Over the past fifty years, the steady march of civil rights has been to New Orleans and this court. It continues but the demands have changed. Relatively clear lines of legality and morality have become more difficult to locate as demands for outcomes have followed the cutting away of obstacles to full participation. With our diverse ethnic makeup, this demand for results in voting has surfaced profound questions of a democratic political order such as the limits on rearranging state structures to alter election outcomes, and majority rule at the ballot box and even in legislative halls, questions Congress has provoked but not answered. Ml this can make a simple voting rights case seem difficult, certainly so with state judges elected on a partisan ballot. Today our difficulties of fitting the Act to the unique features of the state judiciary and sorting out racial and partisan voting are large but the merits of the claims are easily grasped. As we will explain, there is a background to the debate on the large issues that must not be obscured. The evidence of any dilution of minority voting power is marginal at best. We are not persuaded that a violation of the Voting Rights Act has been proved and we reverse.
I. Facts
On July 11, 1988, ten individual voters and the League of United Latin American Citizens sued in federal district court alleging that Texas’ system of electing state trial judges violated § 2 of the Voting Rights Act and the Fourteenth and Fifteenth Amendments in several Texas counties.1 They sued the Governor of Texas,2 the Attorney General, the Secretary of State, and the Chief Justice of the Supreme Court as chair of the Judicial Districts Board. Because this board is responsible for reapportioning the judicial districts, the suit also named each of its members as defendants. On March 12,1989, the district court granted the motions to intervene of the Houston Lawyers’ Association, the Legislative Black Caucus, and two Texas district court judges, in their individual capacities — Sharolyn Wood, 127th District Court in Harris County, and Harold Entz, 194th District Court in Dallas County.
As they have throughout Texas history, Texas voters elect their trial judges in county-wide elections. A voter may vote for all of the trial courts of general jurisdiction in her county. At the same time, each trial court is a distinct court, such as the 134th judicial district court of Dallas County, with countywide jurisdiction and its own history of incumbents. A candidate runs for a particular *838court. Plaintiffs contend that electing trial judges county-wide violates § 2 of the Voting Rights Act by impermissibly diluting the voting power of Hispanics and blacks. Plaintiffs proceed on behalf of language and ethnic minorities in different combinations in different counties. Depending on the county— more specifically, the numbers — they argue that Hispanic voters, black voters, or the combination of both Hispanic and black voters “have less opportunity than other members of the electorate to participate in the political process and elect representatives of their choice.” Plaintiffs aimed their constitutional challenge at Article 5, § 7a(i) of the Texas Constitution, which precludes the creation of judicial districts smaller than a county absent approval by a majority of the voters in that county. They argued that this limitation on the power to redistrict of the Judicial Districts Board, chaired by defendant Chief Justice Phillips, was enacted with discriminatory intent.
On November 8, 1989, the district court found county-wide elections violated § 2 in all nine counties, enjoined future elections, divided the nine counties into electoral sub-districts, and ordered a nonpartisan election for May 5, 1990, with any runoff to be held on June 2. The district court rejected the constitutional arguments, finding that plaintiffs had failed to prove that Texas instituted or maintained the electoral system with discriminatory intent.3 Intervenors Judge Wood and Judge Entz appealed. Unhappy with nonpartisan elections ordered by the district court, the Texas Attorney General first moved the court to alter its interim plan. After the court denied the motion, the Attorney General filed a notice of appeal.4 We stayed the district court’s order pending appeal.
. In our first effort in this ease, a panel held that the Act covers judicial elections but concluded that electing district judges in county-wide elections in Texas did not violate § 2. League of United Latin American Citizens v. Clements, 902 F.2d 293 (5th Cir.1990) (“LULAC /”). We considered the history of judicial elections in Texas and the office of district judge — the court of general jurisdiction. We held that Texas had a special interest in linking the jurisdictional and electoral bases of the trial courts, an interest accented by unwavering support throughout Texas history. Finding no truly informing analogues for resolving such an attack on at-large voting supported by a state interest unique to this judicial office, we looked to the weighing constructs familiar to the Act. We concluded that, as a matter of law, the state interest linking jurisdiction and electoral base outweighed its potentially dilutive effect. LULAC I, 902 F.2d at 308.
A majority of this court sua sponte ordered reconsideration of the panel decision en banc. League of United Latin American Citizens v. Clements, 914 F.2d 620 (5th Cir.1990) (“LULAC IF). The en banc court held by a 7-6 vote that § 2 of the Act did not apply to judicial elections, rejecting the contrary view of the panel.
Houston Lawyers’ Association, as interve-nor, and LULAC petitioned for certiorari. The Supreme Court granted both petitions, consolidated them, and reversed, holding that the Voting Rights Act applies to state judicial elections. Houston Lawyers’ Ass’n v. Attorney General, — U.S.-, 111 S.Ct. 2376, 115 L.Ed.2d 379 (1991). The Supreme Court also held that Texas has a special interest in linking the electoral and jurisdictional bases of district judges. Id. at -, 111 S.Ct. at 2381. The Court did not agree, however, that this state interest outweighed its dilutive effect in all cases, as a matter of law. Rather, the Court held that balancing is a case-specific enterprise, struck by inquiry into the totality of the circumstances. Justice Stevens explained that the state interest in linkage was to be weighed in deciding “whether a § 2 violation occurred.” Id. Justice Stevens made plain that assessing the linkage interest is part of the determination of liabili*839ty and not remedy alone. The Court effectively, came down between the “goes only to remedy” view of the Department of Justice and the “matter of law” view of the concurring opinion in LULAC II.
On remand, the en banc court in turn remanded to the panel. On January 27, 1993, a majority of the panel affirmed the district court’s findings in eight of the nine counties. The panel concluded that plaintiffs failed only in Travis County, a Democratic stronghold. League of United Latin American Citizens v. Clements, 986 F.2d 728 (5th Cir.1993) (LULAC III). For a second time, this court decided, on its own motion, to hear the case en banc.
Although the panel opinion had been vacated, General Morales urged a legislative solution to reforming judicial elections. He submitted a plan to the legislature calling for the election of judges from single-member districts in all Texas counties with populations over 100,000. Recognizing that the Texas Constitution mandates the current system of electing trial judges, see Tex. Const. Art. 5, §§ 7, 7a(i), Morales asked the legislature to submit a constitutional amendment to the voters to implement his plan and urged them to do so in time to moot the LULAC lawsuit. Doubting the necessary legislative support for an amendment, the Governor, the Lieutenant Governor, and minority lawmakers urged Morales to achieve the same result through settlement. Morales drafted an agreement providing for the election of the vast majority of judges in the nine urban counties by subdistricts. Democratic officials who were parties to the suit quickly agreed. But Morales could not obtain the agreement of Chief Justice Phillips, nor the district judges, Judge Wood and Judge Entz.
When a proposed resolution approving the “agreement” reached the floor of the Senate there was no quorum because all but two of the thirteen Republican senators walked out. The Senate later reconvened as a Committee of the Whole, not in formal session, and voting along party lines, adopted a resolution expressing its “sentiment” in support of a federal decree. Voting in the House also followed party lines. Nothing with the force of law could be obtained from the legislature. When the dust settled, the only legislative action was this expression of sentiment in support of a federal decree, and that from a Senate convened in a Committee of the Whole. Failing to obtain any positive enactment from the legislature, Morales requested that we remand to the district court for a hearing and entry of his proposed “consent” decree.
By the decree, 152 judges would run in districts smaller than a county, while 22 would continue to be elected at-large. District boundaries would mirror state representative districts in Dallas, Harris, Bexar, and Jefferson counties. Justice of the peace districts would be used in Tarrant County. In Lubbock, Ector, and Midland counties, judges would run from the existing commissioners court districts. Anticipating the question of how the case can be settled without the agreement of the district court judges, the plan allows Judges Wood and Entz to be elected in a county-wide election. The stated purpose was to deny the defendant district judges standing to object.
Chief Justice Phillips, Judge Wood, and Judge Entz object to the proposed decree and oppose the motion to remand. In addition, three former Chief Justices of Texas, Joe R. Greenhill, Robert W. Calvert, and John L. Hill, are before us as amici objecting to remand — and denying the authority of the Attorney General to bind the State. Judges Wood and Entz have moved to realign General Morales with the plaintiffs, and allow their assumption of the defense of the current system.5 Judge Wood has also moved to disqualify the Texas Attorney General as counsel for the State. When settlement negotiations began, Chief Justice Phillips obtained independent counsel.6 General Morales responded by moving to disqualify Phillips’ counsel. Finally, immediately after oral argument, plaintiffs filed a notice of *840nonsuit of Chief Justice Phillips and the Texas Judicial Districts Board.
II. Motion to Remand
We are asked to remand to the district court for entry of a consent decree, although some of the parties wish to proceed with the appeal. The Attorney General argues that these non-consenting parties are no obstacle. Chief Justice Phillips, General Morales argues, was sued in his official capacity as chair of the Judicial Districts Board and the Attorney General is the exclusive lawyer for the State of Texas. On its face, this is not a remarkable contention. However, General Morales also maintains that in his role as lawyer for the State, he need not represent the State’s policymakers; he can ignore them and impose his own views. That is remarkable. The force of this contention is that the Attorney General is the sole arbiter of State policy when the State’s interest is in litigation. This argument is put forward despite the fact that it leaves his scrambling for legislative support wholly inexplicable; under his presently claimed power, the Attorney General did not need to have the “settlement” adopted by statute. In any event, Texas law does not sanction his actions. Nor are we persuaded that Defendant-Interve-nors, Judges Entz and Wood, lack standing to object to a proposed consent decree that will allow them to run county-wide. We deny the motion to remand.
A. The Authority of the Texas Attorney General
General Morales is not the first Texas Attorney General to have staked such a claim of authority. We rejected a similar effort in Baker v. Wade, 769 F.2d 289 (5th Cir.1985) (en banc). Baker challenged Texas’ anti-sodomy statute, suing Holt, the Dallas City Attorney, and Wade, the Dallas County District Attorney. The district court certified a defendant class of officials responsible for enforcing the statute, with Holt and Wade as representatives, and the Attorney General of Texas intervened on behalf of the State. After the district court declared the statute unconstitutional, Danny E. Hill, Potter County’s district attorney, filed a notice of appeal, concerned that the Attorney General might decide not to appeal. Hill was a member of the class, but was not a named defendant and had not sought to intervene. Hill’s concern was realized when the Attorney General appealed but then withdrew the notice. After failing to persuade the Texas Supreme Court to order the Attorney General to pursue the appeal and unable to obtain leave to intervene from the district court, Hill asked this court for leave to intervene on appeal. We granted this request, explaining:
[Hill] would be seriously prejudiced were he not allowed to intervene, whereas allowing the appeal to proceed would prejudice no one. As a state official empowered by Texas law to enforce criminal laws, his interest and its impairment by the district court’s judgment cannot be questioned.... In this case where the district court has rejected binding Supreme Court authority, the circuit court is entitled to conclude as a matter of law that those interests were inadequately represented by those who failed to pursue the appeal and that the state officer seeking to intervene was a proper party to do so.
Id. at 292.
Attorney General Mattox made a considered decision to accept the district court’s declaration of unconstitutionality. That was a basic policy choice. Baker’s relevant instruction lies in the fact that Attorney General Mattox’s decision did not control. Baker rejected the very power claimed by this Attorney General. The power he would exercise cannot be squared with Baker.
That Attorney General Mattox decided to accept the ruling of the district court and Morales reaches for a similar result by a “settlement” fails to distinguish our holding in Baker. It does not respond to our holding that the Attorney General cannot bind state officials, his clients, to his own policy preferences. It is asserted that Hill as a district attorney, one of hundreds in Texas, was charged with the duty of enforcing the statute held unconstitutional. The law enforcement responsibility of a district attorney and that of the Chief Justice as chair of the redistricting board, however, do not differ in relevant ways. Indeed, that the Chief Jus*841tice may defend the suit is an a fortiori case under Baker. After all, his judicial duties aside, the Chief Justice’s enforcement responsibilities under the redistricting provisions of state law are statewide. A district attorney’s duties, however, run only to the county line. See Crane v. Texas, 766 F.2d 193 (5th Cir.1985).
The Texas Constitution requires the Chief Justice to supervise the state district courts. Article 5, § 7a established the Judicial Districts Board and made the Chief Justice its chair. Tex. Const. Art. 5, § 7a(a) and (b). The constitution charges the Board with the duty of reapportioning the judicial districts as the need arises. Id. § 7a(f). Among other things, the Board is required to consider a district’s case load and population in its reapportionment decisions. Tex.Gov’t Code Atm. § 24.945 (Vernon 1988). Of special importance to this case, the Board may not create districts smaller than a county without a general election. Tex. Const. Art. 5, § 7a(i); Tex.Gov’t Code Ann. § 24.945(e) (Vernon 1988). A redistricting plan may not be proposed or adopted even in anticipation of such an election. Id. Indeed the district court denied leave to intervene in this suit to Midland County concluding it was not a real party in interest. A panel of this court agreed, observing that, unlike the Judicial Districts Board, the county lacked “the power to re-shape judicial districts.” LULAC v. Clements, 884 F.2d 185, 187 (5th Cir.1989). Given the Chief Justice’s role as chair of the Board and his state constitutional duties to manage state judicial districts and the efficiency of the courts, his contention that he has the authority to defend this lawsuit if the Attorney General will not is compelling. If a district attorney has a sufficient interest in protecting the laws he is duty-bound to enforce, we are persuaded that the Chief Justice as chairman of the Judicial Districts Board has a sufficient interest in protecting the current district court system.7
The concerns raised by the Baker dissent are not present here. The dissent was troubled by the fact that Hill was neither a named defendant nor a class representative, had never sought to intervene in the district court, and was not a named party when he filed his appeal. 769 F.2d at 294-95 (Rubin, J., dissenting). Here, Chief Justice Phillips has been a named defendant from the outset.
The state courts have had little occasion to face such a bold claim of authority. The few Texas cases that have grappled with the Attorney General’s authority offer him little comfort. Morales points to Terrazas v. Ramirez, 829 S.W.2d 712 (Tex.1991), but in Terrazas, General Morales also failed in an effort to “settle” a legislative reapportionment case. Following the 1990 census, plaintiffs sued various state and county officials to prevent the use of the new census in reapportioning the legislature, because it allegedly undercounted minorities. The legislature proceeded with reapportionment and plaintiffs also challenged the resulting plans. General Morales defended the legislature’s plans, lost at trial, and appealed directly to the Texas Supreme Court. Then, Morales agreed with the plaintiffs to settle the senate reapportionment challenge. The agreement included a redistricting plan that was submitted to the trial court and promptly accepted by it. Thereafter, five individuals, not parties to the suit, requested the Supreme Court of Texas to direct the trial court to vacate its judgments reconfiguring the senatorial districts, order the Attorney General to rescind the agreement, and direct the Secretary of State to withdraw submission of the plan for preclearance.
A plurality directed the trial court to vacate its judgments, but refused relief against the Attorney General. Four justices held that the trial court erred by failing to weigh all affected interests before entering the proposed decree. In Justice Hecht’s words, “a district court cannot order a reapportionment plan for the State based on nothing more than an agreement of the Governor, the Attorney General, and a few citizens.” Id. at *842714.8 Indeed a majority believed the Attorney General’s “discretion includes the authority to propose a settlement agreement in an action attacking the constitutionality of a reapportionment statute.” Id. at 722 (Hecht, J.) (emphasis supplied).
In approving of the Attorney General’s conduct, however, the plurality noted that he acted “on behalf of the state defendants[,]” giving him the authority “for his clients and even on his own, to suggest possible remedies ... [and] to negotiate a settlement.” Id. (Hecht, J.) (emphasis added). “To hold that he did not would be to give him less authority than any party or any other attorney participating in the case.” Id. (emphasis added). The Attorney General acts as counsel for state officials who are his clients.
Terrazas recognizes that the Attorney General represents officials. It does not follow that by doing so, the Attorney General steps into their shoes and assumes the poli-cymaking roles of those officials, against whom specific relief is sought. We need not and do not decide the authority of the Attorney General when an official is named in his official capacity only to join the State. Plaintiff sought specific relief against the Judicial Districts Board chaired by defendant Chief Justice Phillips. The petitioners who objected to the settlement in Terrazas were not even parties to the suit. The Attorney General’s power to settle for his clients is certainly no less than that of other lawyers, but Terrazas does not say that it is any greater. No lawyer may forge a settlement agreement over the express objection of his client. Here, to the extent that Morales represents the Chief Justice in the Justice’s defense of his constitutionally assigned task, he may not ignore him. As Justice Wallace put it for the Texas Supreme Court in Public Utility Commission of Texas v. Cofer, 754 S.W.2d 121, 125 (Tex.1988):
We emphasize that when a statute confers a right upon the attorney general to represent an agency, it imposes a corollary duty, and the agency has every right to expect the same diligent and faithful representation as any other “client.”
See also Hill v. Lower Colo. River Auth., 568 S.W.2d 473, 478 (Tex.Civ.App. — Austin 1978, writ ref d n.r.e.) (rejecting an attempt by the attorney general to sue the Texas Water Rights Commission “in an effort to substitute his views for that of a lawfully constituted State administrative agency”); Charles Scribner’s Sons v. Marrs, 114 Tex. 11, 262 S.W. 722, 729 (1924) (although attorney general had authority to represent the State Superintendent of Education, he did not have authority “to elect for the state to accept or reject a contract for text-books that is voidable,” a decision for the Board of Education).
The Texas legislature has also recognized that the Attorney General represents the State but does not make its policies. “An admission, agreement, or waiver made by the attorney general in an action or suit to which the state is a party does not prejudice the rights of the state.” Tex.Gov’t Code Ann. § 402.004 (Vernon 1988); see also State v. Reagan County Purchasing Co., 186 S.W.2d 128, 135 (Tex.Civ.App. — El Paso 1944, writ refd w.o.m.) (“acts beyond the scope of [Attorney General’s] delegated power are not binding on the State”). If the Texas Attorney General could make policy for the State, this provision would be superfluous, for he could never violate it. He would in effect be the State. When faced with this statute before, we appropriately noted that “Texas has been at particular pains to attempt to circumscribe the power of the attorney general to make admissions on its behalf.” United States v. Texas, 680 F.2d 356, 368 n. 17 (5th Cir.1982).9
*843Stated another way, the Attorney General’s right to represent state officials or state agencies cannot be gainsaid, see Hill v. Texas Water Quality Bd., 568 S.W.2d 738, 741 (Tex. Civ.App. — Austin 1978, writ ref'd n.r.e.); Morris v. Smiley, 378 S.W.2d 149, 152 (Tex. Civ.App. — Austin 1964, writ refd n.r.e.), but he must in fact represent them. He cannot ignore his clients and bind the State against their wishes.10 This is not to say that the Chief Justice is the sole arbiter. Both he and the Attorney General are named parties to this suit, and each has the right to be heard in this case. The Attorney General’s authority does not allow him to “close either the mouth of [Phillips] or the ears of the courts, when there are complaints that the Attorney General or his assistants are not in fact fulfilling their duty.” Cofer, 754 S.W.2d at 125.
B. Other Motions
We deny the Attorney General’s motion to disqualify Phillips’ counsel. We also deny plaintiffs’ attempt to nonsuit the Texas Judicial Districts Board, including its chair, Chief Justice Phillips. The motion was filed immediately after oral arguments before the en banc court on May 24, 1993. Rule 41(a) governs voluntary dismissals and provides that a plaintiff may dismiss an action without order of the court in two circumstances. The plaintiff must either file the notice of dismissal before the adverse party serves its answer or summary judgment motion, whichever occurs first, or file a stipulation of dismissal signed by all parties who have appeared in the case. Fed.R.Civ.P. 41(a)(1). The notice of nonsuit comes almost five years after the defendants have answered, and none of the defendant-aligned parties has signed the motion. Plaintiffs have no unilateral right to dismiss the Chief Justice and Judicial Districts Board. We will not permit plaintiffs to seek injunctive relief against the office held by Chief Justice Phillips for almost five years and then dismiss him when he declines to settle. See Davis v. Huskipower Outdoor Equipment Corp., 936 F.2d 193, 199 (5th Cir.1991) (affirming refusal to dismiss defendant more than a year after the case was removed to federal court); Radiant Technology Corp. v. Electrovert USA Corp., 122 F.R.D. 201 (N.D.Tex.1988) (motion to voluntarily dismiss under Rule 41 should be denied when plaintiff seeks to circumvent an expected adverse result).
We deny the motion of the district judges as Defendant-Intervenors to realign General Morales with plaintiffs. Morales’ efforts to settle the case do not require this
*844measure. He is entitled to take a position in settlement negotiations that is different from his trial posture. However, if the Attorney General changes his views on the merits of the case, realigning him with the plaintiffs may be appropriate. Cf. Delchamps, Inc. v. Alabama State Milk Control Bd., 324 F.Supp. 117, 118 (M.D.Ala.1971) (allowing Alabama Attorney General, who like the Texas Attorney General took an oath to defend both state and federal law, to realign himself with plaintiffs to challenge the federal constitutionality of a state law). We also deny Judge Wood’s motion to disqualify General Morales as counsel for the State. While we have rejected his claimed power to bind against their will state officials he is charged to represent, he is nonetheless their counsel.
C. The Intervenors
The Attorney General may represent state officials in their official capacities, but there is no contention that General Morales represents Judges Wood and Entz.11 They have intervened in their personal capacities and have elected to obtain their own counsel.12 As we earlier observed, the proposed consent decree would allow Judge Wood and Judge Entz to continue to run county-wide. General Morales urges that they therefore lack standing to either prosecute the suit or object to the proposed decree.
To this point, the standing of the intervening parties has not been questioned. To the contrary, the intervenors played an important role at trial and have, since taken the lead. After the federal district judge’s ruling in favor of plaintiffs, the notice of appeal was first filed by Judges Wood and Entz, not by the Attorney General. Only the district judge’s adherence to nonpartisan elections prodded the Attorney General to appeal. The Houston Lawyers’ Association intervened by the same order as the intervening judges and carried the appeal from our first en banc decision to the United States Supreme Court.13 Even now, no one questions the earlier uncontested standing of the inter-venors; nor could they. Wood and Entz intervened in part to protect their tenure as elected judges. The district court found that they were illegally elected.
Of course, these intervenors must satisfy Article III to appeal on their own. Diamond v. Charles, 476 U.S. 54, 68, 106 S.Ct. 1697, 1706, 90 L.Ed.2d 48 (1986); Didrickson v. United States Department of the Interior, 982 F.2d 1332, 1337-39 (9th Cir.1992); United States v. Western Elec. Co., 900 F.2d 283 (D.C.Cir.1990). A case or controversy between the State and plaintiffs remains. The parties have a right to a determination of that appeal, unless they consent to a remand. See Wheeler v. American Home Products Corp., 582 F.2d 891, 896 (5th Cir.1977) (“once intervention has been allowed, the original parties may not stipulate away the rights of the intervenor”); see also Sheffield v. Itawamba County Bd. of Supervisors, 439 F.2d 35, 36 (5th Cir.1971) (“having instituted a public lawsuit to secure rectification for a constitutional wrong of wide dimension, *845[plaintiffs] cannot privately determine its destiny”). Put another way, the proposed settlement does not deprive this court of its jurisdiction to hear the appeal independently perfected by Judges Wood and Entz, an appeal from a decision that declared their elections illegal.
Even assuming the proposed settlement foreclosed the intervening judges’ standing to protect their tenure, Wood and Entz would still have a sufficient stake in the litigation to satisfy the Constitution. In an earlier opinion in this case we said
[ajsserting interests both as a Texas voter and as a sitting Texas district judge, Judge Sharolyn Wood moved to intervene on the side of the defendant — the state. The court allowed her to intervene in her personal capacity, permitting Dallas County District Judge Harold Entz to do so as well.
League of United Latin American Citizens v. Clements, 923 F.2d 365, 367 (5th Cir.1991) (emphasis added). . 'In the district court, Judge Entz moved to intervene as a defendant to defend on his interests as a judge, a lawyer, and a registered voter in and citizen of Dallas County. The court’s order granting intervention in his individual capacity encompasses all of these interests.
Thus, the proponents of remand view the judges’ intervention too narrowly, for Wood and Entz also have standing as voters. The settlement agreement would deprive voters of the right to vote for all judges with general jurisdiction over their county. The Eleventh Circuit recently confronted a similar situation. Meek v. Metropolitan Dade County, 985 F.2d 1471 (11th Cir.1993), was a voting rights challenge to the at-large election of county commissioners in Dade County, Florida. As here, individual voters challenged a liability finding that elected officials would not contest on appeal. Swann and Sampson were Dade County residents and voters. The district court denied them leave to intervene before trial. In a second request for leave to intervene, Swann and Sampson sought to preserve their right to appeal in the event of an adverse judgment and a decision by defendants not to appeal. The court found the at-large system illegal and, as feared, the County Commission decided not to appeal. When the district court denied their third motion to intervene, Swann and Sampson appealed.
Our sister court held that the district court abused its discretion in denying the intervention and affirmed the district court on the merits. The court held that the voters had standing, a sufficient interest both to intervene and carry the appeal when the state agency declined to do so. In its view, if the court were to deny standing to these voters, it “would be forced to conclude that most of the plaintiffs also lack standing, a conclusion foreclosed by the many cases in which individual voters have been permitted to challenge election practices.” Id. at 1480 (citing Whitcomb v. Chavis, 403 U.S. 124, 91 S.Ct. 1858, 29 L.Ed.2d 363 (1971); Baker v. Carr, 369 U.S. 186, 82 S.Ct. 691, 7 L.Ed.2d 663 (1962)). We agree that the standing of voters in a voting rights case cannot be gainsaid. See also O’Hair v. White, 675 F.2d 680, 688-90 (5th Cir.1982) (en banc); Henderson v. Fort Worth Independent School Dist., 526 F.2d 286, 288-90 (5th Cir.1976).14
D. Consent Decrees
Even if all of the litigants were in accord, it does not follow that the federal court must do their bidding. The proposal is not to dismiss the lawsuit, but to employ the injunctive power of the federal court to achieve a result that the Attorney General and plaintiffs were not able to achieve through the political process. The entry of a consent decree is more than a matter of agreement among litigants. It is a “judicial act.” United States v. Swift & Co., 286 U.S. 106, 115, 52 S.Ct. 460, 462, 76 L.Ed. 999 (1932). “[W]hen [the court] has rendered a consent judgment it has made an adjudication.” Kaspar Wire Works, Inc. v. Leco Eng’g & Machine, Inc., 575 F.2d 530, 538-39 (5th Cir.1978) (quoting IB James W. Moore et a!., Moore’s Federal Practice ¶ 0.409[5]). *846Courts must exercise equitable discretion before accepting litigants’ invitation to perform the judicial act.
A consent decree must arise from the pleaded case and further the objectives of the law upon which the complaint is based. See Local No. 93, Int’l Ass’n of Firefighters v. City of Cleveland, 478 U.S. 501, 525, 106 S.Ct. 3063, 3077, 92 L.Ed.2d 405 (1986). When presented with a proposed judgment, the court “must not merely sign on the line provided by the parties.” United States v. City of Miami, 664 F.2d 435, 440 (5th Cir.1981) (era banc) (Rubin, J.).
Because the consent decree does not merely validate a compromise but, by virtue of its injunctive provisions, reaches into the future and has continuing effect, its terms require more careful scrutiny. Even when it affects only the parties, the court should, therefore, examine it carefully to ascertain not only that it is a fair settlement but also that it does not put the court’s sanction on and power behind a decree that violates Constitution, statute, or jurisprudence. ... If the decree also affects third parties, the court must be satisfied that the effect on them is neither unreasonable nor proscribed.
Id. at 441 (Rubin, J.) (emphasis added); see also Overton v. City of Austin, 748 F.2d 941, 952-53 (5th Cir.1984); Williams v. City of New Orleans, 729 F.2d 1554, 1559 (5th Cir.1984) (era banc) (Williams, J.).
The emphasized passage makes a critical point. A proposed consent decree is generally — as here — a request for the court to exercise its equitable powers. It involves the court’s sanction and power and is not a tool bending without question to the litigants’ will. As Justice Harlan wrote, “parties cannot, by giving each other consideration, purchase from a court of equity a continuing injunction.” System Federation No. 91, Ry. Employees’ Dep’t, AFL-CIO v. Wright, 364 U.S. 642, 651, 81 S.Ct. 368, 373, 5 L.Ed.2d 349 (1961).15
We have recognized that when fewer than all litigants forge a consent decree, issues affecting other parties remain to be adjudicated. City of Miami, 664 F.2d at 440 (Rubin, J.). As eleven judges recognized in the same case, our preferences for settlement and accord are insufficient to justify the imposition of a decree that infringes upon the rights of third parties. See id. at 451 (Gee, J., concurring and dissenting). A consent decree “cannot dispose of the valid claims of nonconsenting intervenors; if properly raised, these claims remain and may be litigated by the intervenor.” Local 93, 478 U.S. at 529, 106 S.Ct. at 3079.
Courts must be especially cautious when parties seek to achieve by consent decree what they cannot achieve by their own authority. Consent is not enough when litigants seek to grant themselves powers they do not hold outside of court. People Who Care v. Rockford Bd. of Educ., 961 F.2d 1335, 1337 (7th Cir.1992). For example, a local government may not use a consent decree to avoid a state law requiring a referendum before the issuance of construction bonds. Dunn v. Carey, 808 F.2d 555, 560 (7th Cir.1986).
We expressed our concern regarding the risks attending consent decrees in Overton v. City of Austin, 748 F.2d 941 (5th Cir.1984). In that case, plaintiffs and the city attorney, acting for the city council, proposed a decree substituting single-member council districts for the at-large council established by the city charter. A dissenting council member maintained that the council lacked the authority to change the existing scheme without a city-wide referendum. Id. at 947 n. 5. In the district court, several black voters sought to intervene as defendants on the ground that subdistricting would curtail their voting power. Id. at 944. The plaintiffs petitioned for a writ of mandamus to compel the district court to implement the proposed decree without further consideration. We refused to issue the writ. In doing so, Over-ton recognized the danger of manipulation faced by federal courts. We may be asked to effectuate substantive results that government officials are not empowered to bring about themselves. Id. at 956. The risk can *847be realized in many ways, but is palpable where sharply divided state officials would draw the federal courts into a partisan political battle.
Our job is to decide a case or controversy. The parties’ high-strung rhetoric does not fully obscure the reality that a live controversy yet exists. By declining to remand this case, we do not slow one whit any march for change in Texas. Its elected leaders are always free to pursue whatever scheme they think best, through the normal political process. Texas links the jurisdiction and electoral bases of its district judges and the still-contested question for this court is its legality.
The procedural posture of this case when the request to remand to the district court was heard is important. The issues in this case were well known to the entire court. The case had been fully tried and its appeal had twice been before a panel of this court and was before the en banc court a second time. The issues had been fully aired in the panel majority and dissenting opinion when this court vacated the panel opinion. In sum, we are asked to remand to the district court to consider entry of a “consent” decree and to decide whether it would “put the court’s sanction on and power behind a decree that violates Constitution, statute, or jurisprudence.” City of Miami, 664 F.2d at 441 (Rubin, J.). More precisely put, any federal decree must be a tailored remedial response to illegality. Cf. Shaw v. Reno, — U.S. -, 113 S.Ct. 2816, 125 L.Ed.2d 511 (1993). We are asked to remand for this determination although we are not persuaded that there is any illegality.
It is not a matter of our withholding announcement of our decision. We could not, in any event, remand without correcting the district court’s misapprehensions of law, found even by our dissenting colleagues. Significant legal errors infected the trial court’s earlier judgment, including its refusal to consider the effect of partisan voting, its finding of liability in Travis County now undefended, its selective aggregation of language . and ethnic minorities, its refusal to accord weight to the State’s linkage interest in the totality of the circumstances, and finally, its heavy reliance upon historical societal discrimination without bringing this history home to this case. We cannot escape this error-correcting task — and when it is done, there is no case. The amicus United States agrees with our conclusion that, once the proper legal standards are determined, the record presents no factual issue that needs revisiting. It follows that the proposed consent decree cannot respond to sufficiently identified illegality — because the record demonstrates that there is none.
E. Chisom v. Edwards
Finally, the parties urging remand point to Chisom v. Edwards, 970 F.2d 1408 (5th Cir.1992), where we remanded a voting rights case for the district court to enter a consent decree. That case challenged the method of electing Louisiana’s Supreme Court Justices. Chisom v. Roemer, — U.S. -,-, 111 S.Ct. 2354, 2358, 115 L.Ed.2d 348 (1991). Our remand in Chisom, however, resulted from different circumstances.
First, all parties joined the motion to remand, as we were careful to point out in our order:
The Joint Motion to Remand to Effectuate Settlement filed by all parties is hereby granted; and this case is remanded to the United States District Court for the Eastern District of Louisiana for the limited purpose of effectuating a settlement. Jurisdiction of the appeals is hereby retained. Upon notification that a consent judgment has been entered by the district court, the appeals will be dismissed. We express no opinion, of course, on the settlement or judgment.
Chisom, 970 F.2d at 1409 (emphasis added). As we have discussed, the same is not true here.16
*848Second, the parties in Chisom came to this court asking for remand carrying a duly enacted state law with them. They did not seek to invoke the preemptive force of the federal law. The decree in Chisom was agreed to by all parties and adopted into law by the state legislature. The consent decree did not set aside any state laws — and not by accident. It was carefully crafted to that end. In Louisiana, the legislature can create more supreme court districts with a two-thirds vote from both houses. La. Const. Art. 5, § 4.17 Article 5, § 3 of the Louisiana Constitution fixes the number of supreme court justices at seven and establishes that each shall serve a ten-year term.18 Because the state wished to create the Orleans district without upsetting the terms of the sitting justices, Louisiana had to temporarily expand the supreme court to eight members.19
While § 3 limits the size of the supreme court to seven justices, Art. 5, § 5(A) permits the Louisiana Supreme Court to “assign a sitting or retired judge to any court.” La. Const. Art. 5, § 5(A). The legislature therefore created an additional place for a judge on the Court of Appeal for the Fourth Circuit, who, upon election, would be assigned to the supreme court to serve, in reality, as the eighth justice. See La.Rev.Stat.Ann. § 13:312.4 (West Supp.1993). This temporary judgeship was to expire with a vacancy on the supreme court from the first district. The vacancy would be filled by an election in the newly created seventh district comprised of Orleans Parish. La.Rev.Stat.Ann. § 13:101.1 (West Supp.1993). Both of these provisions were contained in Act 512 which, after receiving the required two-thirds vote in both houses of the legislature, became law on June 22, 1992. Official Journal of the Proceedings of the Senate of the State of Louisiana, 18th Reg. Sess. at 24 (June 18, 1992); Official Journal of the Proceedings of the House of the State of Louisiana, 18th Reg. Sess. at 31 (June 16,1992). The Louisiana Legislature provided that Act 512 would not go into effect unless the federal court entered a consent decree in Chisom. La. Rev.Stat.Ann. § 13:101.1 (West Supp.1993).
The Texas Legislature refused to take positive action, and the settlement agreement attempts to avoid ’ constitutional requirements. The Texas Constitution requires that judges be elected from districts no smaller than a county, absent a majority vote by the citizens of that county. Tex. Const. Art. 5, §§ 7, 7a(i).20 The settlement agreement is not contingent on approval by the voters of each county. The legislature has not proposed a constitutional amendment. It has made no laws.
*849F. Federalism
Then we have all sides claiming the high ground of federalism. Some of the assertions are creative. The suggestion that state political groups, unable to muster sufficient political force to change the system, can by “agreement” enlist the preemptive power of the federal court to achieve the same end stands federalism on its head. Of course, we defer to legislative will and state decision. Here, the “decision” to which we are asked to defer is a decision by a political faction that the federal court should order the state to change its system. We do not share this curious view of federalism.
III. Racial Bloc Voting
As amended, § 2 of the Voting Rights Act prohibits states from imposing or applying any “standard, practice, or procedure ... 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.” A minority group may establish a violation of this provision by proving “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.”21 Congress intended “to make clear that proof of discriminatory intent is not required to establish a violation of Section 2” by “restoring] the legal standards” which prevailed in constitutional voting discrimination cases prior to Mobile v. Bolden, 446 U.S. 55, 100 S.Ct. 1490, 64 L.Ed.2d 47 (1980). S.Rep. 417 at 2, reprinted in 1982 U.S.Code Cong. & Admin.News at 206. Specifically, the 1982 amendments “codify” the “results test” articulated in White v. Regester, 412 U.S. 755, 93 S.Ct. 2332, 37 L.Ed.2d 314 (1973). Id.
Section 2 claims brought against multimember schemes are governed by the framework established in Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). Under Gingles, plaintiffs challenging an at-large system on behalf of a protected class of citizens must demonstrate that (1) the group is sufficiently large and geographically compact to constitute a majority in a single-member district; (2) it is politically cohesive; and (3) the white majority votes sufficiently as a bloc to enable it usually to defeat the minority’s preferred candidate. Growe v. Emison, — U.S. —,-, 113 S.Ct. 1075, 1084, 122 L.Ed.2d 388 (1993); Gingles, 478 U.S. at 50-51, 106 S.Ct. at 2766. Satisfaction of these three “preconditions,” Voinovich v. Quitter, — U.S. -, -, 113 S.Ct. 1149, 1157, 122 L.Ed.2d 500 (1993), is necessary, Gingles, 478 U.S. at 50, 106 S.Ct. at 2766, but not sufficient to establish liability under § 2. Chisom v. Roemer, — U.S. -, - — , 111 S.Ct. 2354, 2365, 115 L.Ed.2d 348 (1991); Citizens for Better Gov’t v. City of Westwego, 946 F.2d 1109, 1116 (5th Cir.1991) (Westwego III). Plaintiffs must also show that, under the “totality of circumstances,” they do not possess the same opportunities to participate in the political process and elect representatives of their choice enjoyed by other voters. Courts are guided in this second inquiry by the so-called Zimmer factors listed in the Senate Report.22
*850A- central issue here, one that divided the panel and one over which the parties vigorously disagree, concerns Gingles’ white bloc voting inquiry and the closely related Zimmer factor directing courts to examine “the extent to which voting ... is racially polarized.” S.Rep. 417 at 29, reprinted in 1982 U.S.Code Cong. & Admin.News at 206. As the Court in Gingles held, the question here is not whether white residents tend to vote as a bloc, but whether such bloc voting is “legally significant.” Gingles, 478 U.S. at 55, 106 S.Ct. at 2768; Salas v. Southwest Texas Jr. College Dist., 964 F.2d 1542, 1553 (5th Cir.1992). In finding a violation of § 2 in each of the nine challenged counties, the district court held that plaintiffs need only demonstrate that whites and blacks generally support different candidates to establish legally significant white bloc voting. Because “it is the difference between choices made by blacks and whites alone ... that is the central inquiry of § 2,” the court excluded evidence tending to prove that these divergent voting patterns were attributable to factors other than race as “irrelevant” and “legally [in]competent.”
On appeal, defendants contend that the district court erred in refusing to consider the nonracial causes of voting preferences they offered at trial. Unless the tendency among minorities and whites to support different candidates, and the accompanying losses by minority groups at the polls, are somehow tied to race, defendants argue, plaintiffs’ attempt to establish legally significant white bloc voting, and thus their vote dilution claim under § 2, must fail. When the record indisputably proves that partisan affiliation, not race, best explains the divergent voting patterns among minority and white citizens in the contested counties, defendants conclude, the district court’s judgment must be reversed.
We agree. The scope of the Voting Rights Act is indeed quite broad, but its rigorous protections, as the text of § 2 suggests, extend only to defeats experienced by voters “on account of race or color.” Without an inquiry into the circumstances underlying unfavorable election returns, courts lack the tools to discern results that are in any sense “discriminatory,” and any distinction between deprivation and mere losses at the polls becomes untenable. In holding that the failure of minority-preferred candidates to receive support from a majority of whites on a regular basis, without more, sufficed to prove legally significant racial bloc voting, the district court loosed § 2 from its racial tether and fused illegal vote dilution and political defeat. In so doing, the district court ignored controlling authorities: Whitcomb v. Chavis, 403 U.S. 124, 91 S.Ct. 1858, 29 L.Ed.2d 363 (1971), which established a clean divide between actionable vote dilution and “political defeat at the polls”; the 1982 *851amendments, enacted to restore a remedy in cases “where a combination of public activity and private discrimination have joined to make it virtually impossible for minorities to play a meaningful role in the electoral process,” Hearings on the Voting Rights Act Before the Subcomm. on the Constitution of the Senate Comm, of the Judiciary, 97th Cong., 2d Sess. 1367-68 (statement of Prof. Drew Days) (emphasis added); and Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986), where a majority of the Justices rejected the very test employed by the district court as a standard crafted to shield political minorities from the vicissitudes of “interest-group politics rather than a rule hedging against racial discrimination.” Id. at 83, 106 S.Ct. at 2782 (White, J., concurring); id. at 101, 106 S.Ct. at 2792 (O’Connor, J., joined by Burger, C.J., Powell and Rehnquist, JJ., concurring). We must correct these errors.
A. Whitcomb v. Chavis and White v. Regester
The Senate Report indicates that the 1982 amendments to § 2 were intended to “codify” the results test as employed in White and Whitcomb. See S.Rep. 417 at 2, 20-23, 32-33, reprinted in 1982 U.S.Code Cong. & Admin.News at 197-201, 210-11; Gingles, 478 U.S. at 97, 106 S.Ct. at 2790 (O’Connor, J., concurring) (“In enacting § 2, Congress codified the ‘results’ test this Court had employed, as an interpretation of the Fourteenth Amendment, in White and Whitcomb ”); Jones v. City of Lubbock, 727 F.2d 364, 379 (5th Cir.1984) (the amended § 2 “codifies pre-Bolden voting dilution law”). Consequently, “it is to Whitcomb and White that we should look in the first instance in determining how'great an impairment of minority voting strength is required to establish vote dilution in violation of § 2.” Gin-gles, 478 U.S. at 97, 106 S.Ct. at 2790 (O’Con-nor, J., concurring).
In Whitcomb, black citizens residing in one part of Marion County, referred to as the “ghetto” by the Court, claimed that the county’s at-large method of electing members to the state legislature unconstitutionally diluted their votes. The “[s]trong differences” between “ghetto” residents and adjacent communities “in terms of housing conditions, income and educational levels, rates of unemployment, juvenile crime, and welfare assistance,” 403 U.S. at 132, 91 S.Ct. at 1863,23 correlated closely with voting patterns in the county. “Ghetto” residents “voted heavily Democratic,” but since the county’s more affluent white majority consistently voted Republican, black-preferred candidates were defeated in four of the five elections between 1960 and 1968. Id. at 150, 91 S.Ct. at 1872. The Whitcomb Court recognized that the at-large electoral scheme caused the “voting power of ghetto residents [to be] ‘cancelled out,’ ” id. at 153, 91 S.Ct. at 1874, but held that this result by itself did not provide grounds for relief. Noting that blacks enjoyed full access to the political process,24 the Court reasoned that “had the Democrats won all of the elections or even most' of them, the ghetto would have no justifiable complaints about representation.” Id. at 152, 91 S.Ct. at 1873. For this reason, the Court concluded that the “failure of the ghetto to have legislative seats in proportion to its population emerges more as a function of losing elections than of built-in bias against poor Negroes.” Id. at 153, 91 S.Ct. at 1874.
The Whitcomb Court was reluctant to view the plaintiffs’ claims of vote dilution as anything more than “a euphemism for political defeat at the polls,” id., for, absent evidence of a lack of access to the political system, there was no principle by which the Court could distinguish the “ghetto’s” claims and those of other unsuccessful political groups:
*852[A]re poor Negroes of the ghetto any more under-represented than poor ghetto whites who also voted Democratic and lost, or any more discriminated against than other interest groups or voters in Marion County with allegiance to the Democratic Party, or, conversely, any less represented than Republican areas or voters in years of Republican defeat? We think not. The mere fact that one interest group or another concerned with the outcome of Marion County elections has found itself outvoted and without legislative seats of its own provides no basis for invoking constitutional remedies where, as here, there is no indication that this segment of the population is being denied access to the political system.
Id. at 154-55, 91 S.Ct. at 1875. To grant relief to black residents in this ease, the Court held, “would make it difficult to reject claims of Democrats, Republicans, or members of any political organization in Marion County who live in what would be safe districts in a single-member district system but who in one year or another, or year after year, are submerged in a multimember district vote.” Id. at 156, 91 S.Ct. at 1876.
The Court’s assertion that plaintiffs’ racial vote dilution claim was indistinguishable from complaints which might be brought by any unsuccessful interest group hinged on its determination that “ghetto” residents did not suffer from a lack of access to the political process. Despite the presence of vast disparities in virtually every significant measure of socioeconomic status, the Court found that black voters stood on the same footing with whites in vying for representation within Marion County. “Ghetto” residents had in fact experienced a string of losses at the polls in recent years, but these, defeats were shared equally among all members of the Democratic Party.
The Court confronted very different circumstances two years later in White v. Regester, 412 U.S. 755, 93 S.Ct. 2332, 37 L.Ed.2d 314 (1973). The Court confirmed Whitcomb’s rejection of the claim that “every racial or political group has a constitutional right to be represented in the state legislature,” id. at 769, 93 S.Ct. at 2341, and reiterated the standard established in its earlier decision: a minority group must prove “that its members had less opportunity than did other residents in the district to participate in the political processes and to elect legislators of their choice.” Id. at 766, 93 S.Ct. at 2339 (citing Whitcomb, 403 U.S. at 149-50, 91 S.Ct. at 1872). Unlike the plaintiffs in Whit-comb, however, the black residents of Dallas County and the Hispanic voters in Bexar County each established that they had been effectively excluded from the political processes leading to the nomination and election of the Texas House of Representatives. 412 U.S. at 766-70, 93 S.Ct. at 2339-41.
Specifically, black voters in Dallas labored under the yoke of Texas’ long history of official discrimination and were subjected to several procedural devices which, while not invidious in themselves, “enhanced the opportunity for racial discrimination.” Id. at 766, 93 S.Ct. at 2339. “More fundamentally,” the Court noted, the Dallas Committee for Responsible Government, “a white-dominated organization that is in effective control of Democratic Party candidate slating,” had slated only two black candidates in its history, who, not coincidentally, constituted the only two blacks ever to have served in the Dallas County delegation to the Texas House since Reconstruction. Id. at 766-67, 93 S.Ct. at 2340. The DCRG failed to display any “good-faith .concern for the political and other needs and aspirations of the Negro community,” and in fact regularly relied on racial campaign tactics to defeat candidates supported by black residents. Id. at 767, 93 S.Ct. at 2340. Consequently, the Court had no reason to disturb the district court’s conclusion “that ‘the black community has been effectively excluded from participation in the Democratic primary selection process,’ and was therefore generally not permitted to enter into the political process in a reliable and meaningful manner.” Id. (quoting Graves v. Barnes, 343 F.Supp. 704, 726 (W.D.Tex.1972)).
The Court also upheld a similar finding that Mexican-Americans likewise had been “ ‘effectively removed from the political processes of Bexar [County] in violation of all the Whitcomb standards.’ ” Id. 412 U.S. at *853769, 93 S.Ct. at 2341 (quoting Graves, 343 F.Supp. at 733). Like black residents. of Texas, Mexican-Americans “had long ‘suffered from, and eontinue[d] to suffer from, the results and effects of invidious discrimination and treatment in the field of education, employment, economics, health, politics and others.’” Id., 412 U.S. at 768, 93 S.Ct. at 2340 (quoting Graves, 343 F.Supp. at 728). In addition, the district court determined that “cultural and language barrier[s] ... ‘conjoined with the poll tax and the most restrictive voter registration procedures in the nation have operated to effectively deny Mexican-Americans access to the political processes in Texas even longer than the Blacks were formally denied access by the white primary.’ ” Id. (quoting Graves, 343 F.Supp. at 731). The exclusionary effects of past and present discrimination, the Court found, were palpably reflected in low voting registration among Mexican-Americans, the election of only five Bexar County Mexican-Americans to the Texas Legislature since 1880, and the county delegation’s unresponsiveness to the community’s interests. Id., 412 U.S. at 768-69, 93 S.Ct. at 2341. Given that the district court’s findings flowed from “a blend of history and an intensely local appraisal” of conditions in Bexar County, the Court was “not inclined to overturn” its conclusion that the multimember district “invidiously excluded Mexican-Americans from effective participation in political life.” Id. at 769, 93 S.Ct. at 2341. As we will explain, this earlier time in Texas history and the elections at issue' here present stark contrasts. The record before us contains no evidence that past or present discrimination has affected minorities’ political access in any way.
The principles announced and applied in Whitcomb and White are instructive and, we believe, controlling. As Justice White, the author of these opinions, recently indicated, the central “theme” of Whitcomb and White is “that it is not mere suffering at the polls but discrimination in the polity with which the Constitution is concerned.” Shaw v. Reno, — U.S. -, -, 113 S.Ct. 2816, 2835, 125 L.Ed.2d 511 (1993) (White, J., dissenting). Beyond the bounds of this litigation, the clarity with which the Whitcomb Court articulated the principles underlying the “results” test has largely forestalled confusion or doubt, even among those whom plaintiffs might be inclined to count as allies. See, e.g., Jones v. City of Lubbock, 727 F.2d 364, 384 (5th Cir.1984) (“Even where an at-large system interacts with a racially or ethnically polarized electorate to the disadvantage of the minority, the ‘result’ is not necessarily a denial of political access.... [T]he ‘result’ in Whitcomb [is] that polarized voting does not render an at-large system dilutive of minority voting strength”); Pamela S. Karlan, Undoing the Right Thing: Single-Member Offices and the Voting Rights Act, 71 Va.L.Rev. 1, 22 n. 78 (1991). Justice Marshall, for example, provided a clear explanation of the Court’s holding in his dissent in Mobile v. Bolden, 446 U.S. 55, 100 S.Ct. 1490, 64 L.Ed.2d 47 (1980):
In Whitcomb v. Chavis, we again repeated and applied the Fortson [v. Dorsey, 379 U.S. 433, 85 S.Ct. 498, 13 L.Ed.2d 401 (1965) ] [effects] standard, but determined that the Negro community’s lack of success at the polls was the result of partisan politics, not racial vote dilution. The Court stressed that' both the Democratic and Republican Parties had nominated Negroes and several had been elected. Negro candidates lost only when their entire party slate went down to defeat. In addition, the Court was, impressed that there was no finding that officials had, been unresponsive to Negro concerns.
Id. at 109, 100 S.Ct. at 1522 (Marshall, J., dissenting) (citations omitted).
Justice Marshall’s references to the “lack of success at the polls” as a “result” of “partisan politics, not racial vote dilution,” closely tracks the relevant language in Whit-comb, where the Court held that the “can-celling] out” of the “voting power of ghetto residents” was more “a function of losing elections” or “political defeat” than of “built-in bias against poor Negroes.” 403 U.S. at 153, 91 S.Ct. at 1874. Absent evidence that minorities have been excluded from the political process, a “lack of success at the polls” is not sufficient to trigger judicial intervention. Courts must undertake the additional inquiry into the reasons for, or causes of, these electoral losses in order to determine whether *854they were the product of “partisan politics” or “racial vote dilution,” “political defeat” or “built-in bias.” It is only upon concluding that a minority group’s failure to prevail at the polls, that is, their failure to attract the support of white voters, was the “result” or “function” of “racial vote dilution” or “built-in bias,” that a court may find that minority plaintiffs have suffered “a denial or abridgement of the right ... to vote on account of race or color.” In sum, Whitcomb unmistakably prescribes the very inquiry into the causes underlying the lack of support for minority-preferred candidates among white voters with which the district court dispensed.
As Justice Marshall suggested, failures of a minority group to elect representatives of its choice that are attributable to “partisan politics” provide no grounds for relief. Section 2 is “a balm for racial minorities, not political ones — even though the two often coincide.” Baird v. Consolidated City of Indianapolis, 976 F.2d 357, 361 (7th Cir.1992) (citing Whitcomb). “The Voting Rights Act does not guarantee that nominees of the Democratic Party will be elected, even if black voters are likely to favor that party’s candidates.” Id. Rather, § 2 is implicated only where Democrats lose because they are black, not where blacks lose because they are Democrats. While this rule is easier stated than applied, the Whitcomb Court’s application of the “results” test to the facts before it provides helpful and indeed dispositive guidance. As we explain in greater detail below, the Court’s dismissal in Whitcomb of the plaintiffs’ vote dilution claim as a “mere euphemism for political defeat at the polls,” despite evidence of polarized voting, the lingering effects of past discrimination, and little electoral success among minority candidates, precludes finding a violation of § 2 in most, but not all, of the counties at issue.
B. The 1982 Amendments
The Senate Report accompanying the 1982 amendments to § 2 states that Congress intended to “codify” the “results test” articulated and employed in Whitcomb and White. Congress of course retained the statutory language restricting relief under § 2 to “denial[s] or abridgments] of the right ... to vote on account of race or color.” This limitation was not so much the product of legislative discretion as constitutional imperative, given that the scope of Congress’ remedial power under the Civil War Amendments is defined in large part by the wrongs they prohibit. See, e.g., City of Rome v. United States, 446 U.S. 156, 206, 100 S.Ct. 1548, 1576, 64 L.Ed.2d 119 (1980) (Rehnquist, J., dissenting); Oregon v. Mitchell, 400 U.S. 112, 152, 91 S.Ct. 260, 279, 27 L.Ed.2d 272 (1970) (Harlan, J., concurring in part and dissenting in part). Thus, the Senate Report explained that the 1982 amendments avoided constitutional difficulty because “the very terms and operation of [§ 2] confine its application to actual racial discrimination.” S.Rep. 417 at 43, reprinted in 1982 U.S.Code Cong. & Admin.News at 221.
Congress embraced Whitcomb on terms consistent with § 2’s limitation to cases of “actual racial discrimination.” Noting that the claim before the Court in Whitcomb alleged vote dilution on grounds that “black ghetto residents with [distinct] legislative interests had been consistently underrepresented in the legislature,” the Senate Report recounted what it regarded as the relevant facts of the case:
The evidence showed that the ghetto area voted Democratic, that the Republicans won four of the five elections from 1960 to 1968, and that in 1964, when the Democrats won, ghetto area senators and representatives were elected. Nine blacks had in fact been elected to the legislature from the at-large districts between [1960] and 1968.
Id. at 20-21, reprinted in 1982 U.S.Code Cong. & Admin.News at 198. The facts cited by the Senate mirror those previously identified by Justice Marshall in Bolden and stressed here: Plaintiffs were unsuccessful in years in which their party suffered electoral defeat; they were able to elect representatives of their choice when their party prevailed. Not surprisingly, the Senate adopted Whitcomb’s central teaching in presenting what it understood to be the kernel of the decision:
*855The failure of the ghetto to have legislative seats in proportion to its population emerges more as a function of losing elections than of built-in bias against poor Negroes. The voting power of ghetto residents may have been “cancelled out,” as the district court held, but this seems a mere euphemism for political defeat at the polls.
Id. at 21 (quoting Whitcomb, 403 U.S. at 153, 91 S.Ct. at 1874), reprinted in 1982 U.S.Code Cong. & Admin.News at 198.
In keeping with Whitcomb’s sharp distinction between “built-in bias” and “political defeat at the polls,” the Senate Report indicated that a proper application of the results test requires courts to “distinguish^ ] between situations in which racial politics play an excessive role in the electoral process, and' communities in which they do not.” Id. at 33, reprinted in 1982 U.S.Code Cong. & Admin.News at 211. The Senate Report, again following Whitcomb, accorded this inquiry into “racial bloc voting,” that is, whether “ ‘race is the predominant determinant of political preference,’ ” dispositive significance: Absent a showing of “racial bloc voting,” the Senate Report asserted, “it would be exceedingly difficult for plaintiffs to show that they were effectively excluded from fan-access to the political process under the results test.” Id. (quoting S.Rep. 417 at 148 (Report of the Subcommittee on the Constitution), reprinted in 1982 U.S.Code Cong. & Admin.News at 321). Since the results test itself, contrary to critics’ charges, “makes no assumptions one way or the other about the role of racial political considerations in a particular community,” id. at 34, reprinted in 1982 U.S.Code Cong. & Admin.News at 212, the Senate Report emphasized that plaintiffs must supply affirmative proof of “racial bloc voting.” The “mere existence of underrepre-sentation plus a history of dual schools” plainly does not suffice to make out a violation of § 2. Id.
It is difficult to see how the record in this case could possibly support a finding of liability under the approach outlined in the Senate Report.. Plaintiffs have not even attempted to establish proof of racial bloc voting by demonstrating that “race,” not, as defendants contend, partisan affiliation, “is the predominant determinant of political preference.” They have instead maintained, in the very teeth of the Senate Report, that such a showing is unnecessary. Because the district court accepted this argument, the test employed at trial enabled plaintiffs to prevail by proving little more than a lack of success at the polls and a history of discrimination. While this standard finds clear support in Justice Brennan’s plurality opinion in Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986), it “simply was not the approach used by the courts under the White/Zimmer test” and codified by Congress. S.Rep. 417 at 34, reprinted in 1982 U.S.Code Cong. & Admin.News at 212.
C. Thornburg v. Gingles
Justice Brennan’s discussion of the first and second Gingles factors received majority support. Gingles, 478 U.S. at 50-51, 56, 106 S.Ct. at 2766, 2769.25 With respect to the third element, however, five justices rejected Justice Brennan’s proposed standard for proving racial bloc voting. Id. at 83, 106 S.Ct. at 2782 (White, J., concurring); id. at 100-01, 106 S.Ct. at 2792 (O’Connor, J., joined by Burger, C.J., Powell and Rehnquist, JJ., concurring). For this reason, we believe that it is to these opinions, not Justice Brennan’s, that we should look in attempting to define the contours of the inquiry into legally significant bloc voting.
Despite the presence of express language to the contrary in the Senate Report, see S.Rep. 417 at 33 (“racial bloc voting” is established when “race is the predominant determinant of political preference”), reprinted in 1982 U.S.Code Cong. & Admin.News at 211, Justice Brennan held that racial bloc voting or “racially polarized voting” did not describe divergent “voting patterns for which *856the principal cause is race.” Gingles, 478 U.S. at 61, 106 S.Ct. at 2772. Instead, he asserted that “[i]t is the difference between the choices made by blacks and whites — not the reasons for that difference — that [matters].” Id. A consideration of “irrelevant variables” such as partisan affiliation or the race of the candidate, Justice Brennan urged, would “distort[] the equation and yield[] results that are indisputably incorrect under § 2 and the Senate Report.” Id. at 64, 106 S.Ct. at 2773.
Justice Brennan’s assertion that racial political considerations had no role in examining racial bloc voting was squarely rejected by five Justices in Gingles. 478 U.S. at 83, 106 S.Ct. at 2782 (White, J., concurring); id. at 100-01, 106 S.Ct. at 2792 (O’Connor, J., joined by Burger, C.J., Powell and Rehnquist, JJ., concurring). Justice White argued that
Justice Brennan states in Part III-C that the crucial factor in identifying polarized voting is the race of the voter and that the race of the candidate is irrelevant. Under this test, there is polarized voting if the majority of white voters vote for different candidates than the majority of the blacks, regardless of the race of the candidates. I do not agree. Suppose an eight-member multimember district that is 60% white and 40% black, the blacks being geographically located so that two safe black single-member districts could be drawn. Suppose further that there are six white and two black Democrats running against six white and two black Republicans. Under Justice Brennan’s test, there would be polarized voting and a likely § 2 violation if all the Republicans, including the two blacks, are elected, and 80% of the blacks in the predominately black areas vote Democratic.... This is interest-group politics rather than a rule hedging against racial discrimination. I doubt that this is what Congress had in mind in amending § 2 as it did, and it seems quite at odds with the discussion in Whitcomb v. Chavis, 403 U.S. 124, 149-160 [91 S.Ct. 1858, 1872-78, 29 L.Ed.2d 363] (1971).
Id. 478 U.S. at 83, 106 S.Ct. at 2782 (White, J., concurring) (emphasis added). Justice O’Connor joined Justice White in maintaining that evidence that white and minority voters generally supported different candidates did not constitute legally significant racial bloc voting where these patterns were attributable to partisan affiliation rather than the race of the candidate. She therefore rejected Justice Brennan’s position that
evidence that the divergent racial voting patterns may be explained in part by causes other than race, such as an underlying divergence in the interests of minority and white voters.... can never affect the overall vote dilution inquiry. Evidence that a candidate preferred by the minority group in a particular election was rejected by white voters for reasons other than those which made that candidate the preferred choice of the minority group would seem clearly relevant in answering the question whether bloc voting by white voters will consistently defeat minority candidates. Such evidence would suggest that another candidate, equally preferred by the minority group, might be able to attract greater white support in future elections.
I believe Congress also intended that explanations of the reasons why white voters rejected minority candidates would be probative of the likelihood that candidates elected without decisive minority support would be willing to take the minority’s interests into account. In a community that is polarized along racial lines, racial hostility may bar these and other indirect avenues of political influence to a much greater extent than in a community where racial animosity is absent although the interests of racial groups diverge. Indeed, the Senate Report clearly stated that one factor that could have probative value in § 2 cases was “whether there is a significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority group.” S.Rep., at 29. The overall vote dilution inquiry neither requires nor permits an arbitrary rule against consideration of all evidence concerning voting preferences other than statistical evidence of racial voting patterns. Such a rule would give no effect whatever to the Senate Report’s re*857peated emphasis on “intensive racial politics,” on “racial political considerations,” and on whether “racial politics ... dominate the electoral process” as one aspect of the “racial bloc voting” that Congress deemed relevant to showing a § 2 violation. Id., at 33-34. Similarly, I agree with Jtistice White that Justice Brennan’s conclusion that the race of the candidate is always irrelevant in identifying racially polarized voting conflicts with Whitcomb and is not necessary to the disposition of this case. Ante [478 U.S. at 83-84, 106 S.Ct.] at 2783 (concurring).
Id. at 100-01, 106 S.Ct. at 2792 (O’Connor, J., concurring) (emphásis added).
As courts and commentators alike have noted, Justice White and Justice O’Connor were united in their fidelity to Whitcomb’s distinction between vote dilution and partisan politics and in their opposition to Justice Brennan’s attempt to expunge this teaching from the bloc voting inquiry. See, e.g., Baird v. Consolidated City of Indianapolis, 976 F.2d 357, 361 (7th Cir.1992) (“Justice White ... observed] that system leading to the election of black Republicans could not be dismissed as discriminatory. To disregard the race of the victors, Justice White concluded, ‘is interest-group politics rather than a rule hedging against racial discrimination.’ Justice O’Connor agreed”) (citation omitted); Note, Voting Rights Act Section 2: Racially Polarized Voting and the Minority Community’s Representative of Choice, 89 Mich. L.Rev. 1038, 1044 (1991); Note, Defining the Minority Preferred Candidate Under Section 2, 99 Yale L.J. 1651, 1662-63 (1990). The division in Gingles between the Brennan plurality and the five Justices who supported the White/O’Connor approach cuts deep, reflecting quite different visions of voting rights and their statutory treatment. Since these five Justices expressly rejected a test that would permit § 2 liability to attach upon a showing that white and black citizens generally gave their votes to different candidates in favor of an inquiry into the possible explanations of these divergent voting patterns, we believe that it is this view, not Justice Brennan’s, that commands our allegiance. The district court’s failure to accord similar weight to this approach was not justified.
All members of the Court in Gingles agreed that only “legally significant” racial bloc voting is cognizable under § 2. They disagreed sharply, however, on the sort of proof that would implicate this provision. Justice Brennan held that a “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.” Gingles, 478 U.S. at 51, 106 S.Ct. at 2766-67. Justice O’Connor, on the other hand, argued that such a showing did not warrant judicial intervention: “[A] reviewing court should be required to find more than simply that the minority group does not usually attain an undiluted measure of electoral success.” Id. at 99, 106 S.Ct. at 2791 (O’Connor, J., concurring). Instead, she would require a court to “find that even substantial 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].’ ” Id. at 99-100, 106 S.Ct. at 2792 (quoting White, 412 U.S. at 765, 93 S.Ct. at 2339) (alterations in original).
Justice O’Connor’s admonition that federal courts should stay their hand absent proof that “even substantial minority success will be highly infrequent” receives formal expression in her insistence that the racial bloc voting inquiry must include an examination of the causes underlying divergent voting patterns. Both Justice Brennan and Justice O’Connor recognized that racial bloc voting is intimately related to the responsiveness of elected officials to the interests of minorities, one of the factors considered as part of the “totality of circumstances.” As Justice Brennan indicated, “[n]ot only does ‘[v]oting along racial lines’ deprive minority voters of their preferred representatives in these circumstances, it also ‘allows those elected to ignore [minority] interests without fear of political consequences.’ ” Id. 478 U.S. at 48 n. 14, 106 S.Ct. at 3106 n. 14 (quoting Rogers v. Lodge, 458 U.S. 613, 623, 102 S.Ct. 3272, 3278, 73 L.Ed.2d 1012 (1982) (alterations in original)). The close tie between bloc voting and representatives’ responsiveness noted by the Court in Rogers and confirmed by Justice *858Brennan rests on common sense: Public officials need not address concerns expressed by minorities so long as white bloc voting ensures that they will remain minority concerns. The Court in Rogers and Justice Brennan, however, differed sharply over the sort of polarized voting that might provide elected officials with such assurances and federal courts with grounds to intervene. The Court in Rogers held that this close identification was warranted only where racial political considerations were present, that is, where white bloc voting caused “minority candidates [to] lose elections solely because of their race.’’ Rogers, 458 U.S. at 623, 102 S.Ct. at 3279 (emphasis added). Justice Brennan’s approach, by contrast, assumes that political leaders may safely ignore minority concerns even where black and white voters are separated only by differing interests. Put another way, Justice Brennan’s bloc voting test accords governing majorities linked only by the perception of common interests the same permanence and thus relevance under § 2 as white blocs cemented by racial prejudice.
Justice O’Connor not only rejected Justice Brennan’s polarized voting standard but was also unwilling to join in the questionable assumption that minorities are unable to influence elections and secure the attention of public officials where these groups have been unsuccessful in their efforts to elect their preferred representatives. Gingles, 478 U.S. at 100-101, 106 S.Ct. at 2792 (O’Connor, J., concurring). Unlike Justice Brennan, she argued that “Congress also intended that explanations of the reasons why white voters rejected minority candidates would be probative of the likelihood that candidates elected without decisive minority support would be willing to take the minority’s interests into account.” Id. at 100, 106 S.Ct. at 2792 (O’Connor, J., concurring). Following Rogers, Justice O’Connor believed that a minority group’s prospects for future electoral success and the likelihood that elected officials will take account of their interests differ materially “in a community where racial animosity is absent although the interests of racial groups diverge.” Id. (O’Connor, J., concurring). A tendency among whites to cast their votes on the basis of race presents a far more durable obstacle to the coalition-building upon which minority electoral success depends than disagreements over ideology for, as Professor Ely observes, “prejudice blinds us to overlapping interests that in fact exist.” John Hart Ely, Democracy and Distrust 153 (1980). Representatives who owe their office to the support of majorities bound by prejudice need not attend to the interests of minorities, since the bias uniting their constituents ensures that these issues will remain minority concerns. Where, on the other hand, voting patterns correlate with partisan affiliation or perceived interest, the open channels of communication facilitate a recognition of points of common ground that might otherwise go undetected. Elected officials in these communities cannot ignore minority interests because this group might be part of the winning coalition that votes them out of office. The deep division between Justice Brennan and Justice O’Connor on the question of racial bloc voting thus reflects fundamentally different views of political factions and our constitutional and statutory arrangements for accommodating their simultaneous demands for fluidity and fixity.26
*859Given that the divergent voting patterns in this case are in most instances attributable to partisan affiliation rather than race, it is thus far from coincidental that the district court found no evidence of unresponsiveness on the part of elected officials in any of the contested counties. The irony, of course, is that the subdistricting remedy sought by plaintiffs provides most judges with the same opportunity to ignore minority voters’ interests without fear of political reprisal they would possess if elections were in fact dominated by racial bloc voting.
D. Partisan Politics
We need not hold that plaintiffs must supply conclusive proof that a minority group’s failure to elect representatives of its choice is caused by racial animus in the white electorate in order to decide that the district court’s judgment must be reversed. It is true that such a requirement could be inferred from the text of § 2 (prohibiting “denial[s] or abridgement[s] of the right ... to vote on account of race or color”); the caselaw Congress intended to codify in amending the provision, see, e.g., Whitcomb, 403 U.S. at 153, 91 S.Ct. at 1874 (vote dilution does not lie when losses at the polls do not reflect “built-in bias against poor Negroes”); the Senate Report, see S.Rep. 417 at 33 (equating proof of racial bloc voting with evidence that “race is the predominant determinant of political preference”), reprinted in 1982 U.S.Code Cong. & Admin.News at 211; the testimony of prominent supporters of the Act, see, e.g., Hearings on the Voting Rights Act Before the Subcomm. on the Constitution of the Senate Comm, of the Judiciary, 97th Cong., 2d Sess. 1367-68 (statement of Prof. Drew Days) (§ 2 implicated “where a combination of public activity and private discrimination have joined to make it virtually impossible for minorities to play a meaningful role in the electoral process”); and the controlling opinions of the Supreme Court. See Gingles, 478 U.S. at 100, 106 S.Ct. at 2792 (O’Connor, J., concurring) (distinguishing communities where polarized voting is attributable to “racial hostility” and those in which “racial animosity is absent although the interests of racial groups diverge”). There is also a powerful argument supporting a rule that plaintiffs, to establish legally significant racial bloc voting, must prove that their failure to elect representatives of their choice cannot be characterized as a “mere euphemism for political defeat at the polls,” Whitcomb, 403 U.S. at 153, 91 S.Ct. at 1874, or the “result” of “partisan politics.” Bolden, 446 U.S. at 100, 100 S.Ct. at 1517 (Marshall, J., dissenting).
Describing plaintiffs’ burden in terms of negating “partisan politics” rather than affirmatively proving “racial animus” would not be simply a matter of nomenclature. As Judge Wood emphasizes, there are many other possible non-racial causes of voter behavior beyond partisan affiliation. A rule conditioning relief under § 2 upon proof of the existence of racial animus in the electorate would require plaintiffs to establish the absence of not only partisan voting, but also all other potentially innocent explanations for white voters’ rejection of minority-preferred candidates. Factors that might legitimately lead white voters to withhold support from particular minority candidates include, for example, limited campaign funds, inexperience, or a reputation besmirched by scandal. Because these additional factors map only imperfectly onto partisan affiliation, detailed multivariate analysis might then be the evidence of choice. The argument would then be that without this additional inquiry, courts that confine their scrutiny to partisan voting might well find racial bloc voting in circumstances where the losses of minority-pre*860ferred candidates were actually attributable to causes other than race. This result, it is urged, might unfairly tip the scales in favor of liability.
This argument possesses considerable force. Certainly, the allocation of proof in § 2 cases must reflect the central purpose of the Voting Rights Act and its intended liberality as well as the practical difficulties of proof in the real world of trial. In countless areas of the law weighty legal conclusions frequently rest on methodologies that would make scientists blush. The use of such blunt instruments in examining complex phenomena and corresponding reliance on inference owes not so much to a lack of technical sophistication among judges, although this is often true, but to an awareness that greater certitude frequently may be purchased only at the expense of other values. Here, we are told that we cannot ignore the significant and, assertedly, unacceptable substantive consequences that would accompany a more nuanced bloc voting inquiry. Requiring plaintiffs affirmatively to establish that white voters’ rejection of minority-preferred candidates was motivated by racial animus would make racial bloc voting both difficult and, considering the additional analysis that would be needed, expensive to establish. See, e.g., McCrary, Discriminatory Intent: The Continuing Relevance of “Purpose ” Evidence in Vote-Dilution Lawsuits, 28 How.L.J. 463, 492 (1985). Moreover, it would facilitate the use of thinly-veiled proxies by permitting, for example, evidence that a minority candidate was regarded as “unqualified” or “corrupt” to defeat a claim that white voters’ refusal to support him was based on race or ethnicity. The argument continues that an inquiry into causation beyond partisan affiliation seems inconsistent with the fundamental division between “partisan politics” and “racial vote dilution” set out by the Court in Whitcomb and White and confirmed by Congress. Legal standards of necessity reflect a balance of competing considerations. Finally, the argument continues that limiting the racial bloc voting inquiry to a determination whether or not divergent voting patterns are attributable to partisan differences or an underlying divergence in interests best captures the mandate of § 2.27 Having said this, we heed not resolve the debate today. Whether or not the burden of the plaintiffs to prove bloc voting includes the burden to explain partisan influence, the result is the same. This is so even if the partisan voting is viewed as a defensive parry.
Finally, we recognize that even partisan affiliation may serve as a proxy for illegitimate racial considerations. Minority voters, at least those residing in the contested counties in this case, have tended uniformly to support the Democratic Party. At the same time, a majority of white voters in most counties have consistently voted for district court candidates fielded by the Republican Party. Noting this persistent, albeit imperfect correlation between party and race, plaintiffs assert that a determination that partisan affiliation best explains voting patterns should not foreclose § 2 liability in this case because the Republican and Democratic Parties are proxies for racial and ethnic groups in Texas. Whitcomb’s distinction between “racial vote dilution” and “political defeat at the polls” should not control, they contend, for “partisan politics” is “racial politics.”
We fully agree with the plaintiffs that the bloc voting inquiry, like the “question whether the political processes are ‘equally open,’ ” must rest “upon a searching practical evaluation of the ‘past and present reality.’ ” S.Rep. 417 at 30 (quoting White, 412 U.S. at 769-770, 93 S.Ct. at 2341), reprinted in 1982 U.S.Code Cong. & Admin.News at 208. Indeed, the refusal of Congress and the Supreme Court to equate losses at the polls with actionable vote dilution where these unfavorable results owe more to party than race may be traced directly to this “functional” view of political life. Plaintiffs are therefore entirely correct in maintaining that courts should not summarily dismiss vote dilution claims in cases where racially divergent voting patterns correspond with parti*861san affiliation as “political defeats” not cognizable under § 2.
We do not agree, however, that a “functional” and “practical” review of Texas judicial elections exposes political parties as proxies for race or ethnicity. In assessing the record before us, we do not indulge in the hopeful yet unrealistic assumption that decisions to support particular political parties among black and white voters in all cases rest on issues other than race. We instead focus on the same two factors cited by the Court in Whitcomb and the concurring Justices in Gingles. First, white voters constitute the majority of not only the Republican Party, but also the Democratic Party, even in several of the counties in which the former dominates. In Dallas County, for example, 30-40% of white voters consistently support Democrats, making white Democrats more numerous than all of the minority Democratic voters combined. The suggestion that Republican voters are galvanized by a “white” or “anti-minority” agenda is plausible only to the extent that the Democratic Party can be viewed as a vehicle for advancing distinctively minority interests, which clearly is not the case. At the same time, white Democrats have in recent years experienced the same electoral defeats as minority voters.- If we are to hold that these losses at the polls, without more, give rise to a racial vote dilution claim warranting special relief for minority voters, a principle by which we might justify withholding similar relief from white Democrats is not readily apparent. See Whitcomb, 403 U.S. at 153, 91 S.Ct. at 1874.
Second, both political parties, and especially the Republicans, aggressively recruited minority lawyers to run on their party’s ticket. Consequently, white as well as minority voters found themselves not infrequently voting against candidates sharing their respective racial or ethnic backgrounds in favor of their party’s nominee. In particular, the undisputed evidence discloses that white voters in most counties, both Republican and Democratic, without fail supported the minority candidates slated by their parties at levels equal to or greater than those enjoyed by white candidates, even where the minority candidate was opposed by a white candidate. In Dallas County, for example, Judge Wright, a black woman, received the greatest recorded percentage of the white vote (77%) in her race against a white Democrat. To conclude on this record that political parties serve as proxies for race is simply unwarranted. Because the evidence in most instances unmistakably shows that divergent voting patterns among white and minority voters are best explained by partisan affiliation, we conclude that plaintiffs have failed to establish racial bloc voting in most, but not all, of the counties.28
E. Two Objections
The Houston Lawyers’ Association and amicus the United States raise two particular objections that merit additional consideration. These arguments closely track those made by Justice Brennan — arguments rejected by five members of the Supreme Court in Gingles. Nevertheless, the urgency with which they are pressed here warrants a further explanation of the reasons underlying the views expressed by Justice White and Justice O’Connor in their separate opinions.
The Association contends that a require-' ment that plaintiffs prove that their failure to elect representatives of their choice is attributable to white bloc voting rooted in racial considerations is presumptively inconsistent with § 2’s focus on “results.” The Association reads this test to impose on plaintiffs the burden of affirmatively establishing that white voters are motivated by racial animus in selecting candidates. So characterized, the racial bloc voting standard we apply today állegedly contravenes the fundamental purpose of the 1982 amendments by reintroducing the “intent” test announced in Mobile v. Bolden, 446 U.S. 55, 100 S.Ct. 1490, 64 L.Ed.2d 47 (1980). See also Richard L. Engstrom, The Reincarnation of the Intent Standard: Federal Judges and At-Large Election Cases, 28 How.L.J. 495, 498 (1985). That is not so.
*862The Association does not seriously contend that the legislative history accompanying the amendments to § 2 lends direct support for its position. The Senate Report quite unambiguously declares that Congress intended to “make clear that plaintiffs need not prove a discriminatory purpose in the adoption or maintenance of the challenged practice or system in order to establish a violation.” S.Rep. 417 at 27 (emphasis added), reprinted in 1982 U.S.Code Cong. & Admin.News at 205. Moreover, far from suggesting that the presence of racial animus in the electorate was irrelevant, supporters of the 1982 legislation maintained that the amendments were necessary precisely in order to reach such “private discrimination.” See, e.g., Hearings on the Voting Rights Act Before the Subcomm. on the Constitution of the Senate Comm. of the Judiciary, 97th Cong., 2d Sess. 1367-68 (statement of Prof. Drew Days). The Association instead insists that a standard requiring § 2 plaintiffs to show that their failure to elect representatives of their choice is attributable to white bloc voting rooted in racial considerations “frustrated the goals Congress sought to achieve by repudiating the intent test of [Bolden].” Gingles, 478 U.S. at 71, 106 S.Ct. at 2777 (opinion of Brennan, J.). Given the palpable tension between “the goals Congress sought to achieve” and those it actually expressed, it is hardly surprising that the principles the Association purports to locate in the Senate Report bear only a passing resemblance to those offered by Congress. Compare Gingles, 478 U.S. at 70-73, 106 S.Ct. at 2776-78 (opinion of Brennan, J.) with S.Rep. 417 at 36-37, reprinted in 1982 U.S.Code Cong. & Admin.News at 214-15.
More importantly, the Association’s contention that an inquiry into the explanations underlying racially divergent voting patterns somehow conflicts with Congress’ abandonment of the intent requirement announced in Bolden completely ignores the fact that the Senate Report expressly adopted the standard we employ in codifying the “results” test. Indeed, like Justice Marshall in Bolden itself, see 446 U.S. at 109, 100 S.Ct. at 1522 (Marshall, J., dissenting), the Senate Report reiterated Whitcomb’s holding that “[t]he failure of the ghetto to have legislative seats in proportion to its population emerges more as a function of losing elections than of built-in bias against poor Negroes” precisely in order to show that “intent had [not] been required to prove a violation.” S.Rep. 417 at 21 (quoting Whitcomb, 403 U.S. at 153, 91 S.Ct. at 1874), reprinted in 1982 U.S.Code Cong. & Admin.News at 198. In keeping with Whitcomb, the Senate Report equated “racial bloc voting” with proof that “race is the predominant determinant of political preference.” Id. at 33, reprinted in 1982 U.S.Code Cong. & Admin.News at 211. The Association’s assertion that the test we confirm today is inconsistent with “the goals Congress sought to achieve” in amending § 2 becomes plausible only if Whitcomb is purged from our voting rights jurisprudence. It is therefore not coincidental that its brief, like Justice Brennan’s opinion, see Gingles, 478 U.S. at 61-74, 106 S.Ct. at 2772-79, fails to include a citation, let alone a discussion, of the decision Congress intended to codify.
The United States offers a second argument incorporating elements of Justice O’Connor’s as well as Justice Brennan’s opinion in Gingles. The government agrees with Justice O’Connor that an inquiry into the causes underlying polarized voting is appropriate in certain circumstances. It follows Justice Brennan, however, in maintaining that evidence tending to show that divergent voting patterns are attributable to partisan affiliation or a divergence in interests rather than race is irrelevant in assessing whether plaintiffs have established legally significant white bloc voting. We disagree with this argument as well.
The United States’ assertion that partisan affiliation cannot serve to explain voting patterns finds no support in Justice O’Connor’s opinion. The very inquiry it seeks to exclude — whether election returns track “an underlying divergence in the interests of minority and white voters,” — was the only nonracial cause expressly cited in her opinion as a possible explanation of divergent voting patterns. See Gingles, 478 U.S. at 100, 106 S.Ct. at 2792 (O’Connor, J., concurring).
The United States argues that the political differences frequently observed among white *863and minority voters are largely the product of disparities in socioeconomic status, which are themselves attributable to the presence or absence of past discrimination. In this view, a standard that would permit divergence in interest to preclude the establishment of racial bloc voting “would render meaningless the Senate Report factor that addresses the impact of low socioeconomic status on a minority group’s level of participation.” Gingles, 478 U.S. at 69, 106 S.Ct. at 2776.
This argument is not without force; it is, however, clearly foreclosed by the Senate Report. Congress was not unaware that political preference often correlates strongly with socioeconomic status; particularized needs clearly give rise to particularized interests. This observation did not, however, lead Congress to soften the line between partisan politics and racial vote dilution established by the Court in Whitcomb. To the contrary, the Senate Report not o'nly adopted Whitcomb’s holding without modification, but expressly reminded its readers in so doing that the vote dilution claim dismissed by the Whit-comb Court as “a mere euphemism for political defeat at the polls” had been brought by “black ghetto residents with [distinct] legislative interests.” S.Rep. 417 at 20, reprinted in 1982 U.S.Code Cong. & Admin.News at 198. The argument pressed here by the United States has been acknowledged, and rejected, by Congress.
The Senate factor cited by Justice Brennan in support of his refusal to attach relevance to a divergence of interests expressly relates, not to whether minority groups have been able to elect representatives of their choice, but to “the extent to which members of the minority group ... bear the effects of discrimination in areas such as education, employment, and health, which hinder their ability to participate in the political process.” S.Rep. 417 at 29 (emphasis added), reprinted in 1982 U.S.Code Cong. & Admin.News at 206. As the Court in Chisom v. Roemer confirmed, § 2 plaintiffs “must allege an abridgement of the opportunity to participate in the political process and to elect representatives of one’s choice.” — U.S. at -, 111 S.Ct. at 2365 (emphasis in original). The effects of past discrimination, as the text of the Senate Report indicates, pertain solely to the “political access” prong of a § 2 claim. It is by considering these effects in this regard, not in the bloc voting inquiry, that courts give effect to congressional intent. The United States’ approach, by contrast, would allow this single factor to assume dispositive significance in both of these inquiries. In so doing, it would permit liability, to attach, in direct conflict with the Senate Report, upon “the mere existence of underrepresentation plus a history of dual schools.” S.Rep. 417 at 34, reprinted in 1982 U.S.Code Cong. & Admin.News at 212. Electoral losses that are attributable to partisan politics do not implicate the protections of § 2.
IV. Other Legal Errors Affecting the Vote Dilution Inquiry
Defendants cite three additional legal errors that allegedly infect the district court’s findings of illegal vote dilution in each of the counties. Specifically, they argue that the district court erred in: (1) excluding elections pitting Hispanic candidates against white candidates in counties in which the evidence unmistakably showed that black and Hispanic voters were cohesive; (2) refusing to consider the paucity of minority lawyers in assessing the extent to which members of minority groups had been elected to the district court; and (3) finding that the effects of past discrimination hindered the ability of minority groups to participate in the political process despite the presence of little or no evidence suggesting that their participation was in fact depressed. We examine these issues in turn.
A. Cohesiveness of Different Minority Groups
The importance of the distinction in § 2 jurisprudence between illegal vote dilution and political defeat, between protecting racial minorities and fostering the work of political coalitions, raises the stakes for the question whether different racial or ethnic minority groups, usually blacks and Hispanics, may combine to form a single minority group within the meaning of the Voting Rights Act. *864Judges and commentators alike have questioned whether transitory unions rooted in political expedience may be properly equated with those whose source lies in the more enduring bonds supplied by a shared race or ethnicity. League of United Latin American Citizens v. Midland Indep. School District, 812 F.2d 1494, 1505-07 (5th Cir.1987) (Higginbotham, J., dissenting); Katherine I. Butler & Richard Murray, Minority Vote Dilution Suits and the Problem of Two Minority Groups: Can a ‘Rainbow Coalition’ Claim the Protection of the Voting Rights Act?, 21 Pacific L.J. 619, 641-57 (1990). Nevertheless, we have treated the issue as a question of fact, allowing aggregation of different minority groups where the evidence suggests that they are politically cohesive, see, e.g., Midland I.S.D., 812 F.2d at 1500-02, and we need not revisit this question here.
This issue is raised today in the context of the particular elections to which the district court looked as part of its inquiry into racial bloc voting. This court has consistently held that elections between white candidates are generally less probative in examining the success of minority-preferred candidates, generally on grounds that such elections do not provide minority voters with the choice of a minority candidate. See, e.g., Campos v. City of Baytown, 840 F.2d 1240, 1245 (5th Cir.1988); Citizens for a Better Gretna v. City of Gretna, 834 F.2d 496, 503 (5th Cir.1987). For this reason, courts usually focus on those elections involving black or Hispanic candidates in examining whether black or Hispanic voters enjoy an equal opportunity to elect representatives of their choice. Where blacks and Hispanics are cohesive, we have held that the relevant elections are those including either Hispanic or black candidates. See, e.g., Baytown, 840 F.2d at 1245. Defendants contend that the district court erred in refusing to consider elections pitting Hispanic and white candidates in Harris and Tarrant Counties, counties in which plaintiffs proceed on behalf of black voters only, but where the evidence indisputably showed that blacks and Hispanics were politically cohesive. In light of our precedents, we must agree.
Blacks and Hispanics have joined forces for purposes of this suit in Midland, Lubbock, and Ector Counties. In these counties, white-Hispanie elections are relevant in proving legally significant white bloc voting, for the Hispanic candidate provides the combined Hispanic-black minority with a viable minority choice. But plaintiffs contend that where they represent only black voters, white-Hispanic elections in which the Hispanic candidate received the support of black voters are irrelevant. A difference in litigation strategy cannot support this distinction. Cohesion is a fact, not a strategic card to be played at the caprice of a plaintiff. As we stated in Campos, “if the statistical evidence is that Blacks and Hispanics together vote for the Black or Hispanic candidate, then cohesion is shown.” Id. at 1245 (footnote omitted): If blacks and Hispanics vote cohesively, they are legally a single minority group, and elections with a candidate from this single minority group are elections with a viable minority candidate.
Plaintiffs next argue that there is evidence in the record that blacks and Hispanics are not politically cohesive in Harris and Tarrant Counties. They do not tell us to which evidence they refer, and understandably so. The record shows that blacks and Hispanics Were more cohesive in Harris and Tarrant Counties than in Midland and Ector Counties, counties in which plaintiffs represent both blacks and Hispanics and the district court found cohesion.
In Harris County, Taebel studied 45 elections in which he determined the percentage of black and Hispanic votes cast for the minority/winning candidate. In 35 elections the black and Hispanic vote percentages varied by less than 10%. Similarly, the levels of black and Hispanic support for the same candidate were within ten percentage points in 13 of the 17 elections studied in Tarrant County. In Midland County, by contrast, the black and Hispanic voting percentages differed by less than 10% in only 4 of the 8 elections analyzed; in Ector County, this close correlation between the preferences of Hispanic and black voters was shown in just 2 of 10 elections. Under the present law of this circuit, there is no error in the district *865court’s findings of cohesion in Midland, Ector, and Lubbock Counties, because in those counties a significant number of blacks and Hispanics usually voted for the same candidates. Gingles, 478 U.S. at 56, 106 S.Ct. at 2769. But this standard also compels the conclusion that there is also black-Hispanic cohesion in Harris and Tarrant Counties. The district court thus clearly erred in ignoring elections involving Hispanic and white candidates in these counties.29
B. Relevance of Small Number of Minority Lawyers
The absence of minority office holders is typically an important consideration in dilution cases. In this litigation, the small number of minority judges in the target counties has been the cornerstone of the plaintiffs’ proof.
The office of district judge has more eligibility requirements than the age and citizenship prerequisites of many public offices. A person must be a licensed attorney in the state of Texas for four years, and a resident of the district for two years, before becoming eligible for the post. The need for district judges to be experienced lawyers is obvious.
Undisputed evidence shows that in all of the counties, the percentage of minority lawyers was much smaller than the percentage of minority voters. In fact, minority lawyers disproportionately serve as judges, when their percentage among all eligible lawyers is considered. It is true that we have refused “to preclude vote dilution claims where few or no [minority] candidates have sought offices in the challenged electoral system.” Westwego Citizens for Better Gov’t v. City of Westwego, 872 F.2d 1201, 1208 n. 9 (5th Cir. 1989) (Westwego I). That holding is a far cry from the conclusion that the number of minority candidates eligible to run has no relevance. 'Section 2 and the Senate Report instruct us to consider the number of minority candidates elected to office. At the same time, we are instructed to evaluate the totality of the circumstances with a “ ‘functional’ view of the political process.” Gingles, 478 U.S. at 45, 106 S.Ct. at 2764. The cold reality is that few minority citizens can run for and be elected to judicial office. A functional analysis of the electoral system must recognize the impact of limited pools of eligible candidates on the number of minority judges that has resulted. See Southern Christian Leadership Conf. of Ala. v. Evans, 785 F.Supp. 1469, 1476-77 (M.D.Ala.1992).
The record discloses that at times during the 1980’s, the percentage of minority judges in five targeted counties exceeded the percentage of minority lawyers who were eligible to run for district judge. The following table summarizes the evidence.
County Minority Judges as %age of District Judges, 1988 Table IV.B Minority Lawyers as %age of Eligible Lawyers, 1989 Minority Voters as %age of Voting Age Population
Dallas 8.3 1.0 16.0 (black)
Harris 5.1 3.8 18.2 (black)
Tarrant 13.0 2.4 10.4 (black)
Bexar 26.3 11.4 41.4 (Hispanic)
Travis 7.7 2.7 14.4 (Hispanic)
Jefferson 0.0 3.1 24.6 (black)
Lubbock 0.0 5.1 21.6 (both)
Midland 0.0 3.2 19.7 (both)
Ector 0.0 4.0 21.9 (both)
*866In counties with no minority judges, the number of eligible candidates was very small. In Ector County, for example, one survey found five eligible Hispanic lawyers and only one eligible black lawyer. Apparently none of Lubbock County’s 499 lawyers in 1989 was a black attorney eligible for a district judgeship, although the State Bar reported two black lawyers in the county.
The absence of eligible candidates goes a long way in explaining the absence of minority judges. Plaintiffs cannot emphasize the scarcity of successful minority candidates to support the inference of dilution and simultaneously urge that the number of minorities eligible to run is not relevant. Plaintiffs argue that this factor may not be considered because the limited number of minority lawyers was caused by state discrimination in education. We are not persuaded this argument merits exclusion of the evidence. The Voting Rights Act responds to practices that impact voting; it is not a panacea addressing social deficiencies. See Presley v. Etowah County Comm’n, — U.S. -,-, 112 S.Ct. 820, 832, 117 L.Ed.2d 51 (1992).
C. Past Discrimination
The district court also found that Texas’ history of discrimination “touched many aspects of the lives of minorities in the Counties in question including their access to and participation in the democratic system governing this State and their socio-economic status.”30 The district court, however, did not refer to specific facts in the record to support this conclusion. Instead, the court cited a 1980 Civil Rights Commission Report describing civil rights developments in Texas during the years 1968-1978 and a 1981 district court opinion detailing race relations between minority and white residents of one of Texas’ smaller cities during the 1960’s and 1970’s.
Texas’ long history of discrimination against its black and Hispanic citizens in all areas of public life is not the subject of dispute among the parties. Nor has anyone questioned plaintiffs’ assertion that disparities between white and minority residents in several socioeconomic categories are the tragic legacies of the State’s discriminatory practices. Defendants do argue, however, that these factors, by themselves, are insufficient to support the district court’s “finding” that minorities do not enjoy equal access to the political process absent some indication that these effects of past discrimination actually hamper the ability of minorities to participate. We again agree.
It would seem tautological that a factor directing courts to determine whether past discrimination hinders a minority group’s access to the political process would require a showing that the group does not in fact participate to the same extent as other citizens. Nevertheless, prior to the amendments to § 2, this court held that evidence of decreased participation among minorities was unnecessary on grounds that “[[Inequality of access is an inference which flows from the existence of economic and educational inequalities.” Kirksey v. Board of Supervisors, 554 F.2d 139, 145 (5th Cir.1977) (en banc). This standard, however, was challenged by some of our later cases, see, e.g., McIntosh Cty. NAACP v. City of Darien, 605 F.2d 753, 759 (5th Cir.1979), and was decisively rejected by Congress in 1982. As the Senate Report stated:
The courts have recognized that disproportionate educational, employment, income level and living conditions arising from *867past discrimination tend to depress minority political participation. Where these conditions are shown, and where the level of black participation in politics is depressed, plaintiffs need not prove any further causal nexus between their disparate socio-economic status and the depressed level of political participation.
S.Rep. 417 at 29 n. 114 (emphasis added), reprinted in 1982 U.S.Code Cong. & Admin.News at 207 n. 114. As this statement discloses; the Senate Report, while not insisting upon a causal nexus between socioeconomic status and depressed participation, clearly did not dispense with proof that participation in the political process is in fact depressed among minority citizens. In apparently holding that socioeconomic disparities and a history of discrimination, without more, sufficed to establish these Zimmer factors, the district court employed the wrong legal standard.
Nor do we believe that the record before us can support such a finding under the proper test. Plaintiffs have offered no evidence of reduced levels of black voter registration, lower turnout among black voters, or any other factor tending to show that past discrimination has affected their ability to participate in the political process. While there are indications that Hispanic citizens register to vote at a lower rate than white and black citizens, this data provides support for such a finding in only Bexar and Travis Counties, where plaintiffs proceed on behalf of Hispanic voters only.
Plaintiffs contend that the district court could have relied on the opinion offered by Dr. Brischetto, who, during his testimony regarding Bexar County, stated:
Well, certainly having less of these socioeconomic resources or characteristics to draw on, we find that minority voters will participate less in the electoral system. Education is an important resource. For example, it enables people to feel like they are more a part of and take part in the election system to a greater extent. Lacking that they participate less. So it is important, it has an effect certainly on their participation when they are subordinate status in the stratification system.
Brischetto’s statement, as its tone suggests, was not so much a finding as a prediction or hypothesis about what one might expect to find among minorities who still bore the sears of past discrimination. It is for this reason that he could claim that his testimony regarding the participation of Hispanics in Bexar applied with equal force to all of the other counties. In fact, the nature and basis of his opinion became explicit as the testimony shifted to these other locales. In Travis County, for example, he stated only that “stratification ... may very well also be an indication of the fact that Hispanics are less likely to participate fully and effectively in the electoral system in Travis County.” In Lubbock, Brischetto stated only that “I think [such stratification] is an indication that minorities' are less equipped with those resources that they need to participate fully in the political system.” Finally, he testified in the context of Tarrant County that socioeconomic differences “indicate! ] that minorities may have a diminished ability to participate fully in the electoral system because of their lower status and stratification that exists in that community.”
Brischetto’s testimony thus provides support for the common sense proposition that depressed political participation typically accompanies poverty and a lack of education; it certainly does not amount to proof that minority voters in this case failed to participate equally in the political processes! A district court’s findings under § 2 must rest on an “intensely local appraisal” of the social and political climate of the cities and counties in which such suits are brought, White, 412 U.S. at 769, 93 S.Ct. at 2341, not the sort of generalized armchair speculation supplied by Dr. Brischetto. We need evidence, not musings.
Plaintiffs also contend that minority citizens’ lack of financial resources makes it very difficult for minority-preferred candidates to secure funds sufficient to run creditable county-wide campaigns. Here again, the inference plaintiffs ask us to draw might well be true in most cases; regardless of its general validity, however, it is no substitute for proof that a minority group’s poverty has had the predicted effect in this particular *868case. The evidence presented at trial simply does not show that past discrimination has inhibited the ability of minorities to participate in the process. In fact, the record discloses that minority-preferred candidates frequently raised and spent more money than their white opponents.
Witnesses Coronado and Fitch did testify that minority candidates generally were unable to raise the money necessary to run county-wide. When asked about the only district court campaign in which he was personally involved, however, Coronado made no mention of money problems. In fact, he testified that “[Judge Gallardo] ran a very good campaign. I mean he was, he understood the media, had people out working boxes, he had a lot of attorneys of all ethnic groups working in his campaign, a broad base campaign in the community.” Similarly, Fitch asserted that black incumbents had difficulty raising funds, but she attributed this difficulty to “racial discrimination” and black candidates’ “past record of losing.”
In contrast with the highly equivocal testimony of Fitch and Coronado concerning their impressions of the barriers facing minority candidates, nearly all. such candidates who appeared at trial reported that they had outspent their white opponents, often by a very large amount. In Midland County, for example, Watson testified that she outspent her white opponent in the general election for Justice of the Peace by a factor of six. In Dallas County, Joan Winn White, Tinsley, H. Ron White, and Oliver all testified that they had run extensive, well-financed campaigns. In particular, Oliver stated that he spent $300,000 in a losing effort. The same was true of minority-preferred candidates in Harris County. Lee testified that she outspent her white opponent at a rate approaching twelve to one; Berry stated that the ratio in his campaign for district court was even greater. Finally, Leal testified that he raised $85,000 to $90,000 to his opponent’s $1,000. A district court’s findings may only rest on the evidence presented at trial. The record before us does not remotely suggest that the visible scars of discrimination have left minority-preferred candidates and their supporters within minority communities without the funds needed to launch broad-based, county-wide campaigns. In,fact, the available evidence shows just the opposite. For this reason, we must conclude that plaintiffs have not established that the effects of past discrimination have hindered their ability to participate in the political process.
V. Texas’ Linkage Interest
This case involves 172 judicial districts that coincide with nine Texas counties. Given the State of Texas’ county-based system of venue, this venerable structure links the jurisdictional and electoral bases of the district courts. In doing so, the structure advances the state’s substantial interest in judicial effectiveness. Trial judges are elected by a broad range of local citizens, rather than by a narrow constituency. This electoral scheme balances accountability and judicial independence.
As explained in detail below, the state’s interest in maintaining the structure of this single-member judicial office must be weighed in the totality of circumstances to determine whether a § 2 violation exists. The weight of a substantial state interest, determined as a matter of law, is balanced against localized evidence of racial vote dilution. This substantial state interest may be overcome only by evidence that amounts to substantial proof of racial dilution. Otherwise, the at-large election of district court judges does not violate § 2.
A. The Structure of Texas District Courts
The district courts are the primary trial courts in Texas. District judges were first elected in 1850, five years after statehood, and every state constitution since 1861 has provided for their election by county residents. All voters of the entire county elect all the district judges of their county. The political boundaries of each county are the boundaries of the jurisdiction and election base in all of the challenged counties.31 Many counties in Texas have more than one district judge. Even so, trials are presided over by district judges acting alone. The *869only collegial decision-making by district judges in counties with more than one district judge is in the handling of some administrative matters. In some of the counties involved here, district courts are designated to specialize in civil, criminal, or family law cases.
The electoral bases of district judges are linked to the area over which they exercise primary jurisdiction. This linkage has been in place throughout the 143 year history of judicial elections in Texas. By making coterminous the electoral and jurisdictional bases of trial courts, Texas advances the effectiveness of its courts by balancing the virtues of accountability with the need for independence. The state attempts to maintain the fact and appearance of judicial fairness that are central to the judicial task, in part, by insuring that judges remain accountable to the range of people within their jurisdiction. A broad base diminishes the semblance of bias and favoritism towards the parochial interests of a narrow constituency. Appearances are critical because “the very perception of impropriety and unfairness undermines the moral authority of the courts.” John L. Hill, Jr., Taking Texas Judges Out of Politics: An Argument for Merit Election, 40 Baylor L.Rev. 339, 364 (1988). The fear of mixing ward politics and state trial courts of general jurisdiction is widely held. It is not surprising then that states that elect trial judges overwhelmingly share this structure and electoral scheme. See infra note 30. The systemic incentives of subdistrieting are those of ward politics, and would “diminish the appearance if not fact of its judicial independence — a core element of a judicial office.” LULAC II, 914 F.2d at 650 (Higginbotham,' J., concurring).
B. The Role of Function Under § 2
In Houston Lawyers’ Association v. Attorney General, — U.S.-, 111 S.Ct. 2376, 115 L.Ed.2d 379 (1991), the Supreme Court agreed that the interests behind the existing court structure must be considered.
[W]e believe that the State’s interest in maintaining an electoral system — in this case, Texas’ interest in maintaining the link between a district judge’s jurisdiction and the area of residency of his or her voters — is a legitimate factor to be considered by courts among the “totality of circumstances” in determining whether a § 2 violation has occurred. A State’s justification for its electoral system is a proper factor for the courts to assess in a racial vote dilution inquiry____ Because the State’s interest in maintaining an at-large, district-wide electoral scheme for single-member offices is merely one factor to be considered in evaluating the “totality of circumstances,” that interest does not automatically, and in every case, outweigh proof of racial vote dilution.
Id. at -, 111 S.Ct. at 2381.
Justice Stevens noted that Texas’ interest in linking electoral and jurisdictional bases is “a legitimate factor to be considered by courts among the ‘totality of circumstances’ in determining whether a § 2 violation has occurred.” Id. The Court was not persuaded that this “linkage” interest should defeat liability “automatically, and in every case.” Rather, Houston Lawyers’ held that the interest must be weighed against other relevant factors to ascertain whether the interest “outweigh[s] proof of racial vote dilution.” Id. See also Nipper v. Chiles, 795 F.Supp. 1525, 1548 (M.D.Fla.1992) (holding that “a state’s interest in maintaining an electoral system is a legitimate factor to be considered ... in the liability phase of a section two case”).
An examination of Houston Lawyers’ further illuminates why the state interests behind an office’s structure and function must be weighed. The Court held that single-member office elections are within the scope of § 2. Houston Lawyers’, — U.S. at -, 111 S.Ct. at 2380. This holding reached beyond judicial elections. “[T]he coverage of the Act encompasses the election of executive officers and trial judges whose responsibilities are exercised independently in an area coextensive with the districts from which they are elected.” Id. (emphasis added). It appears from this language that an office such as mayor or sheriff is subject to § 2 scrutiny, requiring an analysis of the totality of circumstances to determine whether illegal vote dilution exists. While that analysis is *870not precluded, it must take into account the state interests that are furthered by the structure and function of such single-member offices. Surely by enacting the Voting Rights Act, Congress did not contemplate that the office of mayor in a city would have to be dismantled because its single-member office nature submerged minority voters in the community of voters as a whole, without regard for the interests in preserving that office. Cf. Butts v. City of New York, 779 F.2d 141 (2d Cir.1985) (holding that primary runoffs for single-member offices of mayor, city council president, and city comptroller do not violate § 2).
Therefore, while the Supreme Court rejected the contention that the linkage interest in all cases defeated liability under § 2, the Court endorsed the position that the linkage interest is relevant to a determination of liability. Indeed, by noting that the linkage interest does not “automatically, and in every case, outweigh proof of racial vote dilution,” the Court held that the state interest could outweigh what would otherwise be proof of illegal dilution and thus foreclose liability. As one commentator has noted:
the Court recognized that in balancing the many factors in the totality of the circumstances test, the state interest in district wide judicial elections may, in some cases, outweigh proof of racial voter dilution.
Mary T. Wickham, Note, Mapping the Morass: Application of Section 2 of the Voting Rights Act to Judicial Elections, 33 Wm. & Mary L.Rev. 1251, 1285 (1992).
The issue we face is determining when the linkage interest will outweigh other factors and defeat liability under § 2. In resolving this issue, we reject the polar extremes of the parties. The State of Texas maintains that the linkage interest must defeat liability in every ease, regardless of the other circumstances in the totality. The Supreme Court rejected this position when it held that the linkage interest does not “automatically, and in every case, outweigh proof of racial vote dilution.” Houston Lawyers’, — U.S. at -, 111 S.Ct. at 2381.
We also reject the position of plaintiffs that the linkage interest can never defeat liability under the totality of circumstances if “illegal” dilution is otherwise established. The plaintiffs maintain that only the absence of a compelling state interest in an electoral scheme is relevant to liability, and that such an absence “is an optional factor” that plaintiffs can use to support a finding of illegal dilution. They contend, however, that the existence of a compelling interest can never defeat liability that is otherwise established under the totality of the circumstances. This position is foreclosed by the Supreme Court, which directed that this state interest is to be weighed as part of the totality of the circumstances. Id.
Citing Jones v. City of Lubbock, 727 F.2d 364, 383 (5th Cir.1984), and United States v. Marengo County Comm’n, 731 F.2d 1546, 1571 (11th Cir.1984), plaintiffs urge that the Zimmer factor of a non-tenuous state policy is among the least important of the factors for determining dilution. These decisions state only that defendants cannot defeat liability by using the non-tenuous policy justification of an electoral scheme to prove that scheme “does not have a discriminatory intent.” Marengo County, 731 F.2d at 1571. See also Terrazas v. Clements, 581 F.Supp. 1319, 1345 n. 24 (N.D.Tex.1983) (three-judge panel) (“In the case of tenuousness, the lesser weight is consistent with the change in emphasis from intent to results. The principal probative weight of a tenuous state policy is its propensity to show pretext.”).
The plaintiffs’ argument misses the point. The State of Texas has done more than assert that its interest in this electoral scheme is not tenuous — that is, not a pretext masking discriminatory intent in the adoption or maintenance of the scheme. The interest in linking electoral to jurisdictional base takes on additional and distinct relevance because it advances objectively substantive goals. The inquiry into whether an interest is substantial goes beyond inquiring whether the interest is non-tenuous. A substantial state interest must be more than racially-neutral. Thus, the linkage interest is not examined just because it proves that the state’s practice is premised on a racially-neutral policy and is consistently applied. Cf. S.Rep. 417 at 29 n. 117, reprinted in 1982 U.S.Code Cong. & Admin.News at 207 n. 117.
*871Proof of a merely non-tenuous state interest discounts one Zimmer factor, but cannot defeat liability. It does not follow, however, that proof of a substantial state interest cannot defeat liability. The totality of circumstances inquiry that occurs after a showing of the Gingles prerequisites is not limited to factors listed in the legislative history of the Voting Rights Act. Gingles, 478 U.S. at 45, 106 S.Ct. at 2763; Westwego Citizens for Better Gov’t v. City of Westwego, 946 F.2d 1109, 1120 (5th Cir.1991) (Westwego III). The weight, as well as tenuousness, of the state’s interest is a legitimate factor in analyzing the totality of circumstances. As we have explained, the Voting Rights Act largely codifies Fourteenth Amendment jurisprudence embodied in White v. Regester, 412 U.S. 755, 93 S.Ct. 2332, 37 L.Ed.2d 314 (1973). See Jones, 727 F.2d at 379-80. The substantiality of the state’s interest has long been the centerpiece of the inquiry into the interpretation of the Civil War Amendments and their interplay with the civil rights statutes.
Having rejected the proffered extremes— that the linkage interest either always or never defeats § 2 liability — we turn to when the linkage interest precludes a § 2 violation. This question depends upon the weight of the interest.
C. Weight of State’s Interest Is Matter of Law
The plaintiffs urge that the weight or substantiality of Texas’ linkage interest is an issue of fact for the district court to decide in the first instance, reviewable only for clear error. We disagree. Deciding whether the adoption or maintenance of a system is a pretext for racial discrimination may present a question of fact.32 This question can turn on credibility, an issue best determined by a fact finder. The issue of substantiality, however, is distinct from the conventional Zimmer factor of tenuousness and is a legal determination.
The Supreme Court has held that the finding of dilution is a factual matter reviewable only for clear error. Gingles, 478 U.S. at 78, 106 S.Ct. at 2780-81. A substantial state interest is not inherently preclusive of dilution and is not raised to disprove the existence of dilution. Rather, the state’s interest is weighed against proven dilution to assess whether such dilution creates § 2 liability. Houston Lawyers’, — U.S. at -, 111 S.Ct. at 2381 (weighing of linkage interest on remand goes to determination of whether interests “outweigh proof of racial vote dilution”).
Determining the substantiality of Texas’ linkage interest under the Voting Rights Act, a statute enacted to enforce the guarantees of the Civil War Amendments, is analogous to weighing the asserted state interest in constitutional law contexts. With issues of substantive due process, equal protection, and the First Amendment, the weight of a state’s interest has always been a legal question, not a factual one. For example, in Posadas de Puerto Rico Ass’n v. Tourism Co. of Puerto Rico, 478 U.S. 328, 341, 106 S.Ct. 2968, 2977, 92 L.Ed.2d 266 (1986), the Court had “no difficulty in concluding that the Puerto Rico Legislature’s interests in the health, safety, and welfare of its citizens constitutes a ‘substantial’ governmental interest.” In reaching this conclusion, the Court itself determined the weight of the state interest. See also City of Cleburne v. Cleburne Living Center, 473 U.S. 432, 105 S.Ct. 3249, 87 L.Ed.2d 313 (1985) (weighing state’s interest de novo). We hold that the substan-tiality of Texas’ interest under § 2 is a question of law for this court to determine de novo and not a question of fact that somehow will be described on a county-by-county basis.
D. Determining the Weight of the Linkage Interest
The weight of Texas’ interest is virtually assigned by a Supreme Court decision handed down on the same day as Houston Lawyers’. In Gregory v. Ashcroft, — U.S.-, -, 111 S.Ct. 2395, 2404, 115 L.Ed.2d 410 (1991), the Supreme Court held that the Age *872Discrimination in Employment Act does not apply to judicial offices in Missouri. The plaintiffs had used ADEA to challenge a mandatory retirement age for state judges. The Court noted that “the authority of the people of the States to determine the qualifications of their most important government officials ... lies at the heart of representative government.” Id. at -, 111 S.Ct. at 2402 (internal quotation omitted). Gregory noted that “the States’ power to define the qualifications of their officeholders has force even as against the proscriptions of the Fourteenth Amendment.” Id. at -, 111 S.Ct. at 2405. To protect this power to define the judicial office, Gregory required a clear statement from Congress for an override of qualifications imposed by the State for important state government office. Id. at -, 111 S.Ct. at 2406. This requirement exists even if ADEA was based upon Congress’ powers under the Fourteenth Amendment, rather than the Commerce Clause. Id. at -, 111 S.Ct. at 2405.
“The people of Missouri have a legitimate, indeed compelling, interest in maintaining a judiciary fully capable of performing the demanding tasks that judges must perform.” Id. at-, 111 S.Ct. at 2407. If that interest is compelling, the people of Texas have at least a substantial interest in defining the structure and qualifications of their judiciary. Indeed, Texas’ Attorney General has submitted to this court that linkage is a “fundamental right” that “serves [a] compelling interest” of the State of Texas. Linking electoral and jurisdictional bases is a key component of the effort to define the office of district judge. That Texas’ interest in the linkage of electoral and jurisdictional bases is substantial cannot then be gainsaid.
Our confidence in this conclusion is bolstered by the recognition and pursuit of the linkage interest in other states. Courts have recognized the legitimacy and substance of similar linkage interests in Florida and Alabama. See Nipper v. Chiles, 795 F.Supp. 1525, 1548 (M.D.Fla.1992); Southern Christian Leadership Conf. of Ala. v. Evans, 785 F.Supp. 1469, 1478 (M.D.Ala.1992). Of the twenty-nine states that elect their principal trial court judges, including Texas, Alabama, and Florida, twenty-five employ district-wide elections.33 The overwhelming preservation of linkage in states that elect their trial court judges demonstrates that district-wide elections are integral to the judicial office and not simply another electoral alternative.
The decision to make jurisdiction and electoral bases coterminous is more than a decision about how to elect state judges. It is a decision of what constitutes a state court judge. Such a decision is as much a decision about the structure of the judicial office as the office’s explicit qualifications such as bar membership or the age of judges. The collective voice of generations by their unswerving adherence to the principle of linkage through times of extraordinary growth and change speaks to us with power. Tradition, of course, does not make right of wrong, but we must be cautious when asked to embrace a new revelation that right has so long been wrong. There is no evidence that linkage was created and consistently maintained to stifle minority votes. Tradition speaks to us about its defining role — imparting its deep running sense that this is what judging is about.
On the other hand, plaintiffs’ interests are not well-served by destroying linkage. The inescapable truth is that the result sought by plaintiffs here would diminish minority influ*873ence. Minority voters would be marginalized, having virtually no impact on most district court elections. Given that district judges act alone in exercising their power, that use of the Voting Rights Act is perverse. After subdistricting, a handful of judges would be elected from subdistricts with a majority of minority voters. Creating “safe” districts would leave all but a few subdis-tricts stripped of nearly all minority members. The great majority of judges would be elected entirely by white voters. Minority litigants would not necessarily have their cases assigned to one of the few judges elected by minority voters. Rather, the overwhelming probability would be that the minority litigant would appear “before a judge who has little direct political interest in being responsive to minority concerns.” LULAC II, 914 F.2d at 650 (Higginbotham, J., concurring). Under the totality of circumstances, we must recognize that breaking the link between the electoral base and the jurisdiction of this single-member office would perversely lessen minority influence on the conduct of most litigation.
The distrust of judicial subdistricts does not rest on paternalism. It recognizes Texas’ historic interest in having district judges remain accountable to all voters in their district. Regardless of the race or residency of particular litigants, judges make choices that affect all county residents. Texas has insisted that trial judges answer to all county voters at the ballot box. Unlike legislators or even appellate judges, who make decisions in groups, each district judge holds a single-member office and acts alone. When collegial bodies are involved, all citizens continue to elect at least one person involved in making a particular decision. While subdistricting for multimember offices can enhance minority influence because members from minority subdistricts participate in and influence all of the decisions of the larger body, subdistrict-ing for single-member district court judge-ships would leave minority voters with no electoral influence over the majority of judges in each county. Subdistricting would partially disenfranchise citizens to whom all district judges in a county are now accountable.
By contrast, under the present regime, minority voters participate in all judicial elections in each county. This participation gives minority voters the opportunity to influence all elections, absent significant racial vote dilution. As Justice O’Connor noted in her concurring opinion in Gingles, voters can wield influence over elections even when those votes are cast for losing candidates. Gingles, 478 U.S. at 98-99, 106 S.Ct. at 2791 (O’Connor, J., concurring). Denying importance to this ability to influence asks that all measures of success be found in the win-loss column. This mandates proportional representation as the measure of dilution, contrary to the explicit terms of § 2. Indisputably, subdistricting would assure the absence of minority influence over the judicial process. See LULAC II, 914 F.2d at 649-50 (Higginbotham, J.,. concurring); Southern Christian Leadership Conf. of Ala. v. Evans, 785 F.Supp. 1469, 1478 (M.D.Ala.1992) (Hobbs, J.) (by subdistricting judicial positions, “black voters ... will ... be sacrificing [an] extremely valuable political right — the right to vote for all of the judges who will be serving as judges in the circuit wherein they live”).
Plaintiffs contend that linking jurisdictional and electoral bases does not, in fact, protect these uniquely judicial interests. All of the plaintiffs’ arguments reduce to the single contention that Texas does not consistently apply the policy of linking jurisdictional and electoral bases.
Before addressing these arguments, we note that in assessing the relationship between the end pursued and the means employed, “our scrutiny will not be so demanding where we deal with matters resting firmly within a State’s constitutional prerogatives.” Sugarman v. Dougall, 413 U.S. 634, 648, 93 S.Ct. 2842, 2850, 37 L.Ed.2d 853 (1973). As both Sugarman and Gregory make clear, such matters include “the establishment and operation of its own government, as well as the qualifications of an appropriately designated class of public office holders.” Sugarman, 413 U.S. at 648, 93 S.Ct. at 2851. Examining Texas’ linking of electoral and jurisdictional bases in light of *874these considerations, we find that it serves the substantial interests we described.
The plaintiffs contend that Texas district judges often adjudicate controversies involving litigants who are not residents of the county. Defendants make several responses. First, they maintain that because a district judge’s area of primary jurisdiction is defined by county-specific venue rules, most residents of a county will have their disputes adjudicated by judges they elect. Second, the residency of particular litigants is not that important. Regardless of the identity of litigants in a case, a district judge may make decisions or grant relief that impact primarily upon the residents of the district.
By drawing attention to venue, plaintiffs only remind us of concerns unique to the district judge’s office. Venue rules preserve judicial fairness by preventing forum-shopping and diminishing the chances of biased adjudication. At the same time, the rules keep most local matters in local courts, where local juries are drawn and judges are accountable to voters for the legal and policy choices they make.34 The localized focus of district courts is particularly evident in criminal matters, where venue is based on events related to the offense. Tex.Code Crim.Proc. Ann. eh. 13 (Vernon 1977). Domicile and hence convenience to the defendant have never been a consideration. Since the propriety of venue goes to the authority of the court, it is “quasi-jurisdictional in nature.” George E. Dix, Texas Charging Instrument Law: The 1985 Revisions and the Continuing Need for Reform, 38 Baylor L.Rev. 1, 71 (1986). Thus, Texas law commonly refers to the district court’s venue reach as its jurisdiction in criminal matters. See, e.g., Hodge v. State, 527 S.W.2d 289, 292 (Tex.Crim.App.1975); Tex.Code Crim.Proc.Ann.Arts. 21.-02(5), 21.21(5) (Vernon 1989).
Similarly, family law matters will almost always be handled by the local district court. See, e.g., Tex.Fam.Code Ann. §§ 3.21, 11.04 (Vernon 1986) (concerning venue in divorce and parent-child relationship suits). Quintessentially local matters such as suits against counties or disputes involving title to real property must be tried in the district court of the same county. Tex.Civ.Prac. & Rem.Code §§ 15.001, 15.015 (Vernon 1986). Whatever the area of practice — whether civil, criminal, or family law — the conclusion reached in the concurring opinion in LULAC II remains valid. “[T]he state recognized that elimination of [the] risk and appearance of bias was essential to the office it was creating by an elaborate set of rules controlling venue.” 914 F.2d at 651 (Higginbotham, J., concurring). The argument that Texas’ venue rules somehow abrogate its interest in linking jurisdiction and electoral bases is illusory.
Plaintiffs also challenge the legitimacy of the state interest in linkage by pointing to the use of visiting, judges in the district courts. Judges not elected by a district’s residents — e.g., judges from another district or retired judges — may be temporarily assigned to a district court, when necessary to dispose of its accumulated business, by the Chief Justice or regional presiding judge. Tex.Gov’t Code Ann. § 74.052 (Vernon 1988).35 A typical occasion for such assignments is when the district judge is vacationing or ill. Plaintiffs have not demonstrated that this measure of expedience represents an abandonment of the interests behind linkage. To the contrary, insofar as linkage involves the appearance of judicial fairness and independence, visiting judges are not inconsistent with its purposes. Because visiting judges will not stand for reelection, they do not create the impression of bias that may accompany a judge elected from a narrow constituency.
*875Another challenge to the legitimacy of the linkage interest is based upon Articlé 5, § 7a of the Texas Constitution. Plaintiffs reason that Texas abandoned its linkage interest by allowing the residents of counties to “opt out” of the linkage structure by selecting judges from regions smaller than a county. This contention is without merit. As Chief Justice Phillips explained at trial, § 7a was enacted in 1985 as part of a constitutional and statutory scheme designed to equalize court dockets by allowing the realignment of judicial districts. The provision states that a district smaller than a county may not be created unless approved by a majority of county voters. Tex. Const. Art. 5, § 7a(i). The people of Texas have jealously reserved to themselves, as individual voters, the power to subdivide districts that have always been the size of a county, or larger. Nowhere in Texas’ 254 counties have residents voted to break the link between jurisdiction and electoral base. If anything, § 7a(i)’s unemployment testifies to affirmation, not abrogation, of the interest in linkage.
Moreover, even if one county were to subdivide, the interest in linkage would not be lost in the state as a whole. In Mahan v. Howell, 410 U.S. 315, 93 S.Ct. 979, 35 L.Ed.2d 320, modified 411 U.S. 922, 93 S.Ct. 1475, 36 L.Ed.2d 316 (1973), the Supreme Court recognized that although Virginia divided one county when reapportioning its state legislature, it retained its interest in preserving boundaries of all other political subdivisions. Id. at 327, 93 S.Ct. at 986. Texas’ interest in preserving the structure of its judiciary by linking jurisdictional and electoral boundaries is greater than a state’s interest in observing boundaries in legislative reapportionment, because it serves substantive purposes other than convenience.
Finally, plaintiffs note that some rural district judges, and urban justices of the peace, are elected by a small number of voters. Therefore, plaintiffs contend, judges elected by narrow constituencies are not antithetical to the state’s interest in judicial independence. The structure of the justice courts have no bearing on Texas’ interests in maintaining its system of district courts. Justices of the peace need not be lawyers, and preside over courts whose subject matter jurisdiction is limited to less significant disputes. For instance, the justice court’s criminal jurisdiction is limited to finable misdemeanors. Significantly, the justice of the peace “is powerless to issue injunctions.” Bowles v. Angelo, 188 S.W.2d 691, 693 (Tex.Civ.App. — Galveston 1945, no writ). The justice court is not a court of record, so when its rulings are appealed, the cases are tried de novo before a county court judge — a judge chosen by district-wide election. In great contrast, district courts are Texas’ trial courts of general jurisdiction, charged with trying felony cases and civil matters of unlimited amounts in controversy. As to rural Texas, linkage is preserved, while providing as broad a range of constituents as the countervailing problems of courthouse proximity allow.
E. Other Means to Accommodate the Linkage Interest
Plaintiffs urge that the linkage interest can be accommodated even if the existing scheme were found to be illegal. They offer two alternatives: either a complete overhaul of the existing venue scheme or the use of unconventional electoral methods that preserve at-large voting. The plaintiffs suggest that a scheme of single-member districts may preserve linkage, by making each district judge’s area of primary jurisdiction co-extensive with the single-member district from which the judge is elected. Plaintiffs provide no evidence that such a radical reworking of the venue of Texas courts would be administratively feasible. The district court likewise simply asserted that such an arrangement of venue limited to a single-member district could accommodate Texas’ interests, without a glance at the feasibility of such an arrangement. One look at Harris County cut into a grid of dozens of venue blocks is enough to show the bizarre nature of this proposal.
We cannot conclude that Texas’ interests could be adequately accommodated by such a radical reworking of Texas’ venue rules. The proposal illustrates how different the judicial offices’ at-large election scheme is from legislative and executive at-large elections. Plaintiffs must propose not only changing the means by which Texas’ district *876judges are selected, but also its system of venue, perhaps of case assignment procedures, and maybe even its jury selection methods. The necessity for such proposals is a powerful testament to the reality that linkage is an essential part of the structure of the judicial office, much more than the method of electing the office holder.
The plaintiffs also contend that the linkage interest deserves little weight because it might be accommodated by remedies other than subdistricting. In particular, plaintiffs point to the use of limited voting or cumulative voting. The Supreme Court, of course, “strongly preferfs] single-member districts for federal court-ordered reapportionment.” Growe v. Emison, — U.S. -,-, 113 S.Ct. 1075, 1084, 122 L.Ed.2d 388 (1993). In any event, we do not agree that this argument undermines the substantiality of the state’s interest.
The allegedly illegal facet of the existing electoral scheme is that it employs at-large elections. Both plaintiffs’ amended complaint and plaintiff-intervenors’ complaint-in-intervention assert that the existing “at large scheme” violates § 2, and pray for a court order “that district judges in the targeted counties be elected in a system which contains single member districts.” By employing at-large elections, the people of Texas have linked the electoral and jurisdictional base of the district judge.
Limited and cumulative voting are election mechanisms that preserve at-large elections. Thus, they are not “remedies” for the particular structural problem that the plaintiffs have chosen to attáck. At trial, plaintiffs attempted to prove the three Gingles prerequisites. This test establishes “that the minority has the potential to elect a representative of its own choice in some single-member district” and “that the challenged districting thwarts a distinctive minority by submerging it in a larger white voting population.” Growe, — U.S. at —, 113 S.Ct. at 1084. Plaintiffs then tried to supplement that evidence with proof of Zimmer factors, such as past discrimination and anti-single shot voting rules. The question presented by this lawsuit is whether Texas’ at-large election of district judges violates § 2. To answer that question, we must determine the weight of the state’s linkage interest. We will not discount that interest based upon purported remedies that preserve the challenged at-large scheme. Plaintiffs cannot attack at-large voting as a violation of § 2, and then ignore the special characteristics of the judicial office by insisting that they will embrace a remedy that preserves that scheme. To do so would completely shunt consideration of the interest to the remedy stage, contrary to Houston Lawyers’.
F. Balancing the State’s Interest
In finding that Texas’ interest is substantial, we recognize that it will not always defeat § 2 liability. Substantiality is not quantifiable, and we translate its force in the practical world of trials to the burden required to overcome it. As we see it, plaintiffs cannot overcome a substantial state interest by proving insubstantial dilution. We hold that proof of dilution, considering the totality of the circumstances, must be substantial in order to overcome the state’s interest in linkage established here. As a matter of law, Texas’ interest cannot be overridden by evidence that sums to a marginal ease. It will take more to create a fact issue for trial. We must examine the circumstances in each county accordingly.
We do not now attempt to define in detail what sort of proof of dilution would be substantial enough to override the state’s linkage interest. We do not change the nature or usual means of proof. The Gingles prerequisites and Zimmer factors remain. Two facts are especially relevant to assessing the substantiality of the plaintiffs’ proof of dilution. One is the willingness of the racial or ethnic majority — in this case, white voters— to give their votes to minority candidates. The other critical fact is the ability of minority voters to elect candidates of their choice even when opposed by most voters from the majority. Among the Zimmer factors, proof of racial appeals in elections, non-responsiveness of elected officials to minority voters, and persistent lack of electoral success by minority candidates are most important.
*877VI. Application of Law to Each County
We now turn to the application of these principles of law in each county. As we have explained, the district court’s findings of dilution are infected by erroneous legal principles. Findings that rest upon erroneous views of the law must be set aside. Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 1789, 72 L.Ed.2d 66 (1982). Remanding for reconsideration is inappropriate where “the record permits only one resolution of the factual issue.” Id. at 292, 102 S.Ct. at 1792. Here, under controlling law, the evidence will not support the findings of liability. Our analysis is common to all counties but takes us along different routes in different counties. In the six counties of Dallas, Tarrant, Travis, Midland, Ector, and Lubbock, we hold that the district court clearly prred in finding vote dilution. Even if the district court were correct, the evidence would be outweighed by the State’s substantial interest in linkage. Finally, partisan voting at the least so weakens the proof of dilution that it loses in the weighing of the totality of the circumstances. Even if we assume that some dilution may be inferred, in the three remaining counties the evidence does not outweigh the State’s interest in the totality of the circumstances.
One thread runs throughout the plaintiffs’ case in all of the counties — an insubstantiality of proof that the minority-preferred candidate lost “on account of race.” Except in Dallas County, the district court’s finding of dilution rested on three recitations: (1) the Gingles prerequisites; (2) an invocation of a general history of discrimination; and (3) the number of minority judges was not proportional to the general minority population. The size of some counties and the absence of single shot voting were seen as “enhancing” Zimmer factors. The district court also found two instances of racial appeals in Dallas County, one in a judicial race won by the minority with white support.
A. Dallas County
Plaintiffs proceed on behalf of black voters in Dallas County. The voting age population of Dallas County is 1,106,757. Of this number, 180,294 (16.3%) are black. There were thirty-six different district courts in Dallas County. Until 1987, none of the judges of these districts were black. In 1987 and 1988, three of the district judges, or 8.3% of the total, were black. In 1989, there were two black district judges in Dallas County, 5.5% of the total. Today, five of the thirty-seven36 district judges in Dallas (13.5%) are black. County court judges are also elected at-large in Dallas County; two of those judges are black. Undisputed expert testimony and surveys showed that less than 2.0% of the lawyers in Dallas County are both eligible to serve as district judges and black.
The evidence in Dallas County clearly establishes that judicial elections are decided on the basis of partisan voting patterns. We are left with the inescapable conclusion that plaintiffs have failed to prove that minority-preferred judicial candidates in this county are consistently defeated by racial bloc voting. This is a failure to meet the threshold showing required by Gingles.
Uncontroverted evidence demonstrates that 99% of black Dallas voters support the Democratic candidate in every judicial election. The evidence also indicates that the majority of white voters always voted for the Republican, and thus for candidates other than the black-preferred Democratic candidate. As a result of these voting patterns, the black-preferred Democratic candidate always lost in judicial elections, regardless of the year of the election in Dallas' County.-' The Republican Party dominated every analyzed judicial race. Defendants understandably contend that the defeat of black-preferred candidates is the result of party affiliation rather than racial considerations. According to defendants, elections are determined by straight-party voting in which voters support their party’s ticket regardless of the race of the candidates. The undisputed facts overwhelmingly support this contention. Plaintiffs offer the only answer they have— the evidence is not legally relevant.
Drs. Engstrom and Taebel, plaintiffs’ and defendants’ experts respectively, analyzed seven district court general elections with *878black candidates. The following table summarizes the experts’ analyses, in particular the amount of support from white voters received by each candidate. The first figure represents the “non-black vote” as estimated by plaintiffs’ expert. The range is defined by the homogeneous precinct and bi-variate regression analysis performed by Engstrom. The second figure, in parentheses, represents the white vote as estimated by Taebel. Tae-bel did not analyze the 1984 Tinsley-Maloney race.
Table VI.A
Year Court Candidates Race Party Non-black (white) vote
1980 191st District Winn Howell B W D R 38.6-39.7% (36%)
1984 Crim Dist Ct 2 Baraka Metcalfe B W R D 60.6-61.8% (61%)
" Crim Dist Ct 4 Tinsley Maloney B W D R 28.7-30%
" 301st District White O’Donnell B W D R 30.6-31.9% (31%)
1986 195th District Tinsley Kendall B W D R 36.6-37.5% (31%)
" 256th District Wright Brin B W R D 70.6-71.7% (77%)
1988 95th District Oliver Brown B W D R 36.9-37.9% (38%)
Roughly 61%-77% of white voters consistently supported Republicans, even when black Republicans ran against white Democrats. Virtually all black voters supported the Democratic candidate, even when the Democratic candidate was white, running against black Republicans.
Black Republicans won in two of the seven analyzed district court races. According to Taebel’s study, one of these Republicans, Carolyn Wright, did better among white voters than any other Republican, white or black, winning 77% of the white vote. Other black Republicans received percentages of the white vote comparable to those received by white Republicans. Judge Baraka, the other black Republican district court candidate, took about 61% of the white vote against a white Democrat. County Judge Brashear, another black Republican running at-large against a white Democrat, took 66% of the white vote in his successful race for a county court judgeship.
Just as black Republicans did as well or better than white Republicans, black Democrats also won as large a percentage of the white vote as white Democratic candidates. The white vote for Democratic candidates ranged between 23% and 39%. According to plaintiffs’ exhibits, black Democrat Oliver won about 38% — a larger than average share of the white vote for a Democrat. Winn, another black Democrat, received almost four out of ten white votes. By comparison, white Democrat Brin received no more than 29% of the white vote when running against Wright, a black Republican.
Republican candidates lost the black vote and won the white vote regardless of their public positions on matters related to race. Judge Wright, for instance, had been a member of the Dallas Chapter of the Coalition of 100 Black Women; served as a legal intern for the Lawyers’ Committee on Civil Rights, a project related to civil rights in South Africa; and was a charter member and past vice-chair of the National Political Congress of Black Women. By contrast, the record is silent regarding the record of her white Democratic opponent, Brin. Brin nevertheless won the black vote handily in the general election. At the same time, an overwhelming number of white voters supported Wright.
Dr. Champagne, an expert called by defendants, testified that this voting pattern was the result of straight-ticket voting. According to Dr. Champagne, judicial elections are *879low-profile elections in which the voters know little more about the candidates than what they read on the ballot. The voters, therefore, will make their choice based upon the information that the ballot contains — party affiliation. Because a majority of voters in Dallas County is Republican, Republican candidates prevail in most of the judicial races.
We are unable to find the requisite presence of race in this data. The undisputed facts permit no conclusion but that the defeat of black-preferred candidates was the result of the voters’ partisan affiliation. The black-preferred candidate was always the Democratic candidate, while the majority of white voters always supported the Republican candidate. It is significant to the probe for racial influences that 30% to 40% of the white electorate supported Democratic candidates, although the combination of black and white Democratic votes was insufficient to carry the Democratic candidate. The point is that a black Democratic voter and a white Democratic voter stand in the same position. Both are unable to elect the Democratic judicial candidate they prefer.
We repeat. The race of the candidate did not affect the pattern. White voters’ support for black Republican candidates was equal to or greater than their support for white Republicans. Likewise, black and white Democratic candidates received equal percentages of the white vote. Given these facts, we cannot see how minority-preferred judicial candidates were defeated “on account of race or color.” Rather, the minority-preferred candidates were consistently defeated because they ran as members of the weaker of two partisan organizations. We are not persuaded that this is racial bloc voting as required by Gingles.
Plaintiffs contend that the Democratic Party better represents the political views of black voters in Dallas County. This is doubtlessly the view of black voters, but it is not relevant to whether the minority-preferred candidate is defeated on account of race. To the extent that candidates preferred by black voters are consistently defeated because of their substantive political positions, they are the casualties of interest group politics, not racial considerations. This is not the harm against which § 2 protects. Section 2 protects black voters against defeat on account of race or color, not on account of political platform. See Whitcomb, 403 U.S. at 154-55, 91 S.Ct. at 1874-75. We are sensitive to the reality that political positions can be proxies for racial prejudice. However, where white voters support black candidates of a particular party in larger percentage than they support white candidates of the same party, there is no basis, without more, for us to conclude that the parties’ political positions are proxies for racial bias.
Even assuming arguendo that plaintiffs have met the Gingles threshold by showing racial bloc voting, the totality of circumstances in the record cannot support a § 2 violation. Plaintiff-intervenors Oliver, White, and Tinsley contend that “race considerations pervade elections in Dallas County.” They support this proposition with the district court’s finding that there were two instances of overt or subtle racial appeals in Dallas County elections. In one, judicial candidate Baraka was labeled a “Black Muslim” by his opponent. In another, district attorney candidate Vance printed his own and his opponent West’s pictures in campaign literature, thus informing the' electorate that he was a white candidate running against a black opponent. Nothing in the district court’s opinion indicates that these racial appeals were anything more than isolated incidents. In the only judicial election affected by a racial appeal, Judge Baraka, the black candidate, won both the Republican primary and the general election, winning a majority of the white vote in both elections.
Oliver, Tinsley, and White also contended at trial that voting patterns in nonpartisan elections show that partisan affiliation could not explain the defeats of black-preferred candidates. Dr. Weiser, a statistician with experience in voting rights litigation, presented this data. Weiser examined seven Dallas City Council elections, a presidential primary, and' referenda on public transit funding, a police-review board, and the city council structure. The district court made no findings about the data. Assuming ar-*880guendo that these high-profile elections had any relevance to voting patterns in low-profile judicial elections, the data presented do not support plaintiffs’ argument. For example, although Weiser emphasized that black city council candidate Williams drew 27% and 7% of the vote in the most predominantly black and white precincts, respectively, the fact remains that his opponent Rucker won a majority of the black vote and the election. Other black-preferred candidates, such as Strauss in 1988, prevailed in three of the other six city council elections Weiser studied. In the 1984 Democratic presidential primary, Jesse Jackson won a plurality (46%) of the Dallas vote. Simply put, these nonpartisan elections do not demonstrate the consistent defeat of minority-preferred candidates. To the contrary, the evidence shows that black-preferred candidates won a majority of the white vote and the election in most cases.
Extending our compass to the totality of circumstances fails to bring evidence that racial politics played any role in the defeat of black-preferred candidates. The district court rejected the suggestion that the Republican Party is a white slating organization. Testimony shows that any eligible candidate could run as a Republican, regardless of race. The plaintiff-intervenors testified themselves that they had been heavily lobbied by the Republican Party leadership to run on the Republican ticket. Running as Republicans, the great likelihood is that these former district judges would have been elected, as plaintiff-intervenors conceded at trial.
The plaintiffs presented general evidence of the lingering effects of past discrimination, but offered no specific evidence of depressed levels of black political participation such as low black voter registration or turnout. On the contrary, the minority-preferred candidates ran professional, well-financed campaigns backed by the Democratic Party, a party that, until the late 1970’s, had dominated Dallas County judicial races just as completely as the Republican Party now dominates those races. These Democratic candidates lost because Dallas County shifted from being a county of predominantly Democratic straight-ticket voters to a county of mostly Republican straight-ticket voters.
Plaintiffs made no factual riposte to the overwhelming evidence that election outcomes were the product of partisan affiliation. Rather, plaintiffs’ answer was the legal assertion that the effect of partisan affiliation, virtually admitted, is not relevant. Plaintiffs’ expert, Engstrom, conceded that there is “a stronger association between partisan affiliation and success than there is between the race of the candidate and success,” while clinging to the assertion that partisan affiliation does not explain all of the voting patterns in Dallas County. Finally, he conceded that he had no data that black Democrats generally did worse than white Democrats. In fact, the undisputed facts show that, when one controls for party, black candidates did as well as, or better than, white candidates in winning the white voter and elections. Plaintiff-intervenor White conceded that partisan affiliation determined her electoral defeat in 1984. She admitted that “if I ran as a Republican ... the likelihood is that I would win.”
In short, the facts demonstrate that partisan affiliation, not race, was responsible for the defeat of the minority-preferred candidate in Dallas County. The district court erred in finding racial vote dilution and' a violation of § 2.
B. Harris County
Harris County elects 59 district judges at-large. Three are black, three are Hispanic, and the rest are Anglo. One black county court judge also was elected at-large. Uncontested expert testimony and surveys establish that black lawyers make up at most 3.8% of the eligible lawyers, but comprise 5.1% of Harris County’s district judges. According to plaintiffs’ evidence, 1,685,024 people of voting age reside in Harris County; 305,986 (18.2%) are black, and 222,662 (13.-2%) are Hispanic. Plaintiffs claim to represent all black voters in Harris County.
The district court found a § 2 violation based on the three Gingles prerequisites, two primary Zimmer factors, and three “enhancing” Zimmer factors. The primary Zimmer factors were (1) the general history and lingering effects of past discrimination and (2) *881the small number of successful black judicial candidates. The enhancing Zimmer factors were (1) the large size of Harris County; (2) the prevention of single shot voting by numbered post election; and (3) the majority runoff requirement in primary elections. Defendants contend that, under the proper legal standards, the evidence before the district court amounts at best to a weak case of dilution that is clearly outweighed by Texas’ interest in linkage. We agree.
Defendants argue that the district court’s determination that Harris County district court elections were characterized by legally significant racial bloc voting rests on two fundamental departures from controlling law. They maintain that the district court erred in (1) refusing to consider evidence demonstrating that divergent voting patterns among black and white voters were attributable to partisan affiliation and (2) excluding elections in which the black-preferred candidate was Hispanic despite overwhelming evidence that Harris County black and Hispanic voters were a cohesive group within the meaning of § 2. In light of our previous discussion, these contentions plainly have merit.37
Engstrom studied only 17 district court elections involving black candidates. Taebel studied 45 Harris County judicial elections between 1980 and 1988 with either a black or Hispanic candidate, including 24 district court elections, 9 county court elections, one court of appeals election, one Supreme Court election, and ten primary elections. Taebel examined all but two of the elections analyzed by Engstrom. Including the 42 races listed in Judge Wood’s exhibits, the record before the district court contained a total of 45 general elections that involved minority candidates. Forty of these were indigenous district or county court elections.38
*882As in Dallas County, voters’ preferences were strongly influenced, if not dictated, by partisan affiliation. The black-preferred candidate in Harris County, regardless of race, *883was always the Democratic candidate. For white voters, party affiliation always trumped race in predicting which candidates would be supported. The Republican candidate always won the white vote, generally taking between 55% and 65%, whether the Republican candidate was black, Hispanic, or Anglo. Similarly, Democratic candidates always took almost all black and Hispanic votes, even when a white Democrat ran against a black or Hispanic Republican.
Both the exhibits and expert testimony indicated that party, not race, was the decisive factor in determining electoral outcomes. For example, when white Democrat Schuble defeated black Republican Proctor in a 1986 district court race, Proctor won the majority of the white vote, but lost more than 95% of the black vote to Schuble. Likewise, when Irvin, another black Republican, ran for a county court judgeship against white Democrat Duncan, Irvin won the white vote while Duncan received virtually all of the black vote. Kenneth Hoyt, now a United States District Judge, won the white vote and the election in his bid for the state appellate bench against a white Democratic opponent in 1984. Yet despite the endorsement of the Houston Lawyers’ Association, Judge Hoyt lost virtually all of the black vote.
It is against this backdrop of straight-ticket voting that the limited success of black-preferred candidates described by Engstrom must be assessed. Engstrom limited his study ■ to elections involving black candidates. Since the black-preferred candidate often is not black, this precluded Eng-strom from determining whether whites in Harris County consistently voted as a bloc to defeat black-preferred candidates, as he admitted at trial. Engstrom also excluded judicial elections with Hispanic candidates and races for seats on the county court, which are also conducted on a county-wide basis.
The black-preferred candidate won only two, or 11.8%, of the 17 district court elections analyzed by Engstrom. Ten of these losses by black Democrats, however, occurred in 1980,1984, and 1988, when popular Republican presidential candidates helped Republican judicial candidates to defeat virtually all of their Democratic opponents. The victors included Judge Hoyt, a black candidate running as a Republican. As Eng-strom conceded, white and black Democrats alike- were “wiped out” during these years.
The fortunes of Harris County Democrats, and thus black voters, improved considerably in 1982 and 1986, when either Governor Mark White or Senator Lloyd Bentsen headed the Democratic ticket. As in the Republican years of 1980, 1984, and 1988, success at the top of the ballot carried down to judicial races marked more by anonymity than name identity. Thus, black-preferred candidates won more than a third of the indigenous judicial races in which black candidates participated — 5 out of 14, or 35.7%. Considering elections with Hispanic candidates, black-preferred candidates won 13 out of 24 indigenous judicial elections, or 52.4%. Narrowing the focus to district court races, black voters elected the candidate of their choice 8 out of 14 times. In these years, the black-preferred candidate for district judge won in 57.1% of the elections studied. Even when the results of the lean years of 1980, 1984, and 1988 are included, we find that the black-preferred candidate prevailed in 14 out of 40 (35%) of all indigenous judicial elections with minority candidates. The record also indicates that black-preferred candidates won three out of five exogenous races for appellate and Supreme Court seats during these years.
Plaintiffs insist, however, that partisan affiliation cannot explain all of the results in this case, for in years not dominated by the Republican Party, black Democrats enjoyed less success than other Democrats. In the 1982 district court contests, white and His-panic Democrats won 12 of 14 races, while black Democrats won only one of three. In 1986, black Democrats won two of eight indigenous judicial races; the success rate of white Democrats is not found in the record. Engstrom stated, without discussing the supporting data, that between 1980 and 1988, white Democratic candidates enjoyed a better success rate than their black counterparts.
This evidence may reflect a preference among white Democrats for white and His-panic rather than black candidates. Plain*884tiffs’ assertion that race is at work, however, is contradicted by the success of black candidates in the Democratic primary, where party affiliation plays no part. An exhibit introduced by Houston Lawyers’ Association shows that black candidates won 9 of 16, or 56.3%, of the contested primary races. This showing detracts from the force of plaintiffs’ claim.
The proof of vote dilution is marginal. The undisputed facts show that a majority of white voters invariably supported black Republican candidates, suggesting that the defeat of minority-preferred candidates was largely, although not entirely, attributable to partisan affiliation. Moreover, black voters were consistently able to elect representatives of their choice, even when they were opposed by a majority of white voters. The record indicates that black-preferred candidates prevailed in 14 out of 40 non-exogenous elections in which either black or Hispanic candidates participated — 35% of the time. Limiting the inquiry to district court races, black-preferred candidates still won in 9 of 28 races, or 32.1%.
Black voters could, therefore, repeatedly elect candidates of their choice, even when opposed by a majority of white voters. Far from being submerged in a white majority, black voters were a potent electoral force that could form coalitions with minorities of white voters to elect their preferred candidates. This ability to form coalitions and influence the elections of all judges in Harris County would be lost in the system of single-member districts proposed by the plaintiffs. Instead, black voters might control the election of perhaps ten judges, abdicate any right to vote for the remaining forty-nine, and thus radically reduce the chances of having disputes affecting them decided by a judge they had any hand in electing. A similar observation can be made in the other counties but is strikingly apt in this large urban environment.
The remaining evidence adds little to plaintiffs’ claims of illegal vote dilution. Plaintiffs offered little evidence that past discrimination and socioeconomic disparities between blacks and whites hindered the ability of black residents of Harris County to participate in the political process. In particular, there was no suggestion at trial of a lower-than-average voter registration or turnout rate among black citizens. In addition, the evidence indicated that disproportionate levels of poverty within the black community had no effect on the ability of black judicial candidates to raise the funds necessary to compete on a county-wide basis. At trial, Bonnie Fitch testified, without elaboration, that a few black incumbents experienced some difficulty in obtaining financing for their campaigns, but she attributed these problems to “racial discrimination” and the candidates’ “past record of losing.” Even if this isolated, equivocal testimony could somehow be construed to suggest that a lack of resources among black residents hindered black candidates’ campaigns, it was sharply contradicted by the accounts related by Jackson and Berry, two black judicial candidates. They testified that they were each able to raise sufficient funds and that they in fact outspent their white Republican opponents by ratios exceeding ten-to-one. In light of this evidence, the district court’s finding that the effects of past discrimination hampered the black community’s access to the political process was clearly erroneous.
Likewise, the representation of blacks on the Harris County bench cannot support an inference of racial politics. Three blacks are district judges — 5.1% of the total. By contrast, black attorneys make up at most only 3.8% of the eligible lawyers in Harris County. The fact that blacks constitute a smaller percentage of the district judges than of the county population is therefore not surprising. If judges were chosen at random from the pool of eligible candidates, there would be fewer black district judges on the Harris County bench.
Aside from the number of black judges and the general history of discrimination, the district court found three Zimmer enhancing factors. See Nevett v. Sides, 571 F.2d 209, 218 (5th Cir.1978). Such factors enhance the opportunity of a white majority to engage in racial politics. They do not, however, “meaningfully advance the inquiry into whether race is at issue,” Terrazas v. Clements, 581 F.Supp. 1319, 1346 n. 26 (N.D.Tex.1983) *885(three-judge panel), and therefore do not support an inference of racial politics in Harris County.
The circumstantial evidence of a relation between black voters’ electoral losses and race is, at best, tenuous, given the willingness of white voters to support black Republican candidates and the consistent success enjoyed by black-preferred judicial candidates. Even if the considerable success among black-preferred Hispanic candidates is discounted, the evidence presented at trial hardly amounts to the level of dilution that might outweigh Texas’ substantial interest in linking a trial judge’s jurisdiction with her electoral base. Given the undisputed evidence that nearly all of the losses suffered by black candidates occurred in years when virtually the entire party slate went down in defeat and plaintiffs’ negligible showing under the Zimmer factors, the claim before us reduces itself to a contention that Texas’ 143-year-old electoral scheme must be dismantled in Harris County because a few black candidates — most of them recently-appointed incumbents — failed to attract decisive support from white voters within the Democratic Party. We express no opinion as to whether this minimal proof of dilution might establish a violation of § 2 absent the substantial state interest. Even assuming that it would, we conclude as a matter of law that plaintiffs’ proof at best produces only a marginal case in Harris County, too insubstantial to survive the weighing of the totality of the circumstances particularly so if any appreciable weight is given the linkage interest.
C. Tarrant County
There are 23 district courts in Tarrant County. From 1985 to 1988, three of these judges (13.0%) were black. As of 1989, two district judges are black (8.7%). The defendants’ undisputed evidence indicates that only 2.4% of the eligible Tarrant County lawyers are black. There are 613,698 residents of voting age in Tarrant County. Of this number, 63,851 (10.4%) are black. Plaintiffs proceed on behalf of black voters in Tarrant County.
The evidence indicates that blacks voted cohesively for the Democratic candidate. Dr. Brischetto, plaintiffs’ expert for this county, analyzed four elections: three general elections for district judgeships and the 1988 Democratic presidential primary. In all four elections, the regression estimates show that from 85% to 100% of black voters in Tarrant County supported the black-preferred, Democratic candidate. Taebel’s analysis similarly shows cohesion.
Taebel analyzed nine general elections, including three exogenous elections, in which a black or Hispanic had participated. These included five district court races, one county court race, two Supreme Court races, and a contest for Texas Attorney General. Bris-chetto analyzed only four elections, in which black candidates had participated. As in all other counties, the evidence shows consistent black support for Democratic candidates. The following tables summarize the analyzed races involving black or Hispanic candidates. For each black-preferred candidate, the estimated percentage of the white vote is listed. These are based upon Taebel’s estimates, except those in parentheses, which reflect Brischetto’s regression and homogenous precinct analyses. A “check” mark indicates a victory by the black-preferred candidate.
Table VI.C
Indigenous judicial elections (Tarrant County)
Year Court Candidates Race Party White Vote
1982 233d District Valderas Hines H W D R 36% /
" County Crim Ct 4 Perez Lynch H W D R 48% /
1986 233d District Weaver Valderas W H R D 40%
" Crim Dist Ct 1 Sturns Goldsmith B W R D 43% (51-56%)
*886Year Court Candidates Race Party White Vote
" Crim Dist Ct 4 Drago Salvant W B D R 45% (54-59%) /
1988 Crim Dist Ct 2 Dauphinot Davis W B R D 40% (42-50%)
Exogenous elections (Tarrant County)
Year Court/Office Candidates Race Party White Vote
1986 Attorney General Barrera Mattox H W R D 39%
Supreme Court PI 4 Bates Gonzalez W H R D 38%
Supreme Court PI 3 Gonzalez Howell H W D R 46%
1988 Dem Pres Primary Jackson Dukakis B W (14-16%)
Gore Gephardt W w
Hart Simon w w
Unlike other counties, black judges occupied more than 13% of the district judgeships in Tarrant County for four out of five years — a proportion of the bench that is greater than the proportion of black voters in the county’s population.
The success of black-preferred candidates was also greater in Tarrant County than elsewhere. In those general elections with black candidates, the black-preferred, Democratic candidate won only one out of three general elections — 33.3% of the studied races. However, in nine general elections with either black or Hispanic candidates included in Taebel’s study, the black-preferred candidate won four out of nine, or 44.4% of the elections. In the six indigenous district and county court elections studied, the black-preferred candidate won three out of six, or 50% of the elections. These figures do not indicate the consistent defeat of black-preferred candidates.
The district court, by contrast, found that the black-preferred candidate was consistently defeated in Tarrant County. The district court reached this conclusion by ignoring elections in which Hispanics had participated. This rejection of white-Hispanic elections was erroneous. The undisputed facts, as reflected by Taebel’s 'exhibits, are that a majority of Hispanic voters always supported the same candidate favored by black voters in every general election. The district court found that Hispanic and black voters were cohesive in Midland, Lubbock, and Ector Counties on similar evidence. With virtually identical proof in Tarrant County, the same conclusion must follow, and we hold that it does.
Brischetto included the 1988 Democratic presidential primary in which Jesse Jackson won virtually all of the black vote in Tarrant County, but only between 14% and 16% of the white vote.39 Defendants exhibits include a 1986 Democratic primary for district court in which Ross, the black-preferred candidate, received 57% of the black vote, but lost the white vote and thus came in third out of a field of four candidates. Those exhibits show, however, two other Democratic primaries — ignored by Brischetto — in which black-preferred Hispanic candidates prevailed.40
*887In short, the evidence shows that the black-preferred candidate won 40% of the Democratic primaries and half of the indigenous judicial elections studied, including the elections with Hispanic candidates. The record also shows that black judges have consistently made up a greater proportion of district judges in Tarrant County (13.0%) than the proportion of black voters in the county’s population (10.4%), and far more than the proportion of eligible black attorneys (2.4%).
Furthermore, the undisputed evidence shows that black candidates won as great a share of white votes as white candidates, if we control for party affiliation. For instance, Sturns, a black Republican with a long history of involvement in civil rights and black community organizations, won 57% of the white vote to beat a white Democrat. Sal-vant, another black Republican, also won a majority (55%) of the white vote, although he lost his race for a district judgeship to a white Democrat supported by a combination of black voters, Hispanic voters, and white Democrats. Black Republicans also won the same share (50%) of elections as white Republicans among the races with black or Hispanic candidates.
Finally, blacks have not been underrepresented on the Tarrant County bench. Plaintiffs’ exhibit indicates that, for four out of the five years studied, three of Tarrant County’s district judges were black; for these four years, while blacks made up only 10.4% of the county’s voting age population, more than 13% of the relevant office holders were black. Given this persistent and substantial black presence on the Tarrant County bench, the consistent and substantial success of minority-preferred candidates, and the absence of any evidence of racial politics in Tarrant County, we conclude that, even if the plaintiffs had proven the Cingles prerequisites, the district court clearly erred in finding illegal vote dilution under the totality of circumstances. This is so even if we ignore the fact that blacks and Hispanics voted cohesively in Tarrant County and exclude the races with Hispanic candidates from our analysis. Looking at just the district court general elections involving black candidates, the black-preferred candidate won 33.3% of the time. In light of the evidence just discussed, which excluding Hispanic elections does not change, reducing the relevant success rate from 44.4% to 33.3% is insignificant in the totality of the circumstances. There is no case ás a matter of law in Tarrant County.
D. Travis County
There are 13 district judges elected in Travis County. From 1985 to 1988, one of them was Hispanic, or 7.7% of the total. This judge was defeated in 1988. Hispanic lawyers make up 2.7% of the eligible lawyers in the. county. There are 312,392 voting age residents in Travis County, which encompasses Austin, Texas. Of them, 44,847 (14.-4%) have Spanish surnames. Only 29,067 (9.3%) are black. The district court found that a “minimally contiguous,” predominantly-Hispanic judicial district could be created. Plaintiffs proceed on behalf of Hispanic voters in Travis County.
Plaintiffs’ witnesses stated that the Republican Party is insignificant in Travis County and the proper- testing ground for candidates is the Democratic primary. Plaintiffs analyzed three Democratic primary elections: one for district court and two for county court positions. Defendants analyzed eleven elections: four exogenous general elections, including one state senate, one Attorney General, and two Supreme Court races; four exogenous primary elections, including one state senate, one Supreme Court, and two appellate court primaries; and finally the same three indigenous judicial elections studied by plaintiffs.
By Taebel’s analysis, the Hispanic-preferred candidate won all four of the exogenous general elections. In three of the four, the Hispanic-preferred candidate also won a majority of the Anglo vote. In addition, the Hispanic-preferred candidate won two exogenous primaries, for Supreme Court and state senate. Thus, the Hispanic-preferred candidate prevailed in two of the four (50%) exogenous primaries and two of the seven (28.6%) primaries studied overall. Altogether, the Hispanic-preferred candidate won *88854.5% of the indigenous and exogenous elections analyzed.
The district court found, however, that the three indigenous primary elections for judicial positions were “closer in nature to District Court elections” and sufficed to show a pattern of racial bloc voting sufficient to defeat the Hispanic-preferred candidate. The district court therefore relied solely on the three elections analyzed by both Taebel and Brischetto to find that the Hispanic-preferred candidate lost 100% of the time.
In the one district court and two county court primary elections analyzed by the parties, the Hispanic, and Hispanic-preferred, candidate was defeated by a white majority. In one of these races, however, white voters gave their support to a black candidate, and thereby defeated both the Hispanic Castro and Hughes, a white candidate. Kennedy, the black candidate, had the overwhelming support of black as well as white voters, so it is difficult to conclude that Castro was defeated by a white bloc. Castro and Hughes were defeated by a black-white coalition. Thus, Castro’s defeat is not evidence of the white majority’s ability “usually to defeat the minority’s preferred candidate.” Gingles, 478 U.S. at 51, 106 S.Ct. at 2766-67.
The two remaining indigenous primary elections offer a meager base for liability. The plaintiffs’ case reduces to three facts: (1) Hispanic-preferred candidates Gallardo and Garcia gained only 33% to 37% of the Anglo vote in 1988 Democratic primaries, and failed to win nominations for district and county court elections; (2) only one Hispanic, Gallar-do, served as district judge between 1985 and 1998, while no Hispanic now serves; and (3) Hispanics have suffered from past discrimination in Travis County. We conclude that the district court clearly erred to find illegal vote dilution on this record.
In finding clear error, we repeat Justice Brennan’s admonition that “the usual predictability of the majority’s success distinguishes structural dilution from the mere loss of an occasional election.” Gingles, 478 U.S. at 51, 106 S.Ct. at 2767. It defies common sense to believe that the loss of two primary races in one year constitutes usual and predictable defeat by a white bloc, rather than simply “loss of an occasional election.” However, assuming arguendo that these two elections constitute ■ sufficient proof of the third Gingles prerequisite, they are too meager to prove dilution under the totality of circumstances, as a matter of law.
The plaintiffs contend that Hispanics are underrepresented among district judges in Travis County. Hispanics made up 7.7% of those judges in four out of five recent years, while making up no more than 2.7% of the lawyers eligible under Texas law to fill those posts. Given such a small pool of qualified candidates, it is not surprising that Hispanics have made up a small proportion of the Travis County bench. This result need not be attributed to the interaction of racial bias with the at-large electoral scheme. It is equally likely that the numbers reflect the limited candidate pool. Plaintiffs can point to only one district court election that an Hispanic candidate lost — Gallardo’s race in 1988. Even if Gallardo had prevailed, the percentage of Hispanic judges would not have increased, because Gallardo was the one Hispanic sitting before 1988. While we do not require that any minority candidates run for the office in question, the court cannot ignore this reality while plaintiffs emphasize the absence of minority office holders.
Far from signalling the submergence of minority voting strength by an interaction of electoral process and bias, the undisputed facts indicate that Travis County’s political system is open to Hispanic and white candidates alike. Hispanics won half of the four exogenous primary elections studied, including races for the state senate, appellate courts, and Supreme Court. The Hispanic-preferred candidate also won all four of the general elections analyzed by the defendants. The City of Austin contains most of Travis County’s population. As this court noted in Overton v. City of Austin, 871 F.2d 529, 540 (5th Cir.1989):
Austin has repeatedly elected black and Mexican-American council members during the past 17 years.... [T]he winning minority candidates ixequently received well over fifty percent (50%) of the Anglo vote and were also the preferred candidates of the minorities. Minority candi*889dates have routinely been elected to other posts in Austin and the surrounding Travis County.
Likewise, the defendants here produced evidence that Hispanic county commissioners had been elected from predominantly Anglo districts, and that Trevino, a Hispanic Austin city council member, had been elected in city-wide elections. Against this background, which includes the success of state Senator Barrientos and Justice Gonzalez and other Hispanie-preferred candidates, plaintiffs’ minimal case is plainly insufficient to prove illegal vote dilution. The district court clearly erred in finding otherwise.
E. Bexar County
The voting age population of Bexar County is 672,220. Of these, 46,767 (7.0%) are black, and 278,577 (41.4%) are Hispanic. Nineteen district judges are elected from Bexar County. Of this number, five (26.3%) are Hispanic. Undisputed evidence shows that 11.4% of the eligible lawyers in Bexar County are Hispanic. Plaintiffs proceed on behalf of Hispanic voters in Bexar County.
Plaintiffs and defendants analyzed the six district court general elections with Hispanic candidates between 1982 and 1988. Defendants also studied a 1980 general election with an Hispanic candidate, as well as two appellate court and three county court general elections with either Hispanic or black candidates. As in every other county, Hispanics voted cohesively for the Democratic candidate while a majority of Anglos supported the Republican candidate.
In the twelve judicial elections studied, the Hispanie-preferred Democratic candidate won four times, 33.3%. The Republican candidate usually won the general election, and always won the Anglo vote, regardless of the candidate. The four Democratic victories were: (1) the 1980 appellate court race between Murry and Esquivel; (2) the 1980 district court race between Prado and Priest; (3) the 1988 district court race between Bowles and Míreles; and (4) the 1988 county court race between Patterson and Canales. Priest, an Anglo Democrat, beat Prado, an Hispanic Republican, while Esquivel, Mí-reles, and Canales, all Hispanic Democrats, defeated their Anglo Republican opponents.
Partisan affiliation does not explain, however, the voting patterns in Democratic primary elections. By defendants’ own evidence of Democratic primaries in Bexar County, the Hispanie-preferred 'candidate lost in nine of fourteen elections, prevailing only 35.7% of the time, when Anglo Democrats voted for the Hispanic candidate’s Anglo opponent. Anglo support for the Hispanic candidate was seldom above 30% and as low as 1% — whereas the Hispanic vote for the Hispanie-preferred, and always Hispanic, candidate was above 70% for five of the nine unsuccessful candidates. Plaintiffs’ as well as defendants’ experts agreed, however, that primary elections do not provide a reliable guide where, as here, both parties are competitive, since they involve only a fraction of the electorate.
Partisan affiliation accounts for much of the voting patterns analyzed by the parties. Most Anglo voters are Republicans; most Hispanic voters are Democrats. Anglo voters gave a majority of their votes to Republicans, and Hispanic voters gave a majority of their votes to Democrats, even when Hispanic Republican candidates faced Anglo Democratic opponents. Prado and Barrera, His-panic Republicans, won 70% and 86% of the Anglo vote respectively, when running against Anglo Democratic opponents who received the overwhelming majority of the His-panic vote. Any proof of dilution is meager at best and cannot overcome Texas’ substantial linkage interest as a matter of law.
Because Hispanic voters make up more than 41% of the population, they can elect Democratic candidate with minimal Anglo support and have done so repeatedly. The minority-preferred candidate won four out of twelve elections in which an Hispanic candidate participated — 33.3% of the time — with as little as 17% of the Anglo vote. Hispanic voters are plainly a potent political force that can elect candidates by forming coalitions with small percentages of Anglo voters. If Bexar County were subdistricted, Hispanic voters might elect a few more of their preferred candidates, but only at the expense of losing their influence over the majority of *890Bexar County judges. The perversity of such a result is self-evident.
Finally, the evidence that elections were affected by racial politics preventing the formation of such coalitions is thin. It consisted solely of (1) evidence of low Hispanic voter registration; (2) the usual enhancing factors present in every Texas county — anti-single shot voting and the majority runoff requirement; and (3) the fact that Hispanic judges occupy five of nineteen district judgeships— 26.3% of the total — while Hispanics make up 41.4% of Bexar County’s voting age population. Again, we note that Hispanic attorneys make up only 11.4% of the eligible bar, so that the representation of lawyers on the bench is actually higher than would be produced by random selection from the pool of eligible candidates. This evidence, even if probative in the abstract, is as meager as the evidence in Harris County.
The evidence compels the conclusion that any dilution was marginal and cannot as a matter of law survive the weighing of the totality of the circumstances when Texas’ substantial state interest is added to the mix. If Texas’ linkage interest does not outweigh this evidence of dilution, the state’s interest would be a nullity. We hold that plaintiffs’ proof fails in Bexar County as a matter of law.
F. Jefferson County
Eight district judges are elected from Jefferson County. The record shows that no black judge was elected there between 1985 and 1989.41 Expert evidence establishes that fourteen eligible attorneys in Jefferson County, 3.1% of the qualified bar, are black. The voting age population of Jefferson County is 179,708. Of this number, 44,283 (24.6%) are black. Plaintiffs proceed on behalf of black voters in Jefferson County.
Taebel testified that Jefferson County is the most Democratic of the targeted counties, with 90% of its voters participating in the Democratic primary. Brischetto analyzed eight primary and runoff elections, including the 1988 Democratic presidential primary. Taebel analyzed six exogenous elections involving either black or Hispanic candidates: four primaries and two general elections. Unlike their other studies, Brischetto and Taebel analyzed totally different elections.
In all but one of the primary elections studied by Brischetto, the black vote was cohesive. In one case, the candidate with the greatest black support received a high plurality (47%) of the black vote. A majority of white voters always opposed the black-preferred candidate in the primary elections.
Whether the black-preferred candidate was consistently defeated by a white bloc is a close question. The answer varies with the elections counted and how they are counted. Defendants point to four primaries. In two elections, black candidate Price won the nomination for state representative. In two others, for Supreme Court and Court of Criminal Appeals, black-preferred Hispanic candidates Gonzalez and Martinez participated. Gonzalez won the Jefferson County Democratic vote; Martinez did not. Defendants also rely on two exogenous general elections, for Supreme Court and Attorney General, involving Hispanic candidates Gonzalez, a Democrat, and Barrera, a Republican. In both general elections, the black-preferred candidate — Gonzalez and Mattox, Barrera’s Democratic opponent — prevailed.
Plaintiffs offer five indigenous primaries, ranging back to 1972, in which black candidates participated — four for justice of the peace and one for county court. They also submitted the exogenous 1988 presidential primary. Among these six races, the black-preferred candidate prevailed only once, when Jackson won a plurality in the 1988 presidential primary.
The total of eight elections analyzed by Brischetto includes both the initial primaries and subsequent runoffs for justice of the peace in 1972 and 1974. In the initial prima-*891ríes, black-preferred candidate Freeman failed to win the highest plurality. Plaintiffs would count these results as “defeats” separate from Freeman’s subsequent defeat in the runoffs. Freeman won, however, the second highest number of votes in the initial primaries and thus made those runoffs. We do not consider Freeman’s showings, in the initial primaries to be separate from the runoff elections. Thus, the record reflects four justice of the peace elections, not six.
Unlike Tarrant County, defendants’ evidence does not include estimates of how Hispanic residents in Jefferson County voted. There are no facts showing that Hispanic and black voters were politically cohesive in Jefferson County. Anglo-Hispanic elections are entitled to less weight than white-black races in determining the success of black-preferred candidates.
Nonetheless, confining our consideration to the analyzed elections in which black candidates participated, we must conclude that the plaintiffs failed to prove a substantial case of dilution. The plaintiffs and defendants together produced evidence of eight primary elections in which a black who was also the black-preferred candidate participated. The. black-preferred candidate won three primaries out of these eight elections — -a success rate of 37.5%. All three of the black-preferred candidates’ victories were exogenous: Jackson won the 1988 presidential primary, while Price won two Democratic primaries for state representative.
As in every county but Dallas, the district court found no sign of racial appeals. Likewise, there is no finding of nonresponsiveness on the part of elected officials to the concerns of black constituents. Enhancing factors as well as past discrimination were shown, but — as elsewhere — were not brought home to this case. The minority-preferred candidate prevailed in every general election submitted by the parties.
The plaintiffs’ ease was further weakened by their use of dated statistics: three of the five indigenous elections they submitted were held in 1972, 1974, and 1978. There is no evidence of a practical and searching appraisal of contemporary conditions in Jefferson County. See Nipper v. Chiles, 795 F.Supp. 1525, 1540 (M.D.Fla.1992) (noting limited probative force of “stale” elections).
We have here no more than marginal proof of illegal vote dilution. The evidence is inadequate to prove that black voters were denied an equal opportunity to participate in the political process. It is too insubstantial to survive a weighing of the totality of circumstances when the state’s substantial linkage interest is added to the mix. As a matter of law, the state’s interest outweighs the plaintiffs’ case.
G-. Midland County
Midland County contains 82,636 voting age residents, of whom 6,893 (11.9%) have Spanish surnames and 4,484 (7.8%) are black. There are three district judges in Midland County; none are Hispanic or black. Undisputed evidence shows that seven Hispanic and three black attorneys are eligible for district judgeships. They comprise 3.2% of the lawyers eligible to run for that office. Plaintiffs proceed on behalf of both Hispanic and black voters in Midland County.
Plaintiffs analyzed three general elections in Midland County. Two of them were exogenous races for the Texas Supreme Court. The third was an indigenous race involving a black candidate for a Justice of the Peace position in 1986. Defendants likewise examined Gonzalez’s bids for the Supreme Court in 1986 and 1988, as well as four primary elections in which either a black or Hispanic candidate participated. Defendants also analyzed the Mattox-Barrera race for Texas Attorney General. .
Both parties’ analyses show that the majority of Anglo voters always opposed the candidate preferred by the geographically compact and cohesive combined minority population in the general elections. The minority-preferred candidate was always defeated by this Anglo majority.
We conclude that the district court clearly erred in finding dilution. The undisputed facts indicate that partisan affiliation, not race, caused the defeat of the minority-preferred candidate. The majority of minority voters always cast their votes in favor of the Democratic candidate. The Anglo voters *892cast the majority of their votes for the Republican, regardless of the race of the candidates. Indeed, Barrera, the Hispanic Republican candidate for Attorney General, won 76% of the Anglo vote when running against Mattox, a white Democrat — the second highest vote received by any of the Republicans in the analyzed general elections. Because Republican voters outnumbered Democratic voters, the minority-preferred Democratic candidate consistently lost. The plaintiffs have not established the third prerequisite of Gingles.
Even if plaintiffs could meet the Gingles threshold, the totality of circumstances does not add up to dilution. The plaintiffs can show only a general history of discrimination and a lack of minority judges. These facts prove little. In Midland County, only one minority lawyer has run for local office (county attorney), and none has ever run for a district judgeship. These low numbers reflect the minuscule number of eligible minority candidates. According to the evidence, only ten minority lawyers are eligible to run for the district court seat.
Because the undisputed facts show that partisan affiliation uninfected by racial politics caused the minority-preferred candidates’ defeat, we hold that the district court erred in finding dilution.
H. Lubbock County
Lubbock County residents vote for five district court positions. None of these five judges are black or Hispanic. The surveys introduced by the defendants indicate that 23 Hispanic lawyers in Lubbock County are eligible to run for the district court: The surveys show that no black residing in the county is eligible to do so. The total voting age population is 150,714. Of this number, 22,934 (15.2%) have Spanish surnames and 9,509 (6.4%) are black. Plaintiffs proceed on behalf of the combined Hispanic and black voters in Lubbock County.
None of the parties analyzed indigenous elections in Lubbock County; no minority has ever run for a position on the district court. Plaintiffs analyzed two exogenous primaries and two exogenous general elections, for the Supreme Court and for the Court of Criminal Appeals. Defendants studied the same two general elections, adding an exogenous general election for Attorney General.
Plaintiffs’ and defendants’ evidence shows that blacks and Hispanics tend to vote cohesively. There is also no dispute that the majority of Anglo voters did not support the candidate favored by the minority voters in Lubbock County in any of the elections studied.
As in Dallas and Midland Counties, however, the undisputed facts show that, in general elections, partisan affiliation and not racial politics caused the consistent defeat of the minority-preferred, always Democratic, candidates. The data indicate that, in these counties, over 60% of white voters supported the Republican candidate, while most minority voters supported the Democratic candidate. As a result of this voting pattern, the Democratic and minority-preferred candidate consistently lost to a Republican opponent, regardless of the ethnicity of the candidates.
In the 1986 and 1988 races for the Supreme Court, Hispanic Democrat Gonzalez lost Lubbock County’s vote to white Republican opponents. However, in the contest for Attorney General, Barrera, an Hispanic Republican, defeated Mattox, a white Democrat. Like Gonzalez’s Republican opponents, Barrera took a majority of the Anglo votes, while his white opponent took a majority of the minority votes. In short, as in Midland County, the evidence establishes that voting patterns in Lubbock County were unaffected by the race of the candidates. Rather, they resulted from party loyalty. Therefore, plaintiffs have not met the third Gingles factor.
The plaintiffs point to two exogenous Democratic primary elections for state appellate and Supreme Court positions.42 However, the minority-preferred candidate won a ma*893jority of the votes cast in one of these two elections. Although Martinez was defeated, Gonzalez won a majority of the votes cast in the Lubbock County Democratic primary for the Supreme Court. These primary races, therefore, do not indicate that the minority-preferred candidate was consistently defeated within the meaning of Gingles, and they cannot establish dilution.
I. Ector County
There are four district judges in Ector County. All of them are Anglo. There are fewer than 200 lawyers in the county. Surveys estimate that no more than six of them are black or Hispanic and eligible to become district judges. Ector County, whose principal city is Odessa, has 79,516 voting age residents. 14,147 (17.8%) are Hispanic, while 3,255 (4.1%) are black. Plaintiffs proceed on behalf of the combined minority population in Ector County.
The parties relied on the same exogenous races in Ector County that they produced in Lubbock County. The plaintiffs examined primary and general elections for appellate courts involving Martinez and Gonzalez. The defendants added Barrera’s bid for Attorney General.
The undisputed facts indicate that the minority-preferred, Democratic candidates were consistently defeated in general elections by an Anglo majority voting for their Republican opponents. In -the Democratic primaries, Martinez won a majority of the vote. The minority-preferred candidate won half of the Democratic primary races and therefore was not consistently defeated in the primaries.
As in Lubbock County on virtually identical facts, we find that the district court clearly erred in finding racial vote dilution. The undisputed facts indicate that partisan affiliation controlled the outcomes of the general elections. As in Lubbock County, while His-panic Democratic candidates lost the Anglo vote, Barrera, a Hispanic Republican, won a majority of the Anglo vote running against his white Democratic opponent Mattox.
While partisan affiliation would not explain polarization in the primaries, the facts indicate that the minority-preferred candidate was not consistently defeated by racial polarization in the primary elections. Rather, Martinez won one of the two races analyzed. Plaintiffs have failed to meet the third prerequisite of Gingles.
VII. Conclusion
We would expect over time that the Texas judiciary would reflect the black and Hispanic population eligible to serve — if judges, for example, were drawn from a pool of all persons eligible to serve. In truth, minority lawyers fare better than we would expect from a random process. We do not suggest that because they fare better than they would in a system of random selection, voting rights of blacks and Hispanics could not have been illegally diluted. Rather, the observation is relevant because'it brings'perspective to this battle by drawing borders around 'its asserted implications and deflating over- ■ drawn invocations of large wrongs of history, unremedied and unanswered.
There is no disparity between the number of minority judges and the number of minorities eligible to serve. Rather, the only disparity is between the minority population and minorities eligible to serve as judges. Much can be said about that — of deficits in education and other social shortchangings of black and Hispanic persons. To those who push judicial entry onto this larger field we must answer that our task is more narrowly drawn — to decide if voting rights have been denied. We lack the authority, even if we had the wisdom, to do more. The Voting Rights Act is not an unbridled license — to explore for example the persistent low enrollment of black law students. One small example. This year the law school at Louisiana State University graduated the largest number of black students in its history. This followed intensive recruiting efforts including the inducement of a free education — with stipends. Of the several hundred students graduated, ten were black. This sad story can be repeated at school after school. We are told that this is not relevant. We think that it is.
We decline to reach for social questions beyond the Voting Rights Act by recasting *894its meaning and purpose. Ultimately, we cannot escape the steely truth that we cannot arrive at sound answers if we fail to ask the right questions. We think that today we have asked the correct questions and answered them as best we can.
REVERSED.
. Plaintiffs originally challenged the election of district judges in 44 counties, but by trial, winnowed their targets to the following nine urban counties: Harris County, Dallas County, Travis County, Tarrant County, Jefferson County, Ector County, Bexar County, Midland County, and Lubbock County.
. Plaintiffs early in the case dismissed the Governor.
. This ruling was not appealed.
. Former Secretary of State George S. Bayoud, Jr., a named party defendant, objected to the Attorney General’s decision not to appeal immediately. Bayoud took the position that as chief elections officer of the State of Texas, he was the Attorney General's client and the Attorney General must represent his interests. Bayoud obtained independent counsel and filed a notice of appeal himself.
. Judge Wood also filed a motion to strike the Attorney General's Notice of Action Toward Settlement, which we denied before oral argument.
. We granted Chief Justice Phillips' motion requesting that he be allocated time at oral argument.
. The dissent argues that Chief Justice Phillips was joined solely as a jurisdictional party for Eleventh Amendment purposes. Even if that were true, and it is not, see supra page 901, it would not answer the real question: if the State of Texas is the real party in interest, does the Attorney General possess exclusive authority to choose whether the State’s interests will be asserted on appeal? In Baker, we answered in the negative.
. Justice Gonzalez joined section II.A. of the plurality opinion regarding the entry of redistricting relief.
. Because the office of Attorney General is rooted in the common law, many states, including Texas, refer to their Attorney General’s common law powers. E.g. Martinez v. State, 753 S.W.2d 165, 179 (Tex.App. — Beaumont 1988, writ ref d). Thus, there is some value to looking at how other states have dealt with the issue we face today. In Tice v. Department of Transportation, 67 N.C.App. 48, 312 S.E.2d 241, 246 (1984), the North Carolina court held “that the Attorney General ... is bound by the traditional rule governing the attorney-client relationship, and cannot enter a consent judgment without the consent of the entity represented.” In Georgia, the Attorney General may not "bind his client by settlement for less than the full sum claimed, *843unless express authority be given by the client.” State v. Southwestern R.R., 66 Ga. 403, 407 (1881). The North Dakota Attorney General's power to represent state departments and officers
does not mean that the attorney general, standing in the position of an attorney to a client, who happens to be an officer of the government, steps into the shoes of such client in wholly directing the defense and the legal steps to be taken in opposition or contrary to the wishes and demands of his client or the officer or department concerned.
State ex rel. Amerland v. Hagan, 44 N.D. 306, 175 N.W. 372, 374 (1919), overruled on other grounds, Benson v. North Dakota Workmen's Compensation Bureau, 283 N.W.2d 96 (N.D.1979). According to the Mississippi Supreme Court,
The unique position of the Attorney General requires that when his views differ from or he finds himself at odds with an agency, then he must allow the assigned counsel or a specially appointed Counsel to represent the agency unfettered and uninfluenced by the Attorney General’s personal opinion.
State ex rel. Allain v. Mississippi Public Serv. Comm'n, 418 So.2d 779, 784 (Miss.1982); see also Frazier v. State by and through Pittman, 504 So.2d 675, 691 (Miss.1987) (where attorney general refuses to represent state agency, agency is entitled to its own lawyer and court may retain jurisdiction and entertain the suit). Arizona does not permit its Attorney General to appeal a decision against the wishes of the state agency he represents. Santa Rita Mining Co. v. Department of Property Valuation, 111 Ariz. 368, 530 P.2d 360 (1975). Finally, the authority of the Attorney General of Illinois does not permit him to waive the rights of his client. Cook County v. Patka, 85 Ill.App.3d 5, 40 Ill.Dec. 284, 288, 405 N.E.2d 1376, 1380 (1980).
. Professor Fiss has recognized the problem raised by Attorney General Morales’ actions in this case. "We are left to wonder, for example, whether the attorney general should be able to bind all state officials, some of whom are elected and thus have an independent mandate from the people, or even whether the incumbent attorney general should be able to bind his successors.” Owen M. Fiss, Against Settlement, 93 Yale L.J. 1073, 1079 (1984).
. After trial, certain Bexar County district judges also sought to intervene as defendants, and we have before us an appeal from the denial of their motion. A motion to intervene under Rule 24 must be timely. Fed.R.Civ.P. 24(a), (b); Jones v. Caddo Parish School Bd., 735 F.2d 923, 926 (5th Cir.1984) (en banc). Although the district court did not expressly state that their motion was untimely, it was well within the district court’s discretion to deny the motion on this ground.
. Because we find that the judges' standing in their individual capacities survives the settlement agreement, we are not required to address the ability of Texas district court judges to represent themselves in their official capacities. It appears, however, that Texas law permits them to do so. Tex. Gov't Code § 74.141, titled Defense of Judges provides:
The attorney general shall defend a state district judge, a presiding judge of an administrative region, or an active, retired, or former judge assigned under this chapter in any action or suit in any court in which the judge is a defendant because of his office as judge if the judge requests the attorney general’s assistance in the defense of the suit.
(emphasis added).
."Since an intervenor is bound by future orders, it may appeal from an appealable order unless the intervention has been specifically limited to forbid it.” Matter of First Colonial Corp., 544 F.2d 1291, 1298 (5th Cir.1977). There is obviously no such limitation on the intervenors’ right to appeal in this case.
. Our conclusion that Defendant-Intervenors continue to have standing in their individual capacities to defend the current method of electing trial judges makes it unnecessary for us to consider their motion to modify their intervention to enable them to do so.
. In the same passage, Wright reminds us that "authority to adopt a consent decree comes only from the statute which the decree is intended to enforce." 364 U.S. at 651, 81 S.Ct. at 373.
. For the same reason, Supreme Court authority does not require a remand. In Turnock v. Ragsdale, 493 U.S. 987, 110 S.Ct. 532, 107 L.Ed.2d 530 (1989), the Court granted the parties’ joint motion to defer further proceedings for the parties to submit a proposed consent decree to the district court. Unlike the case before us, the joint motion in Tumock was a true joint motion; there were no objections. See Ragsdale v. Turnock, 941 F.2d 501, 503 (7th Cir.1991) *848(recounting procedural history). In spite of its label, the Attorney General's motion is far from being a joint motion.
. Art. 5, § 4 provides:
The state shall be divided into at least six supreme court districts, and at least one judge shall be elected from each. The districts and the number of judges assigned to each on the effective date of this constitution are retained, subject to change by law enacted by two-thirds of the elected members of each house of the legislature.
. Art. 5, § 3 provides:
The supreme court shall be composed of a chief justice and six associate justices, four of whom must concur to render judgment. The term of a supreme court judge shall be ten years.
. Louisiana’s first effort to create an eighth position, and thereby resolve the Chisom litigation, came in 1989 in the form of a proposed constitutional amendment. However, the voters rejected the proposal. See La. Const. Art. 5, §§ 4, 35, Historical Notes.
.Art. 5, § 7 provides:
The state shall be divided into judicial districts, with each district having one or more judges as may be provided by law or by this Constitution....
Art. 5, § 7a(i) provides:
The legislature, the Judicial Districts Board, or the Legislative Redistricting Board may not redistrict the judicial districts to provide for any judicial district smaller in size than an entire county except as provided by this section. Judicial districts smaller in size than the entire county may be created subsequent to a general election where a majority of the persons voting on the proposition adopt the proposition "to allow the division of_ County into judicial districts composed of parts of_ County.” No redistricting plan may be proposed or adopted by the legislature, the Judicial Districts board, or the Legislative Redistricting Board in anticipation of a future action by the voters of any county.
. Section 2 reads in full:
(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 4(f)(2), as provided in subsection (b) of this section.
(b) A violation of subsection (a) is established if, based on the totality of the 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.
42 U.S.C. § 1973.
. The Senate Report indicates that “[tlypical factors include”:
1. the extent of any history of official discrimination in the state or political subdivision that touched the right of the members of the minority group to register, to vote, or otherwise to participate in the democratic process;
*8502. the extent to which voting in the elections of the state or political subdivision is racially polarized;
3. the extent to which the state or political subdivision has used unusually large election districts, majority vote requirements, anti-single shot provisions, or other voting practices or procedures that may enhance the opportunity for discrimination against the minority group;
4. if there is a candidate slating process, whether the members of the minority group have been denied access to that process;
5. the extent to which members of the minority group in the state or political subdivision bear the effects of discrimination in such areas as education, employment and health, which hinder their ability to participate effectively in the political process;
6. whether political campaigns have been characterized by overt or subtle racial appeals;
7. the extent to which members of the minority group have been elected to public office in the jurisdiction.
Additional factors that in some cases have had probative value as part of plaintiffs’ evidence to establish a violation are:
whether there is a significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority group.
whether the policy underlying the state or political subdivision’s use of such voting qualification, prerequisite to voting, or standard, practice or procedure is tenuous.
S. Rep. 417 at 28-29, reprinted in 1982 U.S.Code Cong. & Admin.News at 206-07. These factors are derived from our decision in Zimmer v. McKeithen, 485 F.2d 1297 (5th Cir.1973), aff'd sub nom East Carroll Parish School Board v. Marshall, 424 U.S. 636, 96 S.Ct. 1083, 47 L.Ed.2d 296 (1976), as well as White. See S. Rep. 417 at 28 n. 113, reprinted in 1982 U.S.Code Cong. & Admin.News at 206 n. 113.
. See also Chavis v. Whitcomb, 305 F.Supp. 1364, 1376-81 (S.D.Ind.1969).
. The Court stated:
We have discovered nothing in the record or in the court's findings indicating that poor Negroes were not allowed to register or vote, to choose the political party they desired to support, to participate in its affairs or to be equally represented on those occasions when legislative candidates were chosen. Nor did the evidence purport to show or the court find that inhabitants of the ghetto were regularly excluded from the slates of both major parties, thus denying them the chance of occupying legislative seats.
Id., 403 U.S. at 149-50, 91 S.Ct. at 1872.
. In order to make out a § 2 vote dilution claim under Gingles, minority plaintiffs challenging an at-large system must prove that: (1) the group is sufficiently large and geographically compact to constitute a majority in a single-member district; (2) it is politically cohesive; and (3) the white majority votes sufficiently as a hloc to enable it usually to defeat the minority’s preferred candidate. Gingles, 478 U.S. at 50-51, 106 S.Ct. at 2766.
. The dissent contends that we have departed from controlling Supreme Court precedent in requiring plaintiffs to show more than divergent voting patterns among white and minority voters in order to establish legally significant bloc voting. The dissent properly points out that a majority of the Court in Gingles held that racial bloc voting rests on proof that "the white majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate.” Gingles, 478 U.S. at 51, 106 S.Ct. at 2766. As the Court’s recent unanimous decision in Voinovich v. Quitter, - U.S. -, --, 113 S.Ct. 1149, 1157, 122 L.Ed.2d 500 (1993), indicates, this standard is hardly controversial. The Justices in Gingles, however, were sharply divided on the crucial, separate issue of the sort of showing necessary to establish "legally significant” bloc voting — that is, the conditions that enable courts to predict that a majority bloc will consistently "defeat the minority's preferred candidate.” The dissent correctly concludes that the approach taken by Justice White and Justice O’Connor, rather than that offered by Justice Brennan, should govern this second inquiry. Thus, we are in full agreement with the dissent that the possible causes of polarized voting must be examined because "they call into question the consistency with which the white bloc will op*859pose minority-preferred candidates.” Dissent at 911.
As we state in the text, we regard evidence that divergent voting patterns are attributable to partisan affiliation or perceived interests rather than race as quite probative on the question of a minority group's future success at the polls. The dissent, however, while apparently willing to consider other possible non-racial causes, asserts that partisan affiliation is insignificant. We are told, in fact, that "the Voting Rights Act, as interpreted in Gingles and succeeding cases, presupposes partisan voting.” Dissent at 910. This ■refusal to distinguish racial politics from partisan politics strikes us as utterly inconsistent with the unbroken line of authority extending from Whit-comb and White through Justice Marshall's dissent in Bolden and the 1982 amendments to the controlling concurring opinions in Gingles the dissent purports to embrace.
. The facts of this case do not require us to determine whether defendants may attempt to prove that losses by minority-preferred candidates are attributable to non-racial causes other than partisan affiliation. We express no opinion on this entirely separate question.
. Defendant Judge Entz has contended throughout this litigation that § 2, as amended, is unconstitutional. In view of our construction of the statute, we need not reach this question.
. The dissent points out that defendants did not ask the trial court to make a specific finding that black and Hispanic voters were politically cohesive in Harris and Tarrant Counties. This observation, while correct, is beside the point, for that is not the claim they raise on appeal. Rather, defendants argue that the district court improperly refused to consider elections involving His-panic candidates studied by Dr. Taebel, their expert. This question is most assuredly before us and, given the overwhelming evidence of cohesiveness among black and Hispanic voters in Harris and Tarrant Counties, is susceptible to only one answer.
. Two separate Zimmer factors guided the court's inquiry:
1. the extent of any history of official discrimination in the state or political subdivision that touched the right of the members of the minority group to register, to vote, or otherwise to participate in the democratic process;
5. the extent to which members of the minority group in the state or political subdivision bear the effects of discrimination in such areas as education, employment and health, which hinder their ability to participate effectively in the political process;
S. Rep. 417 at 28-29, reprinted, in 1982 U.S.Code Cong. & Admin. News at 206.
. One exception is the 72nd District, which encompasses both Lubbock and Crosby Counties.
. We do not decide this issue. Some appellate court decisions appear to have reviewed the tenuousness of state interests without deference to the underlying district court determinations. See, e.g., Zimmer, 485 F.2d at 1307.
. The twenty-five are Alabama, Arizona, California, Florida, Georgia, Idaho, Indiana, Kentucky, Michigan, Minnesota, Montana, Nevada, New Mexico, New York, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, South Dakota, Tennessee, Texas, Washington, West Virginia, and Wisconsin. Among these states, some appoint a portion of their trial judges, while others hold retention elections after initial selection by contested election. See generally 28 The Council of State Governments, The Book of the States 210-12 (1990) (table 4.4).
Mississippi and Louisiana only recently abandoned the link between jurisdiction and electoral base in order to settle prolonged litigation.
North Carolina allows every elector within a district court's jurisdiction to vote for its judge by holding statewide elections after district-wide primaries. See Republican Party of N.C. v. Martin, 980 F.2d 943, 947 (4th Cir.1992) (holding that Fourteenth Amendment challenge to system by Republican Party is justiciable).
. Texas venue law, as amended in 1983, has been influenced by both Spanish and English principles. Besides protecting civil defendants from inconvenient forums, the rules strive to ensure that local matters are tried in local courts. See generally Joseph W. McKnight, The Spanish Influence on the Texas Law of Civil Procedure, 38 Texas L.Rev. 24, 36-40 (1959); Charles T. Frazier, Jr., Note, Venue Procedure in Texas: An Analysis of the 1983 Amendments to the Rules of Civil Procedure Governing Venue Practice Under the New Venue Statute, 36 Baylor L.Rev. 241, 241-44 (1984).
. A visiting judge may not, however, hear a civil case over the objection of a party. Tex. Gov't Code Ann. § 74.053 (Vernon Supp.1993).
. Another district court was added by the legislature.
. The Houston Lawyers' Association maintains that the district court properly refused to consider elections in which black and Hispanic voters gave their united support to Hispanic candidates. The Association does not deny that black and Hispanic voters uniformly supported the same candidates in virtually every election analyzed, thus making them a cohesive group under our precedents. Rather, it objects to our examination of these races on the more general grounds that “minorities protected by the Voting Rights Act are not interchangeable” and that the success of Hispanic candidates "does not tell us anything about the willingness of white voters to support African American candidates.” These objections, of course, go directly to the rule permitting the aggregation of different racial and ethnic minorities itself rather than its application to the facts of this case. We do not understand either the Association or any other plaintiffs to challenge the validity of this general rule. Nevertheless, as we demonstrate below, the weight of Texas' linkage interest precludes the finding of a § 2 violation even if these Hispanic-white elections are excluded.
.
Table VLB
Indigenous judicial elections (Harris County)
1. candidate listed first prevailed
2. bold indicates black-preferred candidate
3. * indicates race Engstrom studied
4. / indicates victory by black-preferred candidate
Year Court Candidates Race Party
1980 80th District * McAfee Bonner W B R D
" 309th District Zimmerman Hinojosa W H R D
" County Crim Ct 6 Musselwhite Muldrow W B R D
1982 157th District Salazar Powell H W D R /
" 208th District* Routt Arnold B W D R /
" 262d District* Shaver James W B R D
" 281st District * Moore Ward W B R D
" 308th District Robertson Leal W H D R
*882" County Crim Ct 6 Musselwhite Muldrow W B R D
" County Crim Ct 9 Leal Kolenda H W D R
1984 80th District* Powell Berry W B R D
" 178th District* Harmon Jackson W B R D
" 215th District* Chambers Lee W B R D
" 339th District Lanford Salinas W H R D
" County Civil Ct 3 Hughes Hobson W B R D
1986 133d District* McCorlde Plummer W B R D
" 157th District Salazar Wittig H W D R
" 180th District Lykos Guerrero W H R D
" 185th District* Walker Godwin B W D R
" 209th District McSpadden Sanchez W H R D
" 232d District Azios Youngblood H W D R /
" 245th District * Schuble Proctor W B D R /
" 281st District* Moore Berry W B R D
" 308th District Robertson Dodier W H D R /
" County Civil Ct 3 Hobson Hughes B W D R /
" County Crim Ct 3 Duncan Irvin W B D R /
" County Crim Ct 4 Anderson Williams W B R D
" County Crim Ct 9 Leal Powell H W D R /
" County Crim Ct 11 Mendoza Pickren H W D R /
" County Crim Ct 13 Atkinson Fitch W B R D
" County Crim Ct 14 Barclay Fisher W B R D
" County Probate Ct 4 McCullough Lee W B R D
1988 80th District * Powell Berry W B R D
" 133d District* McCorlde Plummer W B R D
" 152d District * O'Neill Fitch W B R D
" 179th District Wilkinson Guerrero W H R D
" 215th District * Chambers Jackson W B R D
" 295th District * Downey Lee W B R D
" 333d District * Wilson Spencer W B R D
" 351st District Salinas Pruett H W D R
. We note that there were five other viable Democratic candidates (Dukakis, Gore, Ge-phardt, Hart, and Simon) in the 1988 primary, so that Jackson could expect to receive only 16.7% of the white vote if that vote were randomly distributed.
. Taebel’s exhibits also include a 1982 county court primary that the black-preferred candidate, Hicks, seemed to win by seven votes. This tabulation, however, was based upon election returns *887prior to a recount under which Hicks apparently lost.
. Since trial, district judges in Jefferson County have filed an amicus brief requesting judicial notice that Davis, an African-American, was elected to the county court in 1990. The amicus brief also notes that black Democrat Overstreet and Hispanic Democrat Morales won a majority of the county’s votes in their respective 1990 races for Court of Criminal Appeals and Attorney General.
. As we stated in the discussion of Jefferson County, supra Part VI.F, we hold that the runoff election subsequent to a primary election is a single election for the purposes of computing the success or failure of the minority-preferred candidate. The victor of the runoff election is the victor of the combined primary/runoff race.