Full opinion text
MEMORANDUM AND ORDER SELYA, Circuit JUDGE, sitting by designation. In the spring of 2001, the Massachusetts legislature (the Legislature) redrew the dimensions of its House and Senate districts. The end product of that endeavor— the 2001 Redistricting Act- — forms the backdrop for this litigation. The plaintiffs, African-American and Hispanic voters, complain that the redistricting scheme, as it pertains to House districts in the Boston area, infringes upon rights guaranteed to them by the United States Constitution and denies them equal opportunity “to participate in the political process and to elect representatives of their choice” in violation of section 2 of the Voting Rights Act (VRA), 42 U.S.C. § 1973(b). After careful consideration of the parties’ plenitudinous submissions, we conclude that the Redistricting Act deprives African-American voters of the rights guaranteed to them by section 2 of the VRA. This conclusion renders it unnecessary to decide whether the Redistricting Act also (i) impinges upon the plaintiffs’ constitutional rights, or (ii) deprives Hispanic voters of the rights guaranteed to them by section 2. We start with certain background information and then turn to the merits. At that juncture, we apply the test set forth in Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986), make a series of factual findings, and explain why the plaintiffs prevail on their claim that the Redistricting Act infringes upon rights protected by section 2 of the VRA. We conclude by discussing the development of an appropriate remedy. 1. BACKGROUND We rehearse, in broad brush, the evolution of the Redistricting Act. Thereafter, we recount the procedural history of the litigation. A. The Evolution of the Redistricting Act. In the wake of the most recent decennial census, the Legislature established a joint special committee on redistricting and reapportionment (the Joint Committee) to review existing legislative districts, formulate revisions reflecting the increase in the Commonwealth’s population from 6,016,425 to 6,349,097 residents, and redraw the district lines. This is standard operating procedure, designed to ensure compliance with the one-person, one-vote requirement of the United States Constitution. See Georgia v. Ashcroft, 539 U.S. 461 n. 2, 123 S.Ct. 2498, 2516 n. 2, 156 L.Ed.2d 428 (2003) (“When the decennial census numbers are released, States must redistrict to account for any changes or shifts in population.”); Gray v. Sanders, 372 U.S. 368, 379, 83 S.Ct. 801, 9 L.Ed.2d 821 (1963) (locating the source of the one-person, one-vote requirement in the Equal Protection Clause of the Fourteenth Amendment); see generally Black Political Task Force v. Connolly, 679 F.Supp. 109, 123-30 (D.Mass.1988) (three-judge court) (outlining relevant public policies vis-á-vis one-person, one-vote requirement in connection with 1987 reapportionment of Massachusetts House of Representatives). Insofar as state legislative seats were concerned, the Joint Committee functioned as an integrated body in name only: the House delegation delineated the 160 new House districts and the Senate delegation independently delineated the forty new Senate districts. This case focuses exclusively on the redistricting activities of the House delegation (the Committee). The Speaker of the House, Thomas M. Finneran, named Representative Thomas M. Petrolati to chair the Committee. With the one-person, one-vote requirement in mind, Petrolati and his fellow Committee members determined that the ideal population for each House district was 39,682, plus or minus 5%, that is, somewhere between 37,698 and 41,666 persons. Aided by a specialized computer software program known as Maptitude, they then embarked on the task of shifting the district lines to achieve this numerical goal. The Committee (and the House as a whole) apparently was content to leave the heavy lifting to Finneran, Petrolati, their aides, and the Committee staff. Finneran and Petrolati kept the process on a short leash. As it evolved, Petrolati met one by one with the other 159 members of the House to discuss their concerns. The Committee then held five public hearings at divers locations throughout the Commonwealth. Apart from these public hearings, the Committee did not meet as a body. Moreover, Petrolati neither solicited nor accepted views from community leaders. On October 18, 2001, the Committee filed House Bill No. 4700 (the Committee Plan), which proposed 160 reshaped House districts. The accompanying report described the Committee’s redistricting goals, which included complying with the one-person, one-vote requirement, ensuring contiguity and compactness, acknowledging communities of interest, and attempting to create “minority influence districts.” With regard to this last goal, the report explained: A number of [the newly proposed] districts in Boston and Springfield contain a sizable majority comprising people of color. One proposed district of Boston presently has no incumbent and is over two-thirds minority. This increased the chances of there being an additional minority House member from the City of Boston. The Committee simultaneously wrote to House members, over Petrolati’s signature, lauding the “new minority-majority district in the Roxbury section of Boston,” touting the fact that it was incumbent-free, and representing that minorities composed 68% of its population. Just four days later, the House debated the Committee Plan. Various amendments were proposed and some were incorporated into the final redistricting scheme (the Enacted Plan). Only one of these amendments is relevant here — the so-called Fitzgerald Amendment. This amendment had its genesis in Representative Kevin Fitzgerald’s vacillation over whether to retire from the House. When the Committee drafted its plan, it assumed that Fitzgerald had no interest in running for reelection. Consequently, it proceeded to move the residence of Representative Elizabeth Malia from her previous district (the 11th Suffolk) to Fitzgerald’s home turf (the 15th Suffolk). It then made the 15th Suffolk a majority white district and retained the majority minority character of the 11th Suffolk district — the very majority minority district heralded in the Committee’s report and letter of transmittal. During the course of the floor debate, Fitzgerald (who is white) let it be known that, contrary to the Committee’s assumption, he planned to run for reelection. To facilitate his candidacy, he offered an amendment that had three notable effects on the Committee Plan: it returned Ma-lia’s residence to the 11th Suffolk district, changed the contours of that district so that it would be majority white, and kept the 15th Suffolk as a majority white district in which Fitzgerald would be the lone incumbent. The House adopted this amendment, thereby eliminating the majority minority district that had been a selling point of the original Committee Plan. Like the other floor amendments, the Fitzgerald Amendment was reviewed only for compliance with the one-person, one-vote requirement before being enacted. The Senate adopted the Enacted Plan without substantive debate (we imply no criticism; the House, with equal alacrity, embraced the Senate’s handiwork). On November 8, 2001, the governor signed the Redistricting Act, Chapter 125 of the Acts of 2001, into law. B. The Travel of the Case. On June 13, 2002, African-American and Hispanic voters filed a complaint in this court challenging the Redistricting Act as it applies to House districts in the Boston area. The complaint alleged that the Enacted Plan transgressed the Fourteenth and Fifteenth Amendments as well as section 2 of the VRA. The chief judge of the First Circuit convened this three-judge panel to hear and determine these claims. See 28 U.S.C. § 2284(a). Pretrial discovery proved to be an intensive and demanding process, during which defense witnesses routinely invoked the protections of legislative privilege when asked about the inner workings of the redistricting process. Despite this impediment, the plaintiffs mustered a sufficient case to warrant a trial. At trial, the plaintiffs attempted to show that the Redistricting Act set the boundaries of the House districts in and around Boston in a way that impermissibly diluted the voting strength of African-Americans and Hispanics, notwithstanding the growing population of these two minority groups; that the Committee failed to provide a reasonable explanation for the lines that were drawn; that the Committee deviated from conventional redistricting procedures by, among other things, allowing staffers from Speaker Finneran’s office to play key roles in the process and refusing to consult with minority community leaders; and that the Committee simultaneously packed minority voters into a tiny number of districts and splintered the minority vote in other (more variegated) districts in an effort to protect white incumbents. With respect to this last point, testimony for the plaintiffs focused largely on what they described as the intentional packing of the 6th Suffolk House district and the stripping of African-American voters from the 11th, 12th, and 15th Suffolk districts. The plaintiffs emphasized that, despite the increase in Boston’s minority population from 1990 to 2000 and the concomitant decrease in the white population, the Redistricting Act reduced the number of majority minority districts while increasing the number of majority white districts. They proffered two alternate redistricting schemes as potential remedies for this alleged imbalance. Under Massachusetts law, William F. Galvin, the secretary of state, is charged with the administration and oversight of state elections. See Mass. Gen. Laws ehs. 50-57. The plaintiffs filed their suit against several defendants, including Gal-vin in his official capacity. After the other defendants were dropped as a result of case management determinations, Galvin became the sole defendant. Meanwhile, Petrolati intervened and functioned as the legislative defendants’ designated representative. As a practical matter, the plaintiffs’ grievances are with the House leadership. Having that in mind, we henceforth will refer to Galvin and Petrolati as “the defendants.” Throughout the trial, the defendants maintained that the Enacted Plan provides African-American and Hispanic voters with equal opportunities to participate in the political process and to elect representatives of their choice. The defendants also disclaimed any intent or purpose to deny the plaintiffs equal voting opportunities on account of their race, and provided evidence indicating that incumbency protection was the driving force behind the drawing of the district lines. Finally, the defendants urged us to refrain from engaging in an unnecessary “tinkering exercise” because the Enacted Plan furnishes African-American and Hispanic voters with electoral opportunities that are no worse than “roughly proportional” to their percentage of the relevant population. Both sides presented extensive expert testimony in support of their respective positions. Following prolific briefing and final arguments, we took the matter under advisement. This memorandum comprises our findings of fact and conclusions of law; II. ANALYSIS Section 2 of the Voting Rights Act provides in pertinent part: (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 ... (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 politi cal process and to elect representatives of their choice. 42 U.S.C. § 1973 (emphasis supplied). To prevail under this statute, plaintiffs need not establish that the Legislature acted with a discriminatory purpose; it suffices to prove that the contested standard, practice, or procedure has a discriminatory effect. Bush v. Vera, 517 U.S. 952, 976, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996); Gingles, 478 U.S. at 35-36, 106 S.Ct. 2752. We caution, however, that section 2 does not turn electoral politics into an elaborate series of quotas. Despite its focus on results, the statute does not create a right to have members of a protected class actually elected in numbers equal to their proportion of the population. Johnson v. De Grandy, 512 U.S. 997, 1014 n. 11, 114 S.Ct. 2647, 129 L.Ed.2d 775 (1994). It merely guarantees minorities a level playing field, that is, an opportunity for victory at the polls equal to the opportunity enjoyed by others. Vecinos De Barrio Uno v. City of Holyoke, 72 F.3d 973, 979 (1st Cir.1995). Although relieved of the burden of proving intentional discrimination, plaintiffs who bring section 2 challenges still face formidable hurdles. In Gingles, the Supreme Court limned three threshold conditions that must be fulfilled in order to mount a successful vote dilution claim. First, the plaintiffs must demonstrate that they are part of a minority group that is “sufficiently large and geographically compact to constitute a majority in a single-member district.” 478 U.S. at 50, 106 S.Ct. 2752. Second, they must show that the group is “politically cohesive.” Id. at 51, 106 S.Ct. 2752. Third, they must prove “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.” Id. (citation omitted). Plaintiffs who satisfactorily complete this three-step pavane are not home free. Gingles demands a further sifting of relevant facts. This wide-ranging inquiry must dig into the totality of the circumstances, starting with the litany of factors delineated in the Senate report that accompanied passage of the 1982 amendments to section 2. Id. at 43-46, 106 S.Ct. 2752. The list includes (i) the extent of any history of official discrimination at the polls in the relevant jurisdiction; (ii) the frequency with which members of the minority group have been elected to public office in that jurisdiction; (iii) the presence of voting practices or procedures that stand as obstacles to minority participation; (iv) the extent to which members of the minority group must endure the effects of discrimination in such fields as education, employment, and health care; (v) the level of responsiveness exhibited by elected officials vis-á-vis the particular needs of the minority group; and (vi) the magnitude of racially polarized voting within the relevant jurisdiction. See S.Rep. No. 97-417, at 28-29 (1982), reprinted in 1982 U.S.C.C.A.N. 177, 206-07. Comprehensive as this compendium may appear, it is not exclusive; other factors may be equally instructive in particular cases. One such factor is proportionality — the relationship between the number of majority minority voting districts and the minority group’s share of the relevant population. De Grandy, 512 U.S. at 1000, 1014 n. 11, 114 S.Ct. 2647. Another is related to causation — the extent to which influences apart from racial bias may have caused the white bloc voting identified by the third prong of the Gingles formulation. See Vecinos, 72 F.3d at 983 & n. 4 (noting that other possible influences may include “organizational disarray, lack of funds, want of campaign experience, the unattractiveness of particular candidates, or the universal popularity of an opponent”). When all is said and done, the critical issue in a vote dilution case is whether the challenged districting scheme deprives minority voters of an equal opportunity to participate in the electoral process and to elect candidates of their choice. Id. at 980. In the pages that follow, we conduct our assessment of the plaintiffs’ case, using the Gingles framework. A. Numerosity. The overarching purpose of the first Gingles precondition is to ensure that, in the absence of the challenged practice, procedure, or structure, the minority group would have the potential to elect a representative of its choice in the relevant political subdivision. See Growe v. Emison, 507 U.S. 25, 40, 113 S.Ct. 1075, 122 L.Ed.2d 388 (1993) (extending the Gingles rubric to single-member districts); see also Gingles, 478 U.S. at 50 n. 17, 106 S.Ct. 2752 (observing that without this showing, the plaintiffs cannot claim to have been injured by the challenged practice, procedure, or structure). In a vote dilution case, this showing requires the plaintiffs to demonstrate that an effective and feasible remedy exists. Typically, this demonstration is accomplished by proffering an alternative plan that contains more than the existing number of reasonably compact districts with minority populations large enough to elect minority-preferred candidates. See, e.g., De Grandy, 512 U.S. at 1008, 114 S.Ct. 2647. An inquiring court normally will use the plaintiffs’ alternative plan as a benchmark against which to take the measure of the challenged scheme. See Holder v. Hall, 512 U.S. 874, 880, 114 S.Ct. 2581, 129 L.Ed.2d 687 (1994). At first blush, this aspect of the Gingles test seems fairly straightforward. Appearances can be deceiving, however, and courts have struggled with a host of complexities implicated by its formulation. Although we need not catalogue all of these complexities, a few of them are of particular pertinence here, and, thus, deserve further elaboration. We note, first, that the Supreme Court has explicitly declined to resolve whether plaintiffs may satisfy the first prong of Gingles with evidence that the minority group, although not comprising an absolute majority of the population within a given district, nevertheless wields enough influence that it usually will be able to elect candidates of its choice with the help of crossover votes. See, e.g., De Grandy, 512 U.S. at 1008-09, 114 S.Ct. 2647 (leaving this question open); Voinovich v. Quilter, 507 U.S. 146, 154, 113 S.Ct. 1149, 122 L.Ed.2d 500 (1993) (same); Growe, 507 U.S. at 41 n. 5, 113 S.Ct. 1075 (same). Because this issue is not outcome-determinative in this case, we assume, arguendo, the correctness of the bright-line rule followed by most courts: a minority population is sufficiently large to constitute a “majority” for the purposes of the first Gingles precondition only if it comprises more than 50% of the relevant population in the affected district. See, e.g., Valdespino v. Alamo Heights Indep. Sch. Dist., 168 F.3d 848, 852-53 (5th Cir.1999); Cousin v. Sundquist, 145 F.3d 818, 828-29 (6th Cir.1998); McNeil v. Springfield Park Dist., 851 F.2d 937, 943-45 (7th Cir.1988); Parker v. Ohio, 263 F.Supp.2d 1100, 1104-05 (S.D.Ohio) (three-judge court), aff'd mem., — U.S. -, 124 S.Ct. 574, 157 L.Ed.2d 426 (2003). This does not mean that the existence of such “influence” or “crossover” districts is entirely inconsequential to the case; they are relevant in assessing the totality of the circumstances, see Vecinos, 72 F.3d at 979 n. 2, 990-91, and we shall return to them in that setting. However, we reject the defendants’ effort to use influence or crossover districts as a means of undercutting the plaintiffs’ showing on the first Gingles precondition (by, for example, pointing to districts with African-American populations just under the 50% line to boost the number of “effective” majority minority districts created by the Enacted Plan). These concerns are adequately accounted for not only in addressing the totality of the circumstances but also in the analysis required by the third Gingles precondition. After all, if African-American voters are in fact able to convert influence or crossover districts into “effective” majority minority districts, this circumstance should result in the plaintiffs’ failure to prove that majority white bloc voting regularly operates to defeat the preferred candidate of African-American voters. A second unresolved issue is whether, in determining the size of the minority group in a particular district, a court should concern itself with the percentage of the minority population that is simply of voting age (VAP) or the percentage of the minority population that is composed of citizens who are of voting age (CVAP). To date, the Supreme Court has refrained from choosing between these indices. See De Grandy, 512 U.S. at 1021 n. 18, 1023, 114 S.Ct. 2647 (forgoing a decision between VAP and CVAP in assessing proportionality, but using VAP figures); cf. Growe, 507 U.S. at 38 n. 4, 113 S.Ct. 1075 (declining to choose between total population and VAP). Although the First Circuit has not spoken directly to the point, other courts of appeals lately have tended to support the use of CVAP data, at least when that data is readily available and its use is not contraindicated by other circumstances. See, e.g., Perez v. Pasadena Indep. Sch. Dist., 165 F.3d 368, 372 (5th Cir.1999); Negrón v. City of Miami Beach, 113 F.3d 1563, 1568-69 (11th Cir.1997); see also Barnett v. City of Chicago, 141 F.3d 699, 704-05 (7th Cir.1998) (stating, in dictum, that CVAP is the proper yardstick for measuring proportionality in a section 2 vote dilution case). The cardinal purpose of the first Gingles precondition is to determine whether minority voters have the potential to elect preferred candidates. See Gingles, 478 U.S. at 50 n. 17, 106 S.Ct. 2752 (referring to the potential of minority voters to elect representatives). Because non-citizens cannot vote (or even register to vote), the use of CVAP data, when and where available, seems more concinnous than the use of VAP data. Here, however, we need not make a definitive choice. For one thing, no one has pointed to any material differences in the available VAP and CVAP figures for African-Americans in Boston and, according to the defendants’ expert, non-Hispanic whites and non-Hispanic blacks register to vote at similar rates (56% and 57% of the voting age population, respectively). For another thing, VAP figures may be unreliable when analyzing Hispanic populations, where there are greater numbers of recent immigrants, see Negrón, 113 F.3d at 1568, but there is no serious reason to believe that they are less reliable when analyzing African-American populations. Then, too, the Legislature used only VAP data in drawing the Enacted Plan, and the parties have provided only VAP data for certain analy-ses (e.g, the second and third Gingles preconditions). For these reasons, we employ VAP data in our ensuing analyses. We note, however, that we would reach the same result vis-a-vis the first Gingles precondition utilizing either set of data (VAP or CVAP). A third unresolved issue concerns the appropriate method of characterizing voters as members of a particular race. Unlike previous censuses, the 2000 census permitted interviewees to identify themselves as members of more than one race. This means that there are now two different methods for assigning persons to racial groups. Under the first method — referred to as the “alone” method — a person who has identified himself or herself only as “black” will be counted as black, but a person who has checked both the box for “black” and the box for, say, “Asian” will be counted only as multiracial. Under the second method — referred to as the “combined” or “combo” method — a person who has identified himself or herself as a member of more than one race will be counted in every such category (in our earlier example, the responder will be counted as both black and Asian). The plaintiffs urge us to use the alone method because it avoids an overcount (by placing an individual voter in two or more racial categories, the combo method results in totals exceeding 100% of actual responders) while the defendants urge us to use the combo method because it is more comprehensive in its representation of voter identities (it places in any given racial category everyone who considers himself or herself a member of that race). This is a recent development in election law and the jurisprudence is embryonic. The Supreme Court appears to have employed combo (or combo-like) data in Georgia v. Ashcroft, 123 S.Ct. at 2507 n. 1, but it did so with regard to a section 5 preclearance claim, not a section 2 vote dilution claim, and we regard the question as open. Indeed, there may well be no single rule; context and the availability of data may determine which method best reflects the realities of a specific situation. In all events, the circumstances of this case render it unnecessary to choose one method over the other. As the defense acknowledged, through its expert witness and' in its trial brief, there are no great differences in this case between the two data sets, at least insofar as the second and third Gingles preconditions are concerned. Accordingly, we will provide both sets of data with respect to the first Gin-gles precondition in order to demonstrate that the choice is not outcome-determinative. As for the remaining Gingles preconditions and the totality of the circumstances, we will use an admixture of alone and combo data, determined primarily by the form in which the parties have furnished relevant data to us. We turn now to the proof on the first Gingles precondition. At trial, the plaintiffs established that the African-American population in Suffolk County — which encompasses all of the city of Boston and a few environs — is sufficiently large and compact to allow for the creation of more majority black districts than currently exist. Using the alone method, the Enacted Plan provides for one majority black district in Suffolk County (the 6th) whereas Plan No. 2 — which we deem to be the more useful of the plaintiffs’ two submissions — provides for two majority black districts (the 6th and the ,12th). Using the combo method, the Enacted Plan provides for three majority black districts in Suffolk County (the 5th, 6th, and 12th) whereas Plan No. 2 provides for the same three majority black districts plus a fourth (the 11th). . We enumerate the specifics needed to support these conclusions in Appendix A (attached hereto and incorporated by reference herein). We note, moreover, that the plaintiffs would be equally successful in proving the first Gingles precondition if we were to use CVAP data instead of VAP data. See Appendix B (attached hereto and incorporated by reference herein). To cinch matters, the districts illumined in Plan No. 2 are compact, meet applicable standards of contiguity, and comply fully with the one-person, one-vote requirement. The defendants’ counter-arguments are unpersuasive. First, they insist that the proper benchmark for the first Gingles precondition is the plan initially offered by the plaintiffs (the so-called Harmon Plan), not Plan No. 2. In this respect, the defendants lament that they first became aware of Plan No. 2 after the close of discovery. They add that they received critical documents relating to this new plan only two weeks before trial began. Their argument is unconvincing. In the first place, the defendants have shown no unfair surprise; Plan No. 2 is simply a new iteration of data long in the defendants’ possession. Perhaps more importantly, in requiring plaintiffs “to demonstrate that [the minority group] is sufficiently large and geographically compact to constitute a majority in a single-member district,” Gingles, 478 U.S. at 50, 106 S.Ct. 2752, the Court focused on the potential of minority voters to elect representatives of their choice in some hypothetical district, see Vecinos, 72 F.3d at 979. While plaintiffs must offer an alternative plan, they may do so, with leave of the court, at any time up to the end of trial. Cf. Barnett, 141 F.3d at 702 (“The plaintiff is not required to propose an alternative map that is ‘final’ in the ‘final offer’ arbitration sense, where the parties cannot modify their offers once they have denominated them final_”). Had the defendants so desired, they too could have processed the census data and discovered, early in the redistricting effort, the possibility of creating an additional majority black district. The defendants’ second counter-argument attacks Plan No. 2 frontally. This plan, the defendants aver, fails to show that African-American voting opportunities would be “significantly diminished” if the Enacted Plan were to remain in effect. Since two of the majority black districts configured by Plan No. 2 fall just over the 50% mark (at least when one uses VAP data and the combo method) and one of the Enacted Plan’s districts falls slightly under that mark (42.59%, using the same data and method), they asseverate that Plan No. 2 does precious little to enhance African-American voting opportunities in Suffolk County. We reject this reasoning. Even were we prepared to indulge the defendants’ ipse dixit that 42.59% is “only somewhat under the line,” the defendants’ argument overlooks that neither the Supreme Court nor the First Circuit has held that the first Gingles precondition is concerned with districts that fall under the 50% majority minority cutoff point. Moreover, the Gin-gles line of cases only requires plaintiffs to show the feasibility of one additional majority minority district. See Vecinos, 72 F.3d at 985-86. The plaintiffs have cleared that hurdle here. We find, therefore, that the plaintiffs have carried the devoir of persuasion with regard to the first prong of Gingles. B. Cohesiveness and Bloc Voting. The second and third Gingles preconditions demand slightly different analyses of substantially similar data. For that reason, we think it convenient to discuss them together. We remain aware, however, that each of the two inquiries serves a distinct purpose. The second Gingles precondition focuses on cohesiveness. It measures the potential of minority voters to elect representatives of their choice were they presented with an ideally configured district. The premise is that unless a minority group is politically cohesive, it will not have the voting strength necessary to achieve electoral success, notwithstanding favorable numbers. See Growe, 507 U.S. at 40-41, 113 S.Ct. 1075; Gingles, 478 U.S. at 51, 106 S.Ct. 2752. The third Gingles precondition focuses on majoritarian bloc voting. It delves into the cause of minority voters’ lack of success in existing districts. The premise is that if a majority group votes sufficiently as a bloc to enable it to impose its will and defeat minority-preferred candidates most of the time, the challenged practice, procedure, or structure is likely to be a cause of the minority group’s impotence at the polls. See Growe, 507 U.S. at 40-41, 113 S.Ct. 1075; Gingles, 478 U.S. at 51, 106 S.Ct. 2752. To satisfy the second and third preconditions, plaintiffs typically offer statistical evidence detailing the voting patterns of racial groups in past elections. True to this profile, the plaintiffs’ expert, Dr. Richard Engstrom, presented the results of his analysis of four Boston City Council elections, one city-wide district attorney election, and two Democratic primaries for House seats. The defendants countered with the testimony of Dr. Harold Stanley, who undertook to analyze the results of several House races (involving a mix of primary and general election match-ups) and one city-wide district attorney contest. We found both experts knowledgeable and both used accepted analytic tools, such as regression and homogeneous precinct analyses. See Gingles, 478 U.S. at 52-53, 106 S.Ct. 2752 (endorsing both of these analytic approaches). Dr. Engstrom also employed a third methodology, known as ecological inference—but we have chosen to focus our attention on the two more conventional types of analysis. Under the best of circumstances, combing through columns of statistical data is a daunting task for those of us untrained in the finer points of an admittedly arcane science. Maintaining a focus on the inquiries posed by the second and third Gingles preconditions makes the task more manageable. Refined to bare essence, the second precondition (political cohesion) asks whether the electoral data exhibits clear voting preferences on the, part of the minority group (here, African-Americans). Id. at 56, 106 S.Ct. 2752 (“A showing that a significant number of minority group members usually vote for the same candidates is one way of proving the political cohesiveness necessary to a vote dilution claim .... ”). The third precondition (white bloc voting) requires a four-step inquiry: a court must (i) identify the candidates most preferred by the minority group; (ii) observe whether the white majority votes as a bloc for other candidates; (iii) determine whether the white bloc vote is of a magnitude that usually suffices to defeat minority-preferred candidates; and (iv) assess whether any of the electoral results should be discounted because of special circumstances. Id. at 56-57, 106 S.Ct. 2752; cf. Jenkins v. Manning, 116 F.3d 685, 691 (3d Cir.1997) (outlining a similar inquiry but dividing it into three, rather than four, parts). Although the bulk of the analysis necessarily functions at the level of individual elections, courts must be careful not to become preoccupied with the trees and thereby lose sight of the forest. The ultimate answer to the question of whether racially polarized voting exists to a significant degree will crystallize only after the court steps back to view the landscape as a whole. Although the number of elections that must be studied will vary from case to case, the Supreme Court has cautioned that a pattern of polarized voting extending over a period of time is customarily more probative than the results of any single election. Gingles, 478 U.S. at 57 & n. 25, 106 S.Ct. 2752; accord Vecinos, 72 F.3d at 984 (explaining the need to examine voting practices over time because “[i]n this enlightened day and age, bigots rarely advertise an intention to engage in race-conscious politics”). Above all, an inquiring court should resist the temptation to confine itself to raw numbers in a particular election, instead endeavoring to make a practical, commonsense appraisal of all the evidence. Vecinos, 72 F.3d at 989. With these precepts in mind, we begin our evaluation of the data. 1. The Multi-Race Endogenous Elections. After having conducted an exhaustive review of the case law and the circumstances of this litigation, we conclude that the most probative elections for our purposes are likely to be multi-race endogenous elections. By multi-race, we mean those elections in which at least one white candidate vied for office against at least one African-American candidate. By endogenous, we mean House elections in the seventeen affected Suffolk County districts. Our decision to accord more weight to multi-race elections is supported not only by common sense but also by the case law. See Rural West Tenn. African American-Affairs Council v. Sundquist, 209 F.3d 835, 840-41 (6th Cir.2000) (approving a decision to accord greater weight to results of black-versus-white elections); Jenkins, 116 F.3d at 692, 694-95 (affirming a decision to discount same-race elections); Westwego Citizens for Better Gov’t v. City of Westwego, 872 F.2d 1201, 1208 n. 7 (5th Cir.1989) (“[T]he evidence most probative of racially polarized voting must be drawn from elections including both black and white candidates.”); see generally Vecinos, 72 F.3d at 988 n. 8 (“[EJlections in which minority candidates run are often especially probative on the issue of racial bloc voting.”). We understand that black voters sometimes may consider a white candidate their representative of choice and vice-versa. If no candidate of the voter’s race is in the field, however, that support may well represent something less than a true preference. Cf. Gingles, 478 U.S. at 68, 106 S.Ct. 2752 (plurality op.) (stating, as a fact, “that race of voter and race of candidate is often correlated”). Indeed, the choice presented to minority voters in an election contested only by two white candidates is somewhat akin to offering ice cream to the public in any flavor, as long as it is pistachio. The VRA focuses on the opportunity of minority voters to elect representatives of their choice, and we believe that this opportunity is best and most easily measured in elections that offer black voters the chance to support a viable black candidate against a viable white candidate. See Sanchez v. Colorado, 97 F.3d 1303, 1317 n. 24 (10th Cir.1996). At the same time, we recognize the obvious: in most instances, the best indicator of how voting operates in a particular type of election is how voting historically has operated in that type of election. See Rural West Tenn., 209 F.3d at 841; Johnson v. Hamrick, 196 F.3d 1216, 1222 (11th Cir.1999); see also Vecinos, 72 F.3d at 990 (suggesting that, in general, endogenous elections are more probative than exogenous elections). Consequently, we focus initially on multi-race endogenous elections. We commence our canvass of this electoral subset with the 1994 Democratic primary election in what was then the 5th Suffolk House district. The evidence shows that black voters preferred the black candidate (Golar-Richie) while white voters preferred the white candidate (Roman). Golar-Richie prevailed. On these facts, it appears that racially polarized voting existed, but that white bloc voting lacked the muscle to defeat the black-preferred candidate. In a typical district, these results would supply evidence tending to establish the second Gingles precondition and to defeat the third. But the 5th Suffolk, as it existed in 1994, was not a typical district. Rather, it was a majority black district in which African-Americans comprised 50.87% of the VAP (based on 1990 census figures). This constitutes a special circumstance that robs the election of any probative force with respect to the third Gingles precondition. After all, it should come as no surprise that a black-preferred candidate was successful in • a majority black district with racially polarized voting, which is why Gingles’s third prong directs the court’s attention to those districts in which there is a white majority. See Gingles, 478 U.S. at 51, 106 S.Ct. 2752; see also Growe, 507 U.S. at 40, 113 S.Ct. 1075 (explaining that a showing of white bloc voting is “needed to establish that the challenged districting thwarts a distinctive minority vote by submerging it in a larger white voting population”); cf. De Grandy, 512 U.S. at 1003-04, 114 S.Ct. 2647 (indicating that standard bloc-voting analysis must give way “in a district where a given minority makes up a voting majority”). In sum, we believe that the results of this election support the plaintiffs’ allegation that African-Americans in Boston are politically cohesive without in any way contradicting their allegation that white bloc voting usually serves to defeat black-preferred candidates in majority white districts. See Old Person v. Cooney, 230 F.3d 1113, 1122 (9th Cir.2000) (concluding that a district court erred in drawing “no distinction between jurisdictions in which Indian voters constitute a majority of the voting age population, and those jurisdictions in which white voters were in the majority” while analyzing the third Gin-gles precondition). We turn next to the 1996 Democratic primary election in the same district. This election was largely a reprise of the 1994 primary. African-American voters preferred Golar-Richie, white voters preferred Roman, and Golar-Richie won. For the reasons explicated above, we conclude that these results constitute evidence of political cohesiveness among African-American voters, but do not speak to the existence vel non of legally significant white bloc voting. The 2002 general election for the House seat in the 5th Suffolk district pitted a black candidate (St.Fleur) against a white candidate (Chaparro). In this election, both black and white voters favored St. Fleur (who ran on the Democratic ticket) over Chaparro (who ran as an independent). Here, too, we are satisfied that special circumstances explain the success of the black-preferred candidate (and, therefore, that the results of this election have scant probative value). The reasoning that led us to discount the prior two elections does not apply to this contest. The election occurred in 2002 and the Enacted Plan had revised the district lines. As such, the district had 49.73% black YAP on the alone method and a 55.93% black YAP on the combo method. Although these estimates leave us uncertain whether the district fairly can be characterized as a majority black district, the political climate in Boston — an area in which the vast majority of citizens vote Democrat — makes general election results unreliable barometers of the second and third Gingles preconditions. Based on election statistics compiled by the Elections Division of the Office of the Massachusetts Secretary of State, we take judicial notice, Fed.R.Evid. 201, of the fact that no Republican, independent, or third-party candidate has been elected to the House from a representative district wholly or partly within Boston since 1992. Prior to the election of Representative Althea Garrison in that year, none had been elected since 1970. Because political allegiance can be a tie that binds every bit as much as race, a determination that black and white voters in Boston preferred the Democratic candidate at a general election is hardly news. In our estimation, that fact says less about race than about partisan politics. The relationship between race and party politics is a complicated matter and the removal of partisanship from the equation helps to isolate race for purposes of a vote dilution inquiry. See Sanchez, 97 F.3d at 1317 n. 25 (noting that analysis of Democratic primary elections helps counter the argument that, in a heavily Republican district, “Hispanics don’t lose elections, Democrats do”). For this reason, we deem the results of this election unillumi-nating. This brings us to the 1998 House primary in the 9th Suffolk district. In this race, two white candidates (Branson and Walsh) ran against an African-American candidate (Rushing). Rushing proved to be the preferred candidate of both black and white voters. The rub, however, is that by the time of the 1998 primary, Rushing was an eight-term incumbent who had served in the House for sixteen years. Incumbency is a special circumstance that must be weighed, sometimes heavily, in assaying the probative value of election results. See Gingles, 478 U.S. at 60, 106 S.Ct. 2752; Old Person v. Brown, 312 F.3d 1036, 1048 n. 13 (9th Cir.2002), cert. denied, — U.S. -, 124 S.Ct. 566, 157 L.Ed.2d 429 (2003). Consequently, we decline the defendants’ invitation to treat this election as disproving the plaintiffs’ allegation that legally significant white bloc voting exists in Boston. The 2000 general election in the same House district is no more helpful. Both black and white voters preferred the black incumbent, Rushing, who won over his white opponent (Ashcroft) in a landslide. We discount this result both because of Rushing’s long-time incumbency and because he had the Democratic nomination (Ashcroft ran as a Libertarian). In 2002, there were three candidates in the Democratic primary in the 11th Suffolk House district. Two were white (Malia and McLaughlin) and one was black (Payne-Thompson). The statistical evidence satisfies us that Malia, who won the election, was the preferred candidate of white voters. The record is tenebrous, however, as to which candidate African-American voters preferred because the best available regression analysis produced a negative roll-on. A roll-on is the percentage of a particular group’s voters estimated to have cast a ballot in the contest. A negative roll-on shows that voters in the studied subgroup chose, in large numbers, not to vote for any candidate in the election in question. Thus, a negative roll-on is a powerful indicator that no candidate in the race was that subgroup’s candidate of choice. Given the negative roll-on for African-Americans here, we conclude that this factor renders unreliable the estimates of how black votes were cast. In 1990, two white candidates (Finneran and Bennett) ran against a black candidate (Johnson) in the Democratic primary for the House seat in the 12th Suffolk district. Both white and black voters preferred Fin-neran — a legendary political powerhouse who became Speaker of the House of Representatives a few years later. Although the election results tend to argue against racial polarization, they constitute rather weak proof in the circumstances of this case. The final election in this grouping is the 2002 general election in the 15th Suffolk House district. This three-way race involved an Hispanic candidate (Sanchez— who had the Democratic endorsement), a black candidate (Chidi, who ran as an independent), and a white candidate (Clifford, who also ran as an independent). Although white voters preferred Sanchez, we cannot draw any reliable conclusions from these results because regression analysis produced a negative roll-on estimate for the African-American vote. This concludes our survey of the results of the multi-race endogenous elections. For our purposes, the 1994 and 1996 primary elections in the former 5th Suffolk House district tend to prove that African-Americans vote as a politically cohesive group. Apart from that, the results are largely uninformative. Accordingly, we find it necessary to turn to multi-race exogenous elections as the next best available source of evidence to help us discern whether a pattern of racially polarized voting exists in Boston. 2. Multi-Race Exogenous Elections. In the absence of a sufficient number of useful multi-race endogenous elections, the next most fertile field is composed of mul-ti-race exogenous elections. By this, we mean those elections in which at least one black candidate competed against at least one white candidate for an elected office (other than a House seat) in the general geographic area involved in the plaintiffs’ vote dilution challenge. Courts, as a matter of discretion, have the authority to consider exogenous elections when faced with a paucity of meaningful endogenous elections. See, e.g., NAACP v. Fordice, 252 F.3d 361, 370-71 (5th Cir.2001); Askew v. City of Rome, 127 F.3d 1355, 1381 n. 13 (11th Cir.1997) (per curiam); Cano v. Davis, 211 F.Supp.2d 1208, 1235 (C.D.Cal.2002) (three-judge court) (per curiam). They must, however, furnish explanations for why particular exogenous elections reveal reliable information about relevant voting patterns. Vecinos, 72 F.3d at 990. The affected area here is Suffolk County (and, in particular, Boston). The plaintiffs urge us to consider several exogenous elections as evidence of African-American political cohesion and white bloc voting. Specifically, they point to four at-large Boston City Council elections and one race for Suffolk County district attorney. The defendants argue strenuously against consideration of these exogenous elections. They contend that the different circumstances surrounding the elections render the retrieved data of dubious value in a vote dilution case designed to test the composition of the House of Representatives. This argument has a patina of plausibility. The dynamics of the proffered exogenous elections are somewhat different than those of House elections. For instance, Boston City Council elections are non-partisan, certain candidates run city-wide for four at-large seats instead of district by district, and each voter may cast up to four votes for those seats. The Suffolk County district attorney election also differs from a prototypical House election. Although that election is partisan and involves only a single vote for a single office, the political subdivision is slightly different; whereas the plaintiffs’ vote dilution claim concentrates on those House districts that lie at least partly within the city of Boston, the district attorney election involves votes from all over Suffolk County (a region that includes Boston, Chelsea, Revere, and Winthrop, and, more to the point, includes two Suffolk County House districts that are not within the scope of the plaintiffs’ challenge). The district attorney election is also different in that it does not include votes from Milton (although Milton is in Norfolk County, the Committee tacked several precincts from that community onto Speaker Finneran’s 12th Suffolk district). And, finally, all of the exogenous elections that the plaintiffs urge us to consider may differ from House elections in terms of more amorphous factors (e.g., the importance that voters assign to the office or the campaign strategy that a candidate is likely to employ). In the end, however, we think that the similarities between the proffered exogenous elections and the elections at issue in this case are sufficiently compelling to outweigh the differences. Both sets of elections are for offices with comparable levels of importance within the community. They concern much the same constituencies, and we believe that there is a considerable measure of truth in Dr. Engstrom’s observation that racially polarized voting rarely stops at electoral borders. Thus, we are confident that an in-depth eonsider-ation of the multi-race exogenous elections may allow us to glean meaningful insights into the voting patterns and preferences of Boston-area residents. We turn, then, to those elections. As we do so, we note that although the defendants have scoffed at the probative value of these elections, they have not seriously questioned the accuracy or integrity of the plaintiffs’ regression analysis. We begin with the 1995 councilmanic election. There, a black candidate (Jones) and seven white candidates competed for four at-large seats. Jones received over 81% of the black vote, and the correlation coefficients indicate political cohesion among African-American voters. However, Jones received only 20% of the non-minority vote and he failed to win a seat. In our judgment, the results of this election — Jones ranked first among black voters but last among white voters — evince both a pattern of cohesive black voting and a pattern of majoritarian bloc voting sufficiently strong to defeat the black-preferred candidate. The results of the 1997 councilmanic election reinforce these findings. There, a lone black candidate (Jones) faced seven white candidates in a race for four available seats. Jones received 88.5% of the black vote and was the most prolific vote-getter in that segment of the community. Not surprisingly, the correlation coefficients once again indicate political cohesion among African-American voters. Jones nonetheless failed to finish in the top four (and, hence, lost the election), receiving only 27.8% of the non-minority vote (ranking as the sixth most popular candidate among non-minority voters). We find that this election provides additional evidence of both cohesive African-American voting and majoritarian bloc voting powerful enough to defeat a black-preferred candidate. The results of the 2002 election for Suffolk County district attorney militate in the plaintiffs’ favor. In this contest, a black candidate (Jenkins) ran against a white candidate (Conley). Jenkins received 99.6% of all African-American votes and the correlation coefficients indicate a very high degree of political cohesion among African-American voters. Jenkins nonetheless failed to win election because he received only 22.1% of the non-African-American vote. The results of this election exhibit both African-American cohesiveness and white bloc voting sufficiently staunch to defeat the black-preferred candidate. We give diminished weight-to the indication of legally significant white bloc voting, however, as Conley ran as an incumbent (he had been appointed to fill a vacancy) and carried the Democratic party’s endorsement. As we already have explained, see supra Part 11(B)(1), the Democratic imprimatur has been a salient factor in general elections held within Suffolk County. Incumbency, too, usually helps to propel a candidacy. We also have examined the results of two multi-race exogenous elections touted by the defendants: the 2003 councilmanic election and the 1994 election for Suffolk County district attorney. The first of these witnessed two black candidates, an Hispanic candidate, and five white candidates vying for four at-large city council seats. The Hispanic candidate (Arroyo) was elected with 76.1% of the black vote and 38.4% of the non-minority vote, ranking him first among black voters and fifth among non-minority voters. The correlation coefficients for this election indicate the presence of political cohesion among black voters. The white vote, however, was sufficiently scattered that it did not operate to defeat the black-preferred candidate. Therefore, the election hints at the absence of legally significant white bloc voting (and, to that extent, cuts in the defendants’ favor). The 1994 Suffolk County district attorney election matched a white candidate (Malone) against a black candidate (Martin) for the single seat. Martin ran on the Republican ticket, but he ran as an incumbent (originally having been appointed to the position by a Republican governor). Despite the fact that Malone had the Democratic endorsement, Martin was the preferred candidate of both black and white voters. He won the election handily. This election was plainly dominated by an attractive, well-known incumbent whose appeal cut across racial lines. Consequently, we find the results of this election unin-structive for purposes of either the second or third Gingles precondition. Taken in the ensemble, the results of these multi-race exogenous elections provide compelling evidence that,' in the Boston area, African-Americans vote cohesively, yet the white majority votes sufficiently as a bloc to enable it; most of the time, to defeat black-preferred candidates. 3. Single-Race Endogenous Elections. Although we have concluded that, in general, the results of single-race endogenous elections are less probative than other available data, we believe that those results ought to be mined in hopes that they may yield nuggets of useful information about voting patterns in the Boston area. Cf. Vecinos, 72 F.3d at 988 n. 8 (suggesting that “evidence exhumed from ‘white only’ elections may still be relevant in assessing the totality of the circumstances in a vote dilution case” prosecuted by Hispanics). Of the nineteen elections in this category — which include not only white-versus-white and black-versus-black elections but also two elections that pitted white candidates against Hispanic candidates — only fourteen produced statistically valid estimates. Ten of these were general elections (and, thus, of little utility to us, for reasons previously explicated). Three of the remaining four (the 1998 and 1999 Democratic primaries in the 5th Suffolk House district and the 2000 Democratic primary in the 6th Suffolk House district) took place in majority black districts and, thus, cannot tell us very much about the third Gingles precondition. This winnowing process leaves only the 1994 Democratic primary in the 15th Suffolk House district. That election pitted two white candidates against one another. African-American voters preferred Broo-kins (who lost) and white voters preferred Fitzgerald (who prevailed). These tallies indicate the presence of both African-American political cohesion and legally significant white bloc voting. 4. Single-Race Exogenous Elections. Finally, we have examined the results of one single-race exogenous election: the 2001 election for Boston City Council. In that contest, an Hispanic candidate (Arroyo) faced six white candidates for four at-large city council seats. Although Arroyo received 52.7% of the black vote and 65.6% of the Hispanic vote, he received only 19% of the non-minority vote. This placed him first among African-American and Hispanic voters, but only sixth among other voters. Arroyo lost the election. The regression results indicate the presence of both cohesive African-American voting and white bloc voting staunch enough to defeat a black-preferred candidate. Despite these indications, however, we accord relatively little weight to this evidence because the contest was both single-race and exogenous. 5. Recapitulation. The evidence rehearsed above, when evaluated in connection with the other evidence in this case, leads us to find that the plaintiffs have clearly and convincingly satisfied the second Gingles precondition. The evidence establishes beyond peradventure that African-American voters in the Boston area are politically cohesive. The same evidence, taken as a whole, also suffices to show that white voters, who constitute a majority in most districts, vote sufficiently as a bloc to enable them, as a general rule, to defeat the black-preferred candidates. We find, therefore, that the plaintiffs also have satisfied the third Gingles precondition. C. Totality of the Circumstances. Although satisfaction of the Gingles preconditions takes the plaintiffs a giant step closer to success on the merits, it does not take them to the finish line. We still must canvass the facts needed to bring the totality of the circumstances into proper perspective. See Gingles, 478 U.S. at 79, 106 S.Ct. 2752 (instructing trial courts “to consider the totality of the circumstances and to determine, based upon a searching practical evaluation of the past and present reality whether the political process is equally open to minority voters”) (citation and internal quotation marks omitted). It will, however, be the rare case in which plaintiffs meet the Gingles preconditions and yet fail on their section 2 claim due to the totality of the circumstances. Vecinos, 72 F.3d at 983; Jenkins v. Red Clay Consol. Sch. Dist. Bd. of Educ., 4 F.3d 1103, 1135 (3d Cir.1993). As would be expected, the relevant circumstances are those bearing upon whether African-American voters in the affected districts have the same opportunities as other voters to participate in the political process and elect representatives of their choice. Our canvass includes the factors identified in the Senate report, see supra, to the extent they are relevant here (a few are not). It also includes several environmental factors peculiar to this case. One of the most revealing questions a court can ask in assessing the totality of the circumstances is whether the affected districts exhibit proportionality, that is, whether the number of majority minority districts is in proportion to the minority group’s share of the relevant population. This is to be distinguished from proportional representation, which speaks to the number of minorities elected to office. Whereas proportionality (or the lack thereof) is consistent with the calculus that section 2 demands, “nothing in [section 2] establishes a right to have members of a protected class elected in numbers equal to their proportion in the population.” 42 U.S.C. § 1973(b). While a finding of proportionality does not create a safe harbor, it often may be an important indicium that minority voters have an equal opportunity, regardless of racial polarization, to participate in the political process and elect representatives of their choice. See De Grandy, 512 U.S. at 1017-2