b); see also id. (“[Njothing in this

section establishes a right to have members of a protected class elected in numbers equal to their proportion in the population”).

Were this proviso, coupled with the statute’s plain reference to the minority’s representatives “of choice,” not enough, the plurality in Gingles itself, as noted, expressly stated that “only the race of the voter, not the race of the candidate, is relevant” in determining whether a plaintiff has met Gingles ’ preconditions, Gingles, 478 U.S. at 68, 106 S.Ct. at 2775 (emphasis added), and not a single Justice suggested that only candidates of the minority race could be considered minority-preferred candidates, see infra note 5.

Indeed, as the Third Circuit has held, a candidate of the minority’s race need not even be the minority-preferred candidate. Jenkins v. Red Clay Consol. School Dist. Bd. of Educ., 4 F.3d 1103, 1126 (3d Cir.1993), cert. denied, — U.S. -, 114 S.Ct. 2779, 129 L.Ed.2d 891 (1994). Although at times they suggest otherwise, see Appellant’s Br. at 30 (“Under the law, black voters must have an opportunity not just to elect candidates of their choice, but to elect black candidates of their choice”), plaintiffs no doubt agree with our

interpretation of

Section 2. They themselves conducted a bivariate regression analysis on the eleven elections in which a black candidate was on the ballot precisely in order to determine whether the black candidate or the white candidate was the minority-preferred candidate in those elections. Had they not believed, as we do, that a white candidate may be the minority-preferred candidate, then they would have advanced their claim solely on the basis that only one black candidate had actually won in these elections.

To conduct a regression analysis with respect to a black-white election, ostensibly in part to identify the black-preferred candidate, makes sense only if it is possible for the preferred candidate not to be the black candidate. We recognize that the plaintiffs’ expert in Gingles had only analyzed elections in which a black candidate had appeared on the ballot, see Gingles, 478 U.S. at 52, 106 S.Ct. at 2767, and that the Court there relied upon that data in invalidating North Carolina’s at-large election scheme.

The Court in Gingles, however, never addressed the question of whether plaintiffs’ exclusion of evidence from white-white races rendered their proffer insufficient; it merely assumed for purposes of that case that the black candidates there were the representatives of choice of black voters. Justice Brennan, in fact, could not have been clearer in this regard: Because both minority and majority voters often select members of their own race as their preferred representatives, it will frequently be the case that a black candidate is the choice of blacks, while a white candidate is the choice of whites....

Indeed, the facts of this case illustrate that tendency — blacks preferred black candidates, whites preferred white candidates. Thus, as a matter of convenience, we and the District Court may refer to the preferred representative of black voters as the “black candidate” and to the preferred representative of white voters as the “white candidate." Nonetheless, the fact that race of voter and race of candidate is often correlated is not directly pertinent to a § 2 inquiry.

Under § 2, it is the status of the candidate as the chosen representative of a particular racial group, not the race of the candidate, that is important. Id. at 68, 106 S.Ct. at 2775 (emphasis added, in part) (plurality opinion); cf. Harvell v. Blytheville School District No. 5, 71 F.3d 1382, 1386 (8th Cir.1995) (en banc) (“We do not categorically state that a candidate is the minority-preferred candidate simply because that candidate is a member of the minority. Such stereotyping runs afoul of the principles embodied in the Equal Protection Clause.” (citing Miller v.

Johnson, — U.S. -, -, 115 S.Ct. 2475, 2486, 132 L.Ed.2d 762 (1995))), cert. denied, — U.S.-, 116 S.Ct. 1876, 135 L.Ed.2d 172 (1996). Where the results of not even a majority, much less a substantial majority, of elections are considered, it is simply not possible for the district court to determine whether minority-preferred candidates are “usually” defeated.

We therefore hold that, in assessing whether “the white majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate,” a district court must consider, at a minimum, a representative cross-section of elections, and not merely those in which a minority candidate appeared on the ballot, at least where elections in which minorities were on the ballot do not constitute a substantial majority of the total number of elections. Cf.

Niagara Falls, 65 F.3d at 1016-17 (“The trial court ... properly attempted to analyze white-white elections, as well as to evaluate the success of white candidates in black-white elections, to determine whether whites voted as a bloc to defeat blacks’ preferred candidates.”). Only by considering such a representative sample can the district court meaningfully determine whether a white voting bloc usually defeats the minority-preferred candidate. See, e.g., Uno v.

City of Holyoke, 72 F.3d 973, 985 (1st Cir.1995) (“[T]o be legally significant, racially polarized voting in a specific community must be such that, over a period of years, whites vote sufficiently as a bloc to defeat minority[preferred] candidates most of the time....

In order reliably to tell whether racial groups do (or do not) band together behind particular candidates with regularity, all elections in the relevant timeframe (or, at least, a representative sampling of them) must be studied — not just those elections that, taken in isolation, reveal the cicatrices of racially polarized voting.” (emphasis added)). By focusing exclusively on elections in which a minority candidate appeared on the ballot, it can only be determined whether the “minority candidate” is usually defeated, not, as required by

Section 2, whether the “minority-preferred candidate” is usually defeated. See, e.g., Harvell, 71 F.3d at 1393 n. 2. Our sister circuits interpret

section 2 and the third Gingles element as do we.

As the Tenth Circuit recognized in affirming a district court’s finding that plaintiffs’ evidence was “inadequate” to measure the success of minority-preferred candidates because it did not consider elections in which only whites were on the ballot: [A] per se rule against examining races that have only white candidates ... would be clearly contrary to the [Gingles] plurality opinion, which views the race of the candidates as irrelevant in voting analysis [and] ... is questionable in light of the language of § 2, which seeks to give minorities equal opportunity to “elect representatives of their choice.” Sanchez v.

Bond, 875 F.2d 1488, 1495 (10th Cir.1989), cert. denied, 498 U.S. 937, 111 S.Ct. 340, 112 L.Ed.2d 305 (1990); SCLC v. Sessions, 56 F.3d 1281, 1293-94 (11th Cir.1995) (en banc) (approving district court’s conclusion that the analysis by plaintiffs expert “was flawed because he only analyzed elections ... involving a black candidate.”), cert. denied, — U.S.-, 116 S.Ct. 704, 133 L.Ed.2d 660 (1996); see also Uno, 72 F.3d at 988 n. 8 (“[T]he VRA does not require for a successful

section 2 showing that minority-preferred candidates be members of the minority group_”); Niagara Falls, 65 F.3d at 1016 (district court properly analyzed white-white elections to determine which candidates were black-preferred); Clarke v. City of Cincinnati, 40 F.3d 807, 810 n. 1 (6th Cir.1994) (“Plaintiffs suggest that white candidates cannot be included among blacks’ preferred candidates, but we disagree.” (quotation marks omitted)), cert. denied, — U.S. -, 115 S.Ct. 1960, 131 L.Ed.2d 851 (1995); Nipper v.

Smith, 39 F.3d 1494, 1540 (11th Cir.1994) (en banc) (“[W]e do not foreclose the consideration of electoral races involving only white candidates where the record indicates that one of the candidates was strongly preferred by black voters.”), cert. denied, — U.S. -, 115 S.Ct. 1795, 131 L.Ed.2d 723 (1995); Jenkins, 4 F.3d at 1125 (“[TJhere may be majority candidates who truly may be the minority community’s representative of choice.”); Valladolid v.

City of National City, 976 F.2d 1293, 1297 (9th Cir.1992) (“[Appellants’ argument] that minorities have long been underrepresented on the city council, and that a significant number of minority candidates have failed in their efforts to win election to that body ... too easily equate[s] the presence or absence of minority individuals on the council with the victory or defeat of minority-preferred candidates.”). The Fifth Circuit in Citizens for a Better Gretna v.

City of Gretna, La., 834 F.2d 496 (5th Cir.1987), cert. denied, 492 U.S. 905, 109 S.Ct. 3213, 106 L.Ed.2d 564 (1989), agreed that “Gingles is properly interpreted to hold that the race of the candidate is in general of less significance than the race of the voter,” id. at 503, and it even noted that “Justice Brennan’s plurality opinion is careful not to state that a black candidate is tantamount to the black preference.” Id.

But, it found “implicit” in Gingles ’ undiscussed reliance on data only from black-white elections “the notion that black preference is determined from elections which offer the choice of a black candidate,” and thus concluded that the race of the candidate was less significant than that of the voters “only within the context of an election that -offers voters the choice of supporting a viable minority candidate.” Id. at 503-04; see also id. at 504 (“The various Gingles concurring and dissenting opinions do not consider evidence of elections in which only whites were candidates. Hence, neither do we.”); Campos v.

City of Baytown, Tex., 840 F.2d 1240, 1244 (5th Cir.1988) (“The district court was warranted in its focus on those races that had a minority member as a candidate.” (citing Gretna, 834 F.2d at 503)), cert. denied, 492 U.S. 905, 109 S.Ct. 3213, 106 L.Ed.2d 564 (1989). However, the Fifth Circuit has, since, substantially retreated from its dictum in Gretna in Overton v. City of Austin, 871 F.2d 529 (5th Cir.1989), holding that it is, indeed, permissible to consider elections in which a minority is not on the ballot.

Id. at 538 (relying only on elections in which a minority was on the ballot and on election results rather than extrinsic factors “is not the only permissible way to approach § 2 claims.”); cf. LULAC v.

Clements, 986 F.2d 728, 748 (5th Cir.) (“[A] court may properly give more weight to elections in which the minority-preferred candidate is a member of the minority group.”), on reh’g en banc, 999 F.2d 831, 864 (1993) (“This court has consistently held that elections between white candidates are generally less probative in examining the success of minority-preferred candidates .... ” (emphasis added) (citing Campos, 840 F.2d at 1245; Gretna, 834 F.2d at 503)), cert. denied, 510 U.S. 1071, 114 S.Ct. 878, 127 L.Ed.2d 74 (1994).

In short, though some courts have held that black-white elections are more probative than white-white elections, see, e.g., Uno, 72 F.3d at 988 n. 8; Jenkins, 4 F.3d at 1128; LULAC, 999 F.2d at 864; Smith, 687 F.Supp. at 1316, to our knowledge, no court has held that a white candidate cannot, as a matter of law, be a minority-preferred candidate, and therefore that white-white elections are irrelevant to the Gingles third element inquiry. B.

By failing to consider evidence of elections in which no minority candidate appeared on the ballot, the district court, insofar as can be discerned, could have understated (or overstated) the extent to which minority-preferred candidates were usually defeated in Alamance County. For example, the record includes election data from the 1974 general election because O’Kelley, a black candidate, was on the ballot seeking election to the unexpired term of the seat to which he had been appointed following a vacancy.

But it does not include election data from the 1974 primary election, notwithstanding that the two Democrats elected in the primary for the two full-term seats, both of whom were white, each received 99 + % of the black vote in the general election.

Although it may well be that either or both of these candidates were the minority-preferred candidates in the 1974 primary, and therefore would have counted as minority-preferred electoral successes (both did receive a majority of black support in other primary elections — Long received 84% in 1972, and Newlin 60%, in 1978), it could be that neither was a minority-preferred candidate in the 1974 primary, in which case there were uncounted minority-preferred defeats.

Similarly, the record does not include election results from the 1972 or 1978 general elections, although Long was successful in the 1972 primary with 84% of the black vote (second only to the 98% received by Morris, a black candidate), and Newlin was successful in the 1978 primary with 60% of the black vote (second only to the 78% received by black candidate Harris).

Based upon the record data from other general elections, it may be that both of these candidates received near unanimous support of the black community in their general elections, and thus should have been considered the black-preferred candidates in those elections. But, without the data from these elections, it simply cannot be known whether they, or others, were the black-preferred candidates, and therefore whether the black-preferred candidates were successful or defeated. And there are no election returns, general or primary, for the 1966, 1968, 1970, 1982, 1988, or 1990 election cycles.

The district court therefore likewise could not have known who were the representatives of choice of black voters in those elections, or whether or not they were defeated. In making these observations, we do not suggest that the district court need analyze every election since 1965 in order to determine whether the minority-preferred candidates are usually defeated. We leave to another day the question of precisely how many elections must be considered in order for a district court’s conclusions to be adequately supported. Cf.

Gingles, 478 U.S. at 57, 106 S.Ct. at 2769 (“[A] pattern of racial bloc voting that extends over a period of time is more probative of ... legally significant polarization than are the results of a single election.”); Hines v. Mayor and Town Council of Ahoskie, 998 F.2d 1266, 1272 (4th Cir.1993) (“[The] ‘results’ test [of

section 2 of the VRA] ‘supposes the need to consider multiple electoral contests.’ ” (quoting Baird v.

Consolidated City of Indianapolis, 976 F.2d 357, 359 (7th Cir.1992), cert. denied, 508 U.S. 907, 113 S.Ct. 2334, 124 L.Ed.2d 246 (1993))); Baird, 976 F.2d at 359 (“Any approach that depends on outcomes supposes the need to consider multiple electoral contests — the same position over many years, many positions during the same year, or both.”); Collins I, 816 F.2d at 936 (“[W]hether there is a ‘pattern of racial bloc voting that extends over a period of time’ [is] a critical factor in a claim of vote dilution under § 2.”); see also Uno, 72 F.3d at 985 (“[R]ace-eonseious politics (or its absence, for that matter) can more readily be seen by producing a documentary that spans a series of elections than by taking an isolated snapshot of a single election.”).

Today, we hold merely that it is insufficient to consider selectively only those elections in which minority candidates were on the ballot, at least where such elections are not such a substantial majority of the total elections that a fair assessment can be made of whether the minority-preferred candidates are usually defeated by white bloc voting. III. Plaintiffs challenge the methodology employed by the district court on numerous other grounds.

Specifically, plaintiffs urge: 1) that the district court improperly viewed as minority-preferred candidates some candidates who finished second and third among black voters; 2) that it improperly treated candidates in the general election as black-preferred when those candidates- assertedly received a majority of black votes only because they were Democrats; 3) that it improperly viewed success in the primary election as electoral success; and 4) that it failed to discount the weight attributed to the one repeatedly successful black candidate due to the effects of incumbency.

We address each of these challenges in turn, finding no reversible error by the district court. A. Plaintiffs first contend that, contrary to our holding in Collins v. City of Norfolk, Va., 883 F.2d 1232 (4th Cir.1989) (“Collins II”), cert. denied, 498 U.S. 938, 111 S.Ct. 340, 112 L.Ed.2d 305 (1990), the district court considered as “black-preferred” some successful candidates who finished second and third among black voters, behind a successful black-preferred candidate. We agree that the district court erred, but not in the manner plaintiffs suggest.

Rather, we conclude the district court erred in plaintiffs’ favor by not treating as “black-preferred” some successful candidates who placed second or third among black voters (and with a significant majority of black votes) behind an unsuccessful candidate who was the first choice of black voters. 1. Because Alamance County’s Board of Commissioners is elected from a single, multi-member district, citizens are permitted to vote for either two or three candidates in both the primary and general elections, depending on the number of seats up for election in any given year.

Consequently, the district court in most cases treated the top two or three vote-recipients among black voters as the black-preferred candidates, as did plaintiffs’ expert. If the first choice of black voters did not win an election, however, the district court excluded from consideration even as possible minority-preferred candidates those candidates who placed second or third among black voters behind the losing black-preferred candidate, in attempted reliance on our opinion in Collins II, 883 F.2d at 1238.

In so doing, the district court, as even counsel for the plaintiffs conceded at oral argument, seems to have misread our opinion in Collins II. We did not, in Collins II, adopt a per se rule against considering, as minority-preferred, candidates who finished second or third among minority voters behind an unsuccessful candidate who was the minority community’s first choice.

We held only that, [t]he mere election of a candidate who appears to have received votes from more than fifty percent of minority ballots does not count as a minority electoral success, when each ballot may contain votes for more than one candidate. In such a situation, if there were other candidates, preferred by a significantly higher percentage of the minority community, who were defeated in the same election, then it cannot be fairly said that the minority community has successfully elected representatives of its choice.

Each such situation must be reviewed individually to determine whether the elected candidates can be fairly considered as representatives of the minority community. The presumption must be that they cannot if some other candidate has received significantly more minority votes. Id. at 1238 (emphases added) (quoting Collins I, 816 F.2d at 937).

Collins II, in other words, requires that a two-step inquiry be undertaken by the court in order to determine whether candidates who receive less support from black voters than an unsuccessful first choice may nevertheless be deemed black-preferred candidates in multi-seat elections.

First, if the unsuccessful candidate who was the first choice among minority voters did not receive a “significantly higher percentage” of the minority community’s support than did other candidates who also received a majority among minority voters, Collins II, 883 F.2d at 1238 (quoting Collins I, 816 F.2d at 937); see also id. at 1239 (referring to the level of support given to the unsuccessful candidate as “much greater”), then the latter should also be viewed by the district court as minority-preferred candidates. Cf.

Niagara Falls, 65 F.3d at 1018 (treating as black-preferred a candidate who finished behind an unsuccessful first choice among black voters whose support “was not dramatically higher.”). Second, if the level of support received by the unsuccessful first-place finisher among black voters was “significantly higher” than the support given the second- and third-place finishers by those same voters, then successful candidates who finished behind the unsuccessful first choice “are presumed not to be the minority’s preferred candidates or representatives of choice.” Id. at 1238.

Even so, the district court must still review “[e]ach such situation ... individually to determine whether the elected candidates can be fairly considered as representatives of the minority community.” Id. This is to say that Collins II requires an individualized determination in the circumstance where the unsuccessful first choice of black voters received significantly more support than the next-place finishers of the black voters.

In this circumstance, the individualized determination is undertaken in order to ensure that the second- or third-place finishers, who, based upon their substantial support from the minority community would seem to be minority-preferred candidates, may in fact fairly be characterized as such. In this case, black candidate Morris, for example, received 98% of the black vote in the 1972 Democratic primary election, and white candidates Long, Fleming, and Horne received 84%, 29%, and 16% of the black vote, respectively. Long, Fleming, and Horne won the primary election.

Notwithstanding the impressive support that Long received from black voters in the primary, the district court did not even consider whether Long might have been a black-preferred candidate. Because Morris’ support among black voters was not “significantly higher” than Long’s, the district court appears to have erred in not automatically treating Long as a black-preferred candidate.

But even had Morris’ level of support among black voters been significantly higher than Long’s, the district court still would have erred in not making an individualized determination before rejecting Long as a black-preferred candidate. Morris’ support was significantly higher than the 29% received by Fleming. In such a circumstance, as in the case directly addressed by Collins II, a presumption against viewing Fleming as black-preferred is required by our holding in Collins II, which, as a panel, we are bound by.

However, the district court should have made án individualized determination to assess whether that candidate, too, could properly have been deemed black-preferred. The district court also declined to treat successful white candidate Newlin as a black-preferred candidate in the 1978 Democratic primary, merely because Harris, the first choice of black voters, lost. There were two seats up for election that year, and Harris and Newlin were the top two vote-recipients among black voters, by far, with an estimated 78% and 60% of the black vote, respectively.

Significantly, the third place finisher among black voters was a black candidate, who received only 12% of the black vote. Even were we to conclude that Harris’ support among black voters was “significantly higher” than- Newlin’s, an individualized determination would still have been required before declining to treat Newlin as a black-preferred candidate.

Again, we held in Collins II only that successful candidates who receive more than 50% of the black vote should not automatically be viewed as a black-preferred candidate if they receive significantly less support from black voters than another, losing candidate, not that they should never be treated as black-preferred. 2.

While the district court erred in automatically excluding, as possible minority-preferred candidates, the second- and third-place finishers behind an unsuccessful first choice among black voters, we do not believe the court erred, as plaintiffs contend, in not conducting an individualized determination with respect to every candidate who finished second or third among black voters behind successful black-preferred candidates, before treating them as black-preferred.

Plaintiffs object to the fact that the district court considered some of the second and/or third place finishers as black-preferred in the three primary elections in which the first choice of black voters was successful — 1976, 1980, and 1984. In the 1976 primary, for example, the district court treated both successful black candidate O’Kelley, who won 99 + % of the black vote, and successful white candidate Fleming, who finished second among black voters with 47% of the vote, as black-preferred.

The district court did not, however, consider successful white candidate Paris as black-preferred because he received only 16% of the black vote, even though that percentage placed him in a tie for third among black voters and thus may have been sufficient to win had the election been held only among black voters. See supra note 10. All three candidates went on to win in the general election, with 99 + % of the black community’s vote.

Plaintiffs object to the district court’s treatment of these second and third place finishers as black-preferred candidates on the ground that, “[i]n most of these elections, ... the black community fielded only one [black] candidate,” Appellant’s Br. at 20. In advancing this argument, plaintiffs rely not on the holding in Collins II, but on a single passage from that case in which we said the following: If black voters exercised their right to cast all of their allotted votes, they ran the risk that their second and third choices would be declared their preferred candidates.

Only by single-shot voting — withholding all votes save for their first choice, and forfeiting the opportunity to east all votes allotted to each voter — could the minority be assured that its second and third choices would not be declared its preferred candidates. In contrast, under the at-large system, the white voters can freely east all votes allotted to them without suffering the penalty imposed on the minority voters. 883 F.2d at 1239-40.

Because Collins II, as explained above, addressed only the treatment of second- and third-place finishers behind an unsuccessful, not a successful, minority-preferred candidate, this passage is obviously dictum and not controlling here. And even were it applicable, the passage does not suggest the level of support necessary in order for such candidates to be considered minority-preferred. We believe that it would be unwise to extend the holding in Collins II to instances in which the first choice among black voters was successful.

Where the first choice of black voters was successful, there is simply no reason to presume that the minority community has been unsuccessful in electing representatives of its choice. Accordingly, we now hold that, in multi-seat elections in which voters are permitted to cast as many votes as there are seats, at the very least any candidate who receives a majority of the minority vote and who finishes behind a successful candidate who was the first choice among the minority voters is automatically to be deemed a black-preferred candidate, just like the successful first choice. Cf.

Niagara Falls, 65 F.3d at 1018 n. 18 (noting that parties did not dispute that candidates who received more than 50% were black-preferred). Candidates who receive less than 50% of the minority vote, but who would have been elected had the election been held only among black voters, are presumed also to be minority-preferred candidates, although an individualized assessment should be made in order to confirm that such a candidate may appropriately be so considered. The district court appears generally to have conducted its analysis in the manner that we have described as appropriate.

It considered as black-preferred candidates only those candidates who received substantial support from black voters; it did not unquestioningly denominate as black-preferred the top two or three vote-recipients among the black community. Moreover, the court seems to have made an individualized determination before considering, as black-preferred, candidates such as Fleming who, in 1976, received slightly less than 50% of the black vote (47%), and in rejecting Paris, who in that same election only tied for third, with 16% of the black vote.

We accordingly discern no reason to disturb the district court’s determinations with respect to those candidates who finished second and third among minority voters, behind the minority community’s successful first choice. Those determinations are not clearly erroneous. See Gingles, 478 U.S. at 77-79, 106 S.Ct. at 2780-81. B. Plaintiffs next object to the fact that the district court treated, as black-preferred, Democratic candidates who received 99 + % of the black vote in general elections.

Plaintiffs contend that to consider these candidates as black-preferred ignores that blacks in Alamance County overwhelmingly support Democrats in partisan elections. They argue that Democrats should be treated as black-preferred only if they were also black-preferred in the primary election. Therefore, they would have us discount, for example, the general election victories of Fleming in 1976 and Paris in 1980, both of whom received nearly unanimous black voter support, merely because they received “only” 47% and 48%, respectively, of the black vote in the primary elections.

This argument is flawed in two fundamental respects. First, as we discuss infra, acceptance of this argument would require that we carry forward the candidates’ black-preferred status from the primary to the general election, when the Voting Rights Act is clearly concerned with whether blacks have an equal opportunity to elect the candidate of their choice in both nominations and elections.

See 42 U.S.C. § 1973(b) (“A violation ... is established if ... it is shown that the political processes leading to nomination or election ... are not equally open to participation by [citizens on account of race or color].” (emphasis added)); cf. NAACP v.

City of Columbia, S.C., 850 F.Supp. 404, 417 (D.S.C.1993) (noting, as ironic, the fact that second choice of black voters, behind unsuccessful first choice, who went on to prevail in run-off election with majority of support from black voters, would not be considered preferred by either blacks or whites if level of support received in the initial primary election was the benchmark), aff'd as modified on other grounds, 33 F.3d 52 (4th Cir.1994) (per curiam ), cert. denied, — U.S.-, 115 S.Ct. 1095, 130 L.Ed.2d 1064 (1995).

Thus, candidates who receive 99 + % of the black vote in general elections are the black-preferred candidate in that election, regardless of their level of support in the primary. We reject the proposition that “success of a minority-preferred candidate in a general election is entitled to less weight when a candidate with far greater [minority] support was defeated in the primary.” Niagara Falls, 65 F.3d at 1018.

Such a view is grounded in the belief that minority voters essentially take their marbles and go home whenever the candidate whom they prefer most in the primary does not prevail, a belief about minority voters that we do not share, And such a view ignores altogether the possibility that primary election winners will become the minority’s preferred candidate during the general election campaign, or that where, as here, the overwhelming majority of blacks vote in the Democratic primary, that a Republican could in fact become the black-preferred candidate in the general election by addressing himself to issues of interest to the minority community in a way that appeals to them as participants in the political process.

Second, under plaintiffs’ theory, the district court would be required, contrary to Gingles, to assess the cause of the minority community’s support of the candidate in determining whether the minority-preferred candidates usually are defeated by white bloc voting. See infra note 12.

In Gingles, defendants argued that the plaintiffs were required to use multiple regression analysis, rather than bivariate regression analysis, to ensure that only those defeats of black-preferred candidates that were actually caused by race, and not some other reason such as party affiliation, were considered defeats of black-preferred candidates for purposes of a vote dilution claim. The Gingles plurality explicitly rejected such a requirement in the context of assessing the second and third Gingles preconditions. 478 U.S. at 62, 106 S.Ct. at 2772.

Moreover, even the limited election data in this record demonstrate that black voters do not vote .monolithically, even in partisan general elections. White Democrats Holt and Thompson only received 69% and 83% of the black vote, respectively, in the 1984 general election, and Democrat Morris, who is black, received only 57% of the black vote in the 1986 general election — behind Democrat Bennett, who is white, who received 99 + % of the black vote.

We therefore hold that the district court properly deemed all general election candidates who received 99 + % of the black community’s vote as black-preferred, without reference to the underlying primary election. C. Plaintiffs next contend that the district court erred in not distinguishing between, and separately analyzing, primaries and general elections, in determining whether black-preferred candidates were “usually” successful. With this contention, we agree.

Section 2(

b) of the Voting Rights Act provides that a violation occurs if “the political processes leading to nomination or election ... are not equally open” to all voters. 42 U.S.C. § 1973(b) (emphasis added); see also 42 U.S.C. § 1973i(c)(1) (“The terms Vote’ or ‘voting’ shall include all action necessary to make a vote effective in any primary, special, or general election_”). The statute thus requires that minorities have an equal opportunity to participate not only in primary elections but also in general elections.

From this, we believe it follows that the results in these two phases of the single election cycle must be separately considered and analyzed, and, in recognition of this statutory requirement, that Gingles ’ third precondition can be satisfied by proof that, in either the primary or the general election, the minority-preferred candidate is usually defeated by white bloc voting.

Not to separately consider primary and general elections risks masking regular defeat in one of these phases with repeated successes in the other, and thereby misperceiving a process that is palpably in violation of the Voting Rights Act, as not violative of the Act at all.

By way of illustration, an at-large voting method under which the candidates preferred by the minority community are always successful in the primary election but, because of white bloc voting, are always defeated in the general election (or one in which minority-preferred candidates are essentially barred from the general election ballot because of racial-bloc voting in the primary, see White v.

Regester, 412 U.S. 755, 767, 93 S.Ct. 2332, 2340, 37 L.Ed.2d 314 (1973) (“[T]he black community has been effectively excluded from participation in the Democratic primary selection process.”)), is a voting method under which minorities are, because of white bloc voting, always unsuccessful in attaining elective office. By aggregating primary and general elections, however, a court would ineluctably find in both circumstances that minority-preferred candidates were successful fifty percent of the time, and therefore not “usually” defeated by white bloc voting.

The Supreme Court has held, in an analogous context, that courts must not rely on “aggregated” data “when considering several separate vote dilution challenges in a single case.” See Gingles, 478 U.S. at 59 n. 28, 106 S.Ct. at 2771 n. 28; see also id. at 101, 106 S.Ct. at 2792-93 (O’Connor, J., concurring in judgment) (“The District Court clearly erred in aggregating data from all of the ehal-lenged districts_”).

Because plaintiffs’ challenge is, in reality, two separate' vote dilution challenges&emdash;one to the at-large primary election scheme, and one to the at-large general election scheme&emdash;-we believe that like reasoning must obtain here. Accordingly, we hold that primary and general election results should not be aggregated for purposes of determining whether black-preferred candidates are “usually” successful, and that the district court erred in so doing. Because plaintiffs in this case would not prevail on their

Section 2 claim even were the primaries and general elections considered separately, however, we will not reverse the district court on account of this error. D. Plaintiffs’ final claim is that the district court erred in not discounting the weight attributed to O’Kelley’s electoral successes because he was an incumbent.

See Gingles, 478 U.S. at 57, 106 S.Ct. at 2770 (“[Sjpeeial circumstances, such as the absence of an opponent, incumbency, or the utilization of bullet voting, may explain minority electoral success in a polarized contest.” (emphasis added)); but see Smith, 687 F.Supp. at 1317 (“The materiality of [evidence of incumbency] is questionable.” (citing Gingles, 478 U.S. at 63, 106 S.Ct. at 2772-73 (plurality opinion))). • O’Kelley, one of the successful black-preferred candidates whom the district court considered in determining whether black-preferred candidates usually were defeated, won three straight general elections (and, apparently, 5 elections total, including primaries) after first being appointed to a vacant seat.

Although the district court provided no explanation of its decision not to discount O’Kelley’s successes, we will not disturb that decision, given that plaintiffs’ own expert did not analyze the effects of incumbency for O’Kelley or for any other candidate. To reverse as clearly erroneous the district court’s decision under these circumstances would be to transform what was at most a narrow or “special” circumstance envisioned by the Court only in dicta into a categorical rule that all electoral successes of a minority-preferred incumbent are to be discounted.

See, e.g., Clarke, 40 F.3d at 813-14 (“[I]ncumbency plays a significant role in the vast majority of American elections. To qualify as a ‘special’ circumstance, then, incumbency must play an unusually important role in the election at issue; a contrary rule would confuse the ordinary with the special, and thus ‘make practically every American election a “special circumstance.” ’ ” (quoting Collins II, 883 F.2d at 1250 (Chapman, J., dissenting)).

Of course, as the defendants point out, and as counsel for the plaintiffs conceded at argument, “if incumbency were to be considered to discount the weight to be given to the success of one black candidate, O’Kelley, it likewise would have to be used to discount the weight given to the defeat of other black candidates who lost to incumbents, such as Harris in 1978 and Freeman in 1984.” Appellee’s Br. at 24, citing Nipper v.

Smith, 39 F.3d 1494, 1538-39 (11th Cir.1994)). # * * * * * The purpose of the Voting Rights Act is to ensure to minority citizens an equal opportunity to participate in the democratic processes and to elect candidates of their choice, candidates whom they believe will best represent their political interests. The Act’s purpose is not to ensure the election of candidates of the minority’s race, nor is it to ensure the election of candidates of any particular political party.

In guaranteeing that minorities will be afforded an equal opportunity to participate in the political processes and to elect representatives of “their choice,” the Act at once embodies the presumption and embraces the ideal that, in this Country, a candidate of any color can represent citizens of all colors, that a candidate of either party can represent the interests of all parties. Few presumptions are more fundamental to our system of representative democracy or more revealing of our commitment to equality and individual autonomy.

At bottom, through their collective arguments, plaintiffs urge a reading of the Supreme Court’s decision in Gingles, and therefore an

interpretation of

Section 2, that would be “at war” with this fundamental presumption, this ideal, which we believe not only informed Congress’ enactment of

Section 2, but underlay the Court’s*

interpretation of that provision in Gingles. The reading of Gingles urged upon us would not simply guarantee to minority voters an equal opportunity to elect representatives of their choice, but would require the election of minority candidates: “Under the law, black voters must have an opportunity not just to elect candidates of their choice, but to elect black candidates of their choice.” Appellant’s Br. at 30. Such an

interpretation, of course, is grounded ultimately in the twin presumptions that all members of a particular racial group will always prefer the same candidate and that that preferred candidate will always be of that group’s race. We simply disagree that Gingles requires, or even permits, any such presumptions; indeed, we believe Gingles forbids such presumptions based upon race alone.

And to the extent it can be read otherwise, we are convinced the Court would not so read the case today, given its unequivocal pronouncements that the principles of equal protection are offended by “assumption[s] that voters of a particular race, because of their race, ‘think alike, share the same political interests, and will prefer the same candidates at the polls’.” Miller v. Johnson, — U.S.-,-, 115 S.Ct. 2475, 2486, 132 L.Ed.2d 762 (June 29, 1995) (quoting Shaw v.

Reno, 509 U.S. 630, 647, 113 S.Ct. 2816, 2827, 125 L.Ed.2d 511 (1993)); see also Shaw, 509 U.S. at 648, 113 S.Ct. at 2827 (“ ‘The principle of equality is at war with the notion that District A must be represented by a Negro, as it is with the notion that District B must be represented by a Caucasian, District C by a Jew, District D by a Catholic, and so on.’” (quoting Wright v. Rockefeller, 376 U.S. 52, 66, 84 S.Ct. 603, 611, 11 L.Ed.2d 512 (1964) (Douglas, J., dissenting))). The judgment of the district court is affirmed. AFFIRMED. .

The census data of 1990 shows that Alamance County’s population is 79.8% white, 19.2% black. Voting age population is 81.2% white, 18.0% black. As of April 1994, the population of registered voters was 84.1% white, 15.5% black. While 59% of all registered voters in the county are Democrats, blacks in Alamance County are overwhelmingly Democrats (94%). For this latter reason, when we refer to primary elections, we are referring only to the Democratic primaries. . The full statute provides as follows: (

a) No voting qualification or prerequisite to voting or standard, practice, or procedure shall be imposed or applied by any State or political subdivision in a manner which results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color, or in contravention of the guarantees set forth in

section 1973b(f)(2) of this title, as provided in subsection (

b) of this section. (

b) A violation of subsection (

a) of this

section is established if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens protected by subsection (

a) of this

section 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. The extent to which members of a protected class have been elected to office in the State or political subdivision is one circumstance which may be considered: Provided, That nothing in this

section establishes a right to have members of a protected class elected in numbers equal to their proportion in the population. . In the face of the County’s intended challenge to the use of such statistical evidence, we again caution against overreliance on bivariate ecological regression analysis in the estimation of voter preferences for purposes of a vote dilution claim. See, e.g., Smith v. Brunswick County, Va., Bd. of Supervisors, 984 F.2d 1393, 1400 n. 6 (4th Cir.19931.

Bivariate regression analysis, as used in Voting Rights Act cases, only allows one to determine the correlation between the percentage of the vote received by a particular candidate and the racial composition of voting precincts. See, e.g., Jenkins v. Red Clay Consol. School Dist. Bd. of Educ., 4 F.3d 1103, 1119 n. 10 (3d Cir.1993) (citing Richard L. Engstrom & Michael D. McDonald, Qualitative Evidence in Vote Dilution Litigation: Political Participation and Polarized Voting, 17 Urb.

Law. 369, 374 (Summer 1985)), cert. denied, - U.S. -, 114 S.Ct. 2779, 129 L.Ed.2d 891 (1994); City of Carrollton Branch of the NAACP v. Stallings, 829 F.2d 1547, 1556 (11th Cir.1987), cert. denied, 485 U.S. 936, 108 S.Ct. 1111, 99 L.Ed.2d 272 (1988). The relevant correlation for vote dilution claims, however, is that between the race of the voters and the votes cast for a particular candidate, not that between the racial composition of precincts and the votes cast for a particular candidate, for it is only the former from which one can accurately estimate actual voter preferences.

One can derive estimates of voter preferences from the regressions employed here only by assuming that the level of minority support for the candidate is fairly constant across precincts — an "ecological” assumption that runs counter to the common sense observation that blacks and whites who live in integrated neighborhoods are more likely to vote for candidates of another race.

See, e.g., Bernard Grofman et at, Minority Representation and the Quest for Voting Equality 89 (1992) ("The principal disadvantage [of ecological regression is] the possibility of errors owing to ecological inference.”); id. ("[T]he ‘ecological fallacy’ ... is the error of attributing the average behavior of voters in a given geographic area (ecological unit) to all voters in that area.").

In other words, the bivariate ecological regression analysis merely assumes that which it is the purpose of the analysis to prove — political cohesiveness and constancy of candidate preference among minority voters and racial bloc voting among majority voters. See Overton v. City of Austin, 871 F.2d 529, 539 n. 12 (5th Cir.1989) ("[Bivariate ecological regression analysis] assumes that the voters in most precincts voted according to their ethnicity.”); id. at 539 ("|T]rial court should not ignore the imperfections of the data used nor the limitations of [bivariate ecological regression] analysis”). .

We do not imply that the third Gingles element is met if plaintiffs merely show that white bloc voting defeats the minority-preferred candidate more often than not. The terms used by the Gingles Court are “usually," “normally,” and "generally.” 478 U.S. at 49, 51, 56, 106 S.Ct. at 2766, 2766-67, 2769. We need not in this case specify a meaning for these terms; suffice it to say that they mean something more than just 51%. See Clarke v.

City of Cincinnati, 40 F.3d 807, 812-13 (6th Cir.1994) (finding that 47% success rate was "no reason to find that blacks' preferred black candidates have ‘usually’ been defeated.”), cert. denied, - U.S. -, 115 S.Ct. 1960, 131 L.Ed.2d 851 (1995); Uno v. City of Holyoke, 72 F.3d 973, 985 (1st Cir.1995) ("|T]o be legally significant, racially polarized voting in a specific community must be such that, over a period of years, whites vote sufficiently as a bloc to defeat minority [-preferred] candidates most of the time.” (emphasis added)).

Indeed, Justice O’Connor suggested in Gingles that a finding of "usually defeated” should be based on a success rate by minority-preferred candidates that is less than proportional representation, although greater than mere tokenism. 478 U.S. at 99-100, 106 S.Ct. at 2791-92 (O'Connor, J., concurring in judgment) (“[A] reviewing court should be required to find more than simply that the minority group does not usually attain an undiluted [that is, proportional] measure of electoral success.

The court must find that even substantial [that is, less than proportional but more than token] minority success will be highly infrequent under the challenged plan before it may conclude, on this basis alone, that the plan operates 'to cancel out or minimize the voting strength of [the] racial grou[p].’ ” (quoting White v.

Regester, 412 U.S. 755, 755, 93 S.Ct. 2332, 2339, 37 L.Ed.2d 314 (1973))); see also Jenkins, 4 F.3d at 1122 ("If minority-preferred candidates are consistently able to prevail in representative numbers ..., then it cannot be said that white voters vote sufficiently as a bloc usually to prevent the election of the minority preferred candidates." (emphasis added)). .

Although Justice White, who joined the remainder of Justice Brennan’s opinion, did not join this portion of the opinion, neither Justice White nor Justice O’Connor, who concurred only in the judgment, as much as hinted that only elections in which a black candidate was on the ballot need be considered; they each merely argued that Justice Brennan’s conclusion that the race of the candidate is always irrelevant was in conflict with Whitcomb v. Chavis, 403 U.S. 124, 91 S.Ct. 1858, 29 L.Ed.2d 363 (1971), and was not necessary to the disposition of the case.

Id. at 82, 83, 106 S.Ct. at 2783, 2783 (White, J., concurring); id. at 83, 101, 106 S.Ct. at 2783, 2792-93 (O'Connor, J., concurring in the judgment, joined by Burger, C.J., and Powell and Rehnquist, JJ.). .

Although the First Circuit referred in this passage to "minority candidates,” not to "minority-preferred candidates,” it is evident from its citation to Gingles at p. 56, 106 S.Ct. at 2769 that the court meant by "minority candidate" the "minority-preferred candidate.” At that page in Gingles, the Supreme Court refers only to the "minority’s preferred candidates” and the "representatives of their choice.” It does not once refer to minority candidates. . See also Smith v.

Clinton, 687 F.Supp. 1310, 1316 (E.D.Ark.1988) (three-judge panel) ("We assume without deciding that all of the evidence [of white-white elections] offered by the defendants is admissible and properly to be considered.”); Brown v. Board of Commissioners of the City of Chattanooga, 722 F.Supp. 380, 391 (E.D.Tenn.1989) ("[A] review of white/white elections as well as white/black elections is necessary to determine the existence of racially polarized voting.”); Southern Christian Leadership Conference of Alabama v.

Evans, 785 F.Supp. 1469, 1473 (M.D.Ala.1992) ("Court will ... 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. The Court must keep in mind the mandate of the statute that-

Section 2 does not require that minorities be given equal opportunity to elect minority candidates, but that they be given equal opportunity to 'elect representatives of their choice.'"). . We do not decide here the extent to which, if any, white-white elections are deserving of less evidentiary weight than elections in which a minority candidate is on the ballot.

It seems to us, however, that if white-white elections are entitled to less weight, then they are so only on the question of whether racial polarization exists, not on the question of whether, because of that polarization, minority-preferred candidates are usually defeated. Cf.

Harvell, 71 F.3d at 1382 (distinguishing between "racial cohesiveness," the second Gingles precondition, and “racial polarization, ... the essential element of the third Gingles precondition,” and holding that in measuring whether racial polarization "permits the majority usually to defeat the minority's preferred candidates,” evidence such as "elections in which there was no African-American candidate [on the ballot] ... become[s] relevant.”). . Three other white candidates received 9%, 8% and 5% of the black vote, respectively. .

Our holding in Collins II actually does not address this circumstance directly, since Fleming received less than majority support among black voters; Collins II addressed only the circumstance where a candidate receives more than 50% of the minority community's vote. Nevertheless, we do not believe that the mere failure to achieve a threshold of 50% in a multi-candidate election necessarily means that a candidate cannot be viewed as a black-preferred candidate.

If the election had been held only among black voters, a hypothetical inquiry given some credence by the Gingles plurality, see 478 U.S. at 68, 106 S.Ct. at 2775 ("The essence of a submergence claim is that minority group members prefer certain candidates whom they could elect” if the election were held in a majority-black district), then Fleming would have won, because he was the third-place finisher among black voters in an election in which three seats were to be filled. .

The level of support that may properly be deemed "substantial” will vaiy, of course, depending on the number of candidates on the ballot and the number of seats to be filled. . Although only a four-Member plurality joined

Section III(

C) of the opinion, in which the rejection of defendant's multiple regression claim is found,

Section III(

C) is lengthy, and the dispute that cost Justice White's vote on the

section was not over the "cause” issue.

Although most of our sister circuits have by now adopted the position that an inquiry into cause is relevant, see, e.g., LULAC, 999 F.2d at 850 ("[The] rigorous protections [of the Voting Rights Act], as the text of § 2 suggests, extend only to defeats experienced by voters ‘on account of race or color.’ ”), they are still in disagreement about the stage in the vote dilution inquiry in which causation evidence is appropriate, compare Nipper, 39 F.3d at 1515 ("[I]f the evidence shows, under the totality of the circumstances, that the community is not motivated by racial bias in its voting patterns, then a case of vote dilution has not been made.”) with Clarke, 40 F.3d at 812-14 (considering "cause” of the success rate of black-preferred candidates in assessing third Gingles element).

We think the best reading of the several opinions in Gingles, however, is one that treats causation as irrelevant in the inquiry into the three Gingles preconditions, see Gingles, 478 U.S. at 62, 106 S.Ct. at 2772 (Brennan, J., plurality opinion), but relevant in the totality of circumstances inquiry, see 478 U.S. at 100, 106 S.Ct. at 2792 (O'Connor, J., concurring in judgment) (agreeing that use of racial polarization data “solely” to prove Gingles second and third elements cannot be rebutted by “evidence that the divergent racial voting patterns may be explained in part by causes other than race,” but disagreeing with plurality's claim "that such evidence can never affect the overall vote dilution inquiry.”).

Of course, whether causation is relevant to the second and third precondition inquiries or to the totality of circumstances inquiry, it would be just as relevant for determining whether particular minority-preferred candidates lose for some reason other than racial polarization, as it would be for determining, as plaintiffs urge, whether particular minority-preferred candidates win merely because of party affiliation.