Georgia v. Ashcroft

Opinion for the court filed by District Judge SULLIVAN, in which Circuit Judge HARRY T. EDWARDS joins, and in which Senior District Judge OBERDORFER joins in Parts III.C.2. and III.C.3. Concurring opinion filed by Circuit Judge HARRY T. EDWARDS, in which District Judge SULLIVAN joins. Opinion concurring in part and dissenting in part filed by Senior District Judge OBERDORFER.

SULLIVAN, District Judge.

This is an action for declaratory judgment commenced by the State of Georgia under Section 5 of the Voting Rights Act of 1965, 42 U.S.C. § 1973c (1994) (“Section 5”). The State seeks a declaratory judgment that the redistricting plans passed by the Georgia General Assembly for the United States Congressional seats and the State Senate and House seats do not “have the purpose and will not have the effect of denying or abridging the right to vote on account of race or color” or membership in a language minority group. 42 U.S.C. § 1973c.

The Voting Rights Act imposes weighty obligations on jurisdictions with a history of racial discrimination in their electoral processes.1 Congress enacted the Act with the “firm intent[ ] to rid the country of racial discrimination in voting” by a “complex scheme of stringent remedies.” *30South Carolina v. Katzenbach, 383 U.S. 301, 315, 86 S.Ct. 803, 812, 15 L.Ed.2d 769 (1966). The Georgia General Assembly is well aware of its statutory and constitutional responsibilities, as the State has spent much of the last decade defending its legislative reapportionment plans against claims of racial gerrymandering, brought pursuant to Section 2 of the Voting Rights Act and the United States Constitution. See, e.g., Abrams v. Johnson, 521 U.S. 74, 117 S.Ct. 1925, 138 L.Ed.2d 285 (1997) (affirming court-ordered Congressional redistricting plan); Miller v. Johnson, 515 U.S. 900, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995) (holding that Congressional redistricting plan violated equal protection clause).

The State’s obligations under Section 5, however, differ significantly from those under Section 2 of the Voting Rights Act.2 Section 5 requires specific jurisdictions to. comply with “preclearance” procedures before implementing any new “voting qualification or prerequisite to voting, or standard, practice, or procedure with respect to voting.”3 42 U.S.C. § 1973c. Two avenues for preclearance are provided by the Act. Id. The covered jurisdiction may seek a declaratory judgment from a three-judge District Court for the District of Columbia that the new practice does not have the purpose or effect of denying or abridging the right to vote on account of race or color. Id. In the alternative, the jurisdiction may submit its proposed procedures to the Attorney General for approval; the procedures are deemed approved if, after 60 days following the filing of a completed submission, the Attorney General has not raised any objections to the proposed procedures. Id.

The Supreme Court has characterized Section 5 as “an unusual, and in some aspects a severe, procedure for insuring that states would not discriminate on the basis of race in the enforcement of their voting laws.” Allen v. State Board of *31Elections, 393 U.S. 544, 556, 89 S.Ct. 817, 22 L.Ed.2d 1 (1969). Section 5 was intended to provide an efficient and rapid mechanism for preclearing changes in voting procedures, while expressly providing that such preclearance in no way affects the ability of individuals to challenge that plan on other grounds. Id. at 549, 556, 89 S.Ct. 817.

Section 5 prohibits States from diminishing the opportunities of African American voters to exercise their electoral power. Beer v. United States, 425 U.S. 130, 141, 96 S.Ct. 1357, 1364, 47 L.Ed.2d 629 (1976). Georgia has demonstrated that African American voters increasingly have been able to make their voices heard at the ballot. The record indicates, however, that there are areas within the State where racially polarized voting persists. In these areas, white voters consistently vote against the preferred candidates of African Americans in local and district elections, so the strength of African American votes rests in substantial part on the sheer numbers of African American voters in a district. Where there is evidence of racially polarized voting, a redistricting plan that reduces African American votes in a district with no offsetting gains elsewhere raises the specter of impermissible retrogression. In this situation, the State is hard-pressed to demonstrate that there has been no “backsliding” in African American voting strength. Reno v. Bossier Parish School Bd., 528 U.S. 320, 335, 120 S.Ct. 866, 145 L.Ed.2d 845 (“Bossier II”). And such a failure is fatal in a Section 5 case, because the burden is on the State to show that the redistricting plan will not adversely affect the opportunities of African American voters to effectively exercise their electoral franchise. Beer, 425 U.S. at 141, 96 S.Ct. 1357.

After carefully reviewing the evidence in the record before us, we hereby grant a declaratory judgment that the United States Congressional redistricting plan, Act No. 2EX11, and the State House redistricting plan, Act No. 2EX23, satisfy the requirements of Section 5. We hold, however, that the State of Georgia has not met its burden of proof under Section 5 with regard to the State Senate redistricting plan. The State has not demonstrated by a preponderance of the evidence that the State Senate redistricting plan does not have the purpose and will not have the effect of denying or abridging the right to vote on account of race or color. Accordingly, the State’s request for a declaratory judgment that the State Senate plan meets the requirements of Section 5 is denied.

I. Procedural History and Preliminary Matters

In this case, the State has foregone the option of applying to the Attorney General for preclearance of its redistricting plans, and has filed suit in this court. Section 5 essentially freezes the existing districting plans in Georgia unless and until a declaratory judgment is obtained from this court that the proposed reapportionment plans are without discriminatory purpose or effect. Reno v. Bossier Parish School Bd., 520 U.S. 471, 477, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997) (“Bossier I”). The State of Georgia filed suit on October 10, 2001, requesting that the court enter a declaratory judgment that the Congressional, State House and State Senate plans do not have a discriminatory purpose or effect. Georgia’s general primary is scheduled for July 16, 2002, and the State has recently received preclearance to allow candidates for Congress and the state legislature to qualify for the primary from June 10 to June 21, 2002. See Defs.’ Opp’n to Mot. for Expedited Trial at 6 (discussing November 26, 2001 preclearance of Georgia Act 2EX 10 (2001)). In light of *32the “extraordinary” remedy mandated by the Voting Rights Act, the court has acted with all possible speed to expeditiously resolve this matter. Allen, 393 U.S. at 563, 89 S.Ct. 817.

The court’s scheduling order set a demanding briefing schedule, while permitting the parties to engage in extensive discovery up until the commencement of the trial. See Order, Civ. Action No. 01-2111, Dec. 20, 2001. Indeed, with the consent of the parties, commencement of the trial was deferred for three days to enable the parties to complete discovery.

At the time of the initial scheduling conference, the United States had not yet identified its position with respect to each of the submitted plans. Upon consideration of a motion by the State of Georgia, a response thereto and oral argument at the scheduling conference, the court required the United States to identify its legal position by no later than December 31, 2001.

Two motions to intervene were filed early in the proceedings, one by four African American citizens of Georgia, Patrick Jones, Roielle Tyra, Della Steele and Georgia Benton (“Jones”), and one by Michael King, an African American lawyer and resident of Senate District 44. Both motions were denied without prejudice following the court’s order that the United States identify its legal position. Order, Civil Action No. 01-2111, Dec. 20, 2001; Order, Civil Action No. 01-2111, Dec. 21, 2001. The court invited the movants to file amicus curiae briefs, but held that, without clarification of the United States’ legal position, it could not determine if the existing parties adequately represented the interests of the putative intervenors. See Fed.R.Civ.P. 24(a).

On December 31, 2001, the United States identified its position with respect to the proposed redistricting plans. On January 4, 2002, Jones filed a renewed motion to intervene. After receiving a response and reply to this motion, on January 10, 2002, the court granted Jones’ motion to intervene and required the inter-venors to comply with the Court’s initial scheduling and pretrial order. See Order, Civil Action No. 01-2111, Jan. 10, 2002 (granting intervention as to State House and State Senate plans); Order, Civil Action No. 01-2111, Jan. 30, 2002 (granting intervention as to Congressional redistricting plans). As discussed below, Mr. King renewed his motion to intervene in an untimely fashion, and was denied leave to intervene.

With the consent of the parties, Judge Sullivan presided over the four-day trial.4 Following the conclusion of the trial, the parties submitted proposed findings of fact and conclusions of law, and post-trial mem-oranda of law. Closing arguments were heard by the three-judge panel on February 26, 2002.

There are several preliminary matters that the court must address before focusing on the three reapportionment plans. Pending before the court are: (1) a motion for leave to file an amicus curiae brief submitted by the American Civil Liberties Union (“ACLU”); (2) plaintiffs motion to strike the Jones intervenors for lack of standing; (3) Mr. King’s motion to stay proceedings and motion for reconsideration of his motion to intervene; and (4) *33defendants’ and intervenors’ motions to exclude portions of plaintiffs expert testimony.

A. ACLU’s Motion for Leave to Participate as Amicus Curiae

The ACLU has moved the court for leave to participate as amicus curiae in this case in support of defendants’ position that the Senate Plan is retrogressive. The court is of the opinion that the limitations on amicus filings outlined by the Seventh Circuit in National Organization for Women v. Scheidler, 223 F.3d 615 (7th Cir.2000), are applicable here. The ACLU has presented no unique information or perspective that can assist the court in this matter, and seeks only to make additional legal arguments on behalf of the United States, a more than adequately represented party. Accordingly, the court denies the ACLU’s motion for leave to file an amicus curiae brief.

B. Standing of Jones Intervenors

The State of Georgia has challenged in-tervenors’ standing to contest the reapportionment plans. In the context of a Section 2 challenge, the Supreme Court has stated that “[w]here a plaintiff resides in a racially gerrymandered district, ... the plaintiff has been denied equal treatment because of the legislature’s rebanee on racial criteria, and therefore has standing to chabenge the legislature’s action.” United States v. Hays, 515 U.S. 737, 744-45, 115 S.Ct. 2431, 132 L.Ed.2d 635 (1995).

The State argues that two of the four individual intervenors reside in a benchmark Senate District disputed by the parties, but would be removed from this District under the Senate redistricting plan. Nevertheless, whether intervenors reside in the proposed or benchmark districts at issue in this matter does not affect their standing for purposes of challenging the redistricting plans as retrogressive. The plans are statewide and the drawing of one district’s boundaries necessarily affects neighboring districts. Furthermore, the removal of intervenors from a majority-minority district is sufficient to provide intervenors with standing to challenge the proposed district.

The State also raises concerns that intervenors’ interests in the btigation may diverge from the statements of counsel. Two of the intervenors appeared to testify at their depositions that they would prefer to reside in a majority-white, Repubbcan district. This conflicts with their counsel’s representations that intervenors are harmed by a decrease in overall minority voting strength caused by reductions of minority population in the districts. However, we are reluctant to strike interve-nors solely on the basis of this alleged contradiction. Intervenors’ sworn declarations clearly allege an injury caused by diminution of minority voting strength. Deposition testimony may cast doubt on the extent of that injury, but it does not eliminate intervenors’ standing.

C.King Motion for Reconsideration of Motion to Intervene

Pending before the court are two motions filed by putative intervenor, Michael B. King. King is an African American attorney and registered voter who resides in Georgia Senate District 44. King is proceeding pro se and first sought to intervene on December 19, 2001.5 On Deeem-*34ber 21, 2001, King’s motion was denied without prejudice because the United States had not yet identified its position with respect to the redistricting plans in question and the court was unable to determine whether King’s interests would be adequately represented by the existing parties. See Fed.R.Civ.P. 24(a). However, King was granted permission to file an amicus curiae brief by no later than January 14, 2002.6 Chambers manually faxed King a copy of the court’s December 21, 2001 order.7

An initial scheduling and pretrial order in this matter was issued on December 20, 2001. This order set forth a series of deadlines designed to expedite trial proceedings. In particular, the order required defendants to identify their position with respect to the redistricting plans by no later than December 31, 2001. Chambers manually faxed a copy of the initial scheduling and pretrial order to King, in light of the expedited nature of the proceedings, King’s failure to subscribe to the automated faxing program of the Clerk’s office and his representation to chambers’ staff that he had not viewed the filings in this case.

On January 15, 2002, King filed a renewed motion to intervene. On January 16, 2002, this court issued a scheduling order directing the parties to file any and all responses to King’s motion by no later than January 17, 2002 at noon, and ordering that any and all replies be filed by no later than January 18, 2002 at noon. Also on January 16, 2002, chambers contacted the parties in this matter and attempted to contact King in order to inform them of the contents of the court’s order. Counsel for the parties indicated that they had not received a copy of King’s renewed motion to intervene. King’s telephone answering system was full on January 16 and on the following two days. In light of these circumstances, chambers faxed a copy of the scheduling order to King on January 16, 2002, and faxed a copy of King’s renewed motion to plaintiffs counsel for distribution to all parties.8

On the evening of January 18, 2002, King contacted chambers and stated that he had received the facsimile copy of the January 16, 2002 order. He noted that the deadline for his submission had passed and indicated that he had just received the order. At no future date did King file any *35reply to the parties’ responses to his motion to intervene.

As is evident from our initial scheduling and pretrial order, this matter was scheduled to proceed to trial on February 1, 2002. Pretrial statements, expert reports and direct testimony were all filed by January 18, 2002, and the pretrial conference was scheduled for January 25, 2002. On January 23, 2002, the court issued an order rescheduling the pretrial conference from 10:00 a.m. to 12:00 p.m. on January 25, 2002, and faxed King a copy of this order. King did not appear at the pretrial conference, nor did he communicate in any way with chambers concerning the expedited schedule for this matter. On January 30, 2002, this court denied King’s motion to intervene without prejudice. This denial was based on King’s “failure to appear at the January 25, 2002 pretrial conference, of which he had notice, his consistent failure to communicate with Chambers and with counsel for the parties in this matter, and the expedited nature of these proceedings.” Order, Civil Action No. 01-2111, Jan. 30, 2002.

King sought to intervene as a matter of right pursuant to Fed.R.Civ.P. 24(a). However, any application for intervention must be timely. See National Ass’n for Advancement of Colored People v. New York, 413 U.S. 345, 93 S.Ct. 2591, 37 L.Ed.2d 648 (1973) (“NAACP”). In NAACP, the Supreme Court discussed the legal standard for considering the timeliness of a motion to intervene in an action for declaratory judgment brought pursuant to Section 4(a) of the Voting Rights Act. 413 U.S. at 365-66, 93 S.Ct. 2591. The Court held that the three-judge court properly exercised its discretion in determining from all of the circumstances that a motion to intervene was untimely. Id. (“Although the point to which the suit has progressed is one factor in the determination of timeliness, it is not solely disposi-tive. Timeliness is to be determined from all the circumstances.”). King “failed to protect [his] interest in a timely fashion” by repeatedly failing to communicate with the court, to keep apprized of the case and to comply with local filing requirements. Id. at 367, 93 S.Ct. 2591. King knew or should have known that the proceedings were subject to expedited review.

King’s legal interest has not been adversely affected by the Court’s denial of his motion to intervene. Id. at 368 (considering ability of movant to take future action to protect interests). This court’s denial of King’s motion to intervene in no way forecloses his ability to challenge Georgia’s senate reapportionment plan. We have consistently stressed the expedited nature of our review of this matter. In denying King’s renewed motion to intervene, the court was mindful of the fact that King’s failure to act in a timely matter had “the potential for seriously disrupting the State’s electoral process.” Id. at 369, 93 S.Ct. 2591 (discussing Section 4 proceedings).

On February 7, 2002, King filed a motion to stay proceedings in this court, a motion for reconsideration of his motion to intervene, and a notice of appeal to the Supreme Court. The motion for a stay referred to a “January 4, 2002 hearing,” which King moved to stay pending the court’s consideration of his motion for reconsideration and his appeal. No hearing was scheduled in this matter on January 4, 2002. The court commenced trial in this case on February 4, 2002 and it is possible that King intended to request a stay of the trial. Nevertheless, the motion was filed on February 7, 2002, the last day of the four-day trial.

Plaintiff argues that, upon receipt of the notice of appeal, this court lost jurisdiction to consider King’s motions. *36We agree. The Supreme Court has provided clear direction as to how to proceed when a defendant simultaneously files a notice of appeal with the appellate court and a motion for reconsideration with the district court. “The filing of a notice of appeal is an event of jurisdictional significance — it confers jurisdiction on the [appellate court] and divests the district court of its control over those aspects of the case involved in the appeal.” Griggs v. Provident Consumer Discount Co., 459 U.S. 56, 58, 103 S.Ct. 400, 74 L.Ed.2d 225 (1982); accord United States v. DeFries, 129 F.3d 1293 (D.C.Cir.1997). While it is clear to the court that King’s renewed motion to intervene and motion to stay are as untimely as his previous motion, King’s notice of appeal divested this court of jurisdiction to consider King’s motions.

D. Motion to Exclude Dr. Epstein’s Testimony

The United States and intervenors seek to exclude the testimony of the State’s expert, Dr. David Epstein, regarding his analysis of white crossover voting in the benchmark Senate Districts, see U.S.Ex. 122, and his conclusion that it was proper to assess voting trends on a statewide, rather than a regional basis. Although the testimony was provided at the eleventh hour to the United States, the United States was able to cross-examine Epstein with respect to his analysis. The cross-examination of Epstein effectively highlighted problems with Epstein’s conclusion that there was no statistically significant variation in the degree of white crossover. Furthermore, the court offered the United States an opportunity to reopen cross-examination in order to permit the United States’ expert to assist counsel in the cross-examination. The United States declined this opportunity. The court finds that the introduction of Epstein’s calculations of white crossover voting are not unduly prejudicial. Epstein relies on the table in question only for the limited conclusion that a statewide probit analysis was proper; neither he nor the State suggests that the table is reliable evidence of white crossover voting in the Senate Districts.

Intervenors urge this court to strike Epstein’s testimony on the basis that it is not competent expert testimony. They argue that Epstein’s probit analysis does not represent reliable or relevant evidence. However, Epstein testified that probit analysis is a standard statistical methodology. The court finds that Epstein’s report is reliable and relevant evidence.

II. Findings of Fact

At first glance, the evidentiary record in this matter appears extensive. Yet, considering that the State has chosen to present three statewide reapportionment plans to the court, the record in fact is rather slim. The State of Georgia, the United States and intervenors have all contributed to the evidentiary record before the court. The State introduced statistical data on the existing and proposed districts, including political performance, total population and voting age populations, as well as break-downs of that data by race. The State relied on the testimony of two expert witnesses, State legislators, United States Representative John Lewis from Georgia, and the director of the redistricting office in Georgia, Linda Meggers. In response, the United States presented the court with a greater amount of and more detailed evidence, including voter registration data, precinct-level information, data and maps demonstrating exactly how district lines would be redrawn by the proposed plans, and testimony of numerous social leaders and local elected officials from the contested districts. The United States also provided the only expert report that consid*37ered the prevalence of racially polarized voting. But the United States’ evidence was extremely limited in scope — focusing only on three contested districts in the State Senate plan. That evidence was not designed to permit the court to assess the overall impact of each of the three plans. Finally, while intervenors challenged all three plans, they present little evidence other than proposed alternative plans and an expert report critiquing the State’s expert report.

A. Reapportionment and Elections in Georgia: Background

The Georgia General Assembly has plenary authority under the Constitution and laws of the State of Georgia to enact, subject to the approval or veto of the Governor, legislation to reapportion the State Senate and House of Representatives, as well as of Georgia’s designated number of seats in the U.S. House of Representatives. Ga. Const., Art. Ill, § II, ¶ II; O.C.G.A. §§ 21-2-3, 21-2-4, 28-2-1; 28-2-2. In fact, Georgia’s State Constitution mandates that the General Assembly reapportion the Senate and the House of Representatives as necessary after each United States decennial census. Georgia Const., Art. Ill, § II, 2. The State Constitution further provides that the districts shall be composed of contiguous territory. Id.

The current United States Congressional districts are the result of a court-drawn remedial map, which was put in place after a legislative impasse and a court decision that two Congressional districts were unconstitutionally based on race in the effort to increase their minority population percentages. Abrams v. Johnson, 521 U.S. 74, 117 S.Ct. 1925, 138 L.Ed.2d 285 (1997). This court-drawn remedial map is the benchmark plan9 for this court’s consideration of plaintiffs complaint for declaratory judgment. Pl.’s Proposed Findings of Fact & Conclusions of Law (“PPFF”^ 53.

The current State House and State Senate plans were put in place as a result of a mediated plan, which was adopted by the Georgia General Assembly in 1997. Johnson v. Miller, 929 F.Supp. at 1561-67; PPFF ¶ 68. Pursuant to a settlement agreement in a section 5 lawsuit, both of the current plans were submitted to the Department of Justice and were pre-cleared on April 29, 1997. See DOJ File No. 95-3656 (granted as reconsideration of original 1995 submission); PPFF ¶ 68. Since the adoption of the Senate plan in 1997, it has been amended with minor changes three times and those amendments were precleared by the U.S. Department of Justice on April 23,1998 (DOJ File No. 98-98-0759 and 98-0912), September 20, 1999 (DOJ File No.1999-0989) and August 28, 2000 (DOJ File No.2000-2682). Id. ¶ 69. Since the adoption of the House plan in 1997, it has been amended once. Id. That change was precleared by the U.S. Department of Justice on April 23, 1998 (DOJ File No. 98-0759 and 98-0912). Id. The 1997 plans, as amended, constitute the benchmark plans for the State House and State Senate redistricting plans submitted to the court for consideration.

Following the 2000 national census, the Georgia legislature enacted redistricting plans. These plans are intended to take effect by the time of the next general election day, scheduled for Tuesday, November 5, 2002, at which time Georgia voters will elect candidates to the United *38States Congress and the Georgia General Assembly. O.C.G.A. ¶ 21-2-9; PPFF ¶ 16. In the upcoming election cycle for Congress and the Georgia General Assembly, candidates for partisan offices will qualify for either the Democratic Party or Republican Party nomination between 9:00 а.m. on June 19, 2002 and 12:00 p.m. on June 21, 2002. O.C.G.A. § 21-2-153(c)(l); PPFF ¶ 13. The primary for nomination to partisan office in Georgia will next be held on August 20, 2002. O.C.G.A. § 21-2 — 150(b)(1); PPFF ¶ 14. Any run-off election necessary after the August 20 primary election will be held on Tuesday, September 10, 2002. O.C.G.A. § 21-2-501(g); PPFF ¶ 15.

In Georgia, a candidate seeking nomination to a state or federal office in a regular partisan primary must receive a majority of the votes cast in the primary or in the primary run-off election. O.C.G.A. § 21-2-501(a); PPFF ¶ 18. A candidate is elected to office in a regular general election upon receipt of a 45% plurality vote. O.C.G.A. §§ 21 — 2 — 501(g), 21-2-2(22); Id. ¶ 19. In the event no candidate receives such a plurality, a runoff election is then held 21 days later. Id.

B. Georgia Demographics

The 1990 census showed that the total population of the State of Georgia was б,478,216 persons. PPFF ¶ 21. The 1990 census also showed that 1,746,565 persons in Georgia, or 26.96%, identified themselves as black.10 Id. 22. 1,737,165 persons, or 26.82% identified themselves as non-Hispanic and black only. Id. The 1990 census reflected that black voting age population (“BVAP”) was 1,168,142, or 24.58% of the total voting age population (“VAP”) in the state. Id. ¶¶ 25, 26. The total VAP of non-Hispanic individuals identifying themselves as black only was 24.46% of the total VAP. Id. ¶ 26.

The 2000 decennial census shows that the total population of the State of Georgia has increased by 1,708,237 individuals since 1990, and is now 8,186,453. PPFF ¶ 27. There are 6,017,219 people in Georgia who are of voting age. Id. ¶ 31.

The 2000 census allowed individuals for the first time to identify themselves as more than one race. As a result of this change, the parties dispute the proper calculation of the African American population of Georgia. In the 2001 special redistricting session, the State of Georgia defined “black” as including non-Hispanic and Hispanic black persons of a single race, and “black combo” as all persons who identified themselves as black in combination with any other racial or ethnic category on the 2000 census form. U.S.Ex. 702, 18:13-24. Consequently, for purposes of this matter, Georgia has counted its black population as including all black multi-racial Hispanic and non-Hispanic responses. In contrast, the Department of Justice, in accordance with a Guidance issued by the Department in January, 2001, has counted as black those non-Hispanic individuals who identify as black only, or as black and white, but not individuals who identified as black and another minority race. See Guidance Concerning Redistricting and Retrogression Under Section 5 of the Voting Rights Act, 42 U.S.C. § 1973c, 66 Fed.Reg. 5, 411 (Jan. 18, 2001).11

*39 Total Population

The 2000 census showed that 2,349,542 residents of Georgia identified themselves as black only, representing 28.7% of the total state population. PPFF ¶27. The total non-Hispanic population identifying themselves as black only represents 28.5% of the total state population. Id. 0.22% of the total population identified themselves as black and white only, and 0.21% of the total population identified themselves as non-Hispanic and black and white only. Id.

The total population of Georgia identifying as black and only one other racial category was 0.47% of the total population; and 0.20% of the population identified as black and one other non-white racial category. Id. ¶ 29. An additional 0.07% of the total population identified as black and more than one other racial category. Id. ¶ 30.

Total Voting Age Population

The total voting age population (“VAP”) of Georgia identifying themselves as black only was 1,602,985 people, or 26.64% of the total VAP. Id. ¶ 32. The voting age population of non-Hispanic individuals identifying themselves as black only is 26.52% of the total VAP. U.S. Proposed Findings of Fact & Conclusions of Law (“USPFF”) ¶ 124.12 The VAP identifying as black and white represents 0.08% of the VAP; the total non-Hispanic black and white VAP is 0.07% of the total. PPFF ¶ 34. Those identifying themselves as black and one other racial category constitute 0.28% of the VAP, id. ¶ 35, while those identifying themselves as non-Hispanic black and one other racial category comprise 0.25% of the VAP. Id. Those identifying themselves as black and more than one other racial category represent 0.05% of the VAP. Id. ¶ 36.

Registered voters may identify their race when they register to vote. However, unlike the 2000 census, voters are not permitted to identify as more than one race. At the time of the November 1992 general election there were 3,177,061 people registered to vote in Georgia; of these, 698,305 people, or 21.97%, identified themselves as black. Id. ¶ 37. In November 1994, 3,003,527 people were registered to vote in Georgia, 21.92% of whom identified as black. Id. ¶ 38. In November 1996, 24.38% of persons registered to vote identified themselves as black, id. ¶ 39, and, in November 1998, 24.73% of persons registered to vote identified as black. Id. ¶ 40. At the time of the November 2000 general election, 3,856,676 persons registered to vote in Georgia, of which 980,587, or 25.42%, identified themselves as black. Id. ¶ 41.

C. The 2001 Reapportionment Process

During the 2001-02 session of the General Assembly, there were 180 seats in the Georgia House of Representatives and 56 seats in the Georgia Senate. During the 2001-02 session of the Georgia General Assembly, 34 of the 180 Representatives and 11 of the 56 Senators were African American. All of the African American Representatives and Senators are Democrats. Eric Johnson dep. at 17:13-16; Lynn Westmoreland dep. at 17:18-24.

On March 22, 2000, the results of the 2000 decennial census for Georgia became generally available from the Census Bureau, prompting the General Assembly to act. PPFF ¶ 70. The Senate formed a *40committee to address the issues of reapportionment.13 During the 2001-02 session, the chairman, vice-chairman and secretary of the Senate Reapportionment Committee were Senators Tim Golden, Robert Brown and Hugh Gillis respectively. Id. at ¶ 6. During the 2001-02 session of the General Assembly, that Committee had 24 members, six of whom were African American. Id. at ¶ 7. Those six were Vice-Chairman Robert Brown of the 26th district, Senator Ed Harbison of the 15th, Senator David Scott of the 36th, Senator Nadine Thomas of the 10th, Senator Regina Thomas of the 2nd and Majority Leader Charles Walker of the 22nd. Id.

The House of Representatives also formed a committee to address the issues of reapportionment. The chairman, vice-chairman and secretary of the House Legislative and Congressional Reapportionment Committee were Representatives Tommy Smith, Jay Shaw and Carl Von Epps, respectively. Id. ¶ 8. During the 2001-02 session of the General Assembly, the House Reapportionment Committee has 29 members, six of whom were African American. Id. ¶ 9. Those six were Carl Von Epps of the 131st district, the Secretary of the Reapportionment Committee and Chairman of the Legislative Black Caucus, David Lucas of the 124th, Lester Jackson of the 148th, Arnold Ragas of the 64th, Kasim Reed of the 52nd, and LaNett Stanley-Turner of the 50th. Id.

Prior to the 2001 special sessions to consider reapportionment issues, the House and Senate Reapportionment Committees held joint hearings as follows: April 17 — Watkinsville; April 18 — Atlanta; April 25 — Augusta; April 30 — Perry; May 10 — Brunswick; May 15 — Valdosta; May 23 — Dahlonega. Id. ¶ 71.14

Also prior to the 2001 special session, the House and Senate Reapportionment Committees adopted guidelines providing for public access to committee hearings and meetings, public access to redistricting data and materials and general guidelines for the presentation and introduction of plans to the committees. Id. ¶ 72.

The Senate Reapportionment Committee met formally on April 12, June 28, July 12, August 1, August 6, August 9, August 27, August 28, September 4, September 7 and September 13, 2001. Id. ¶ 73. Transcripts or records of those proceedings were provided to the Department of Justice as a part of this litigation. Id. The House Legislative and Congressional Reapportionment Committee met on April 11, June 29, July 10, July 20, July 24, July 26, July 31, August 13, August 14, August 16, August 22, August 28, August 29, September 4 and September 10. Id. Transcripts or records of those proceedings were provided to the Department of Justice. Id.

On June 21, 2001, Governor Roy Barnes issued a proclamation calling the General Assembly into special session for purposes of reapportioning the State Senate and House of Representatives.15 Id. ¶ 74. *41This first special session of the General Assembly began on August 1 and ended on August 17, 2001. Id.

Linda Meggers has worked with Georgia’s Legislative Redistricting Office since 1971. Id. ¶ 76. She worked full-time with the office since 1973, and has served as the Director since 1978. Id. She is intimately familiar with the demographics, changing demographics, and political geography of the entire state. Pl.Ex. 22 at 10-16. Meggers provided direct testimony, and she was twice deposed by the United States.

Ms. Meggers gave an overview of the 2001 reapportionment process, testifying that there were significant differences in this redistricting process compared with past years. Id. The data and technology available in 2001 allowed for sophisticated analysis of the political performance of prospective districts:

Political geography is exactly what we are talking about here, new Senate districts, congressional districts, House districts; that’s political geography. So we could draw a proposed House district, House piece of geography, and have the census data, and immediately analyze it politically. If that district had existed in 1996 or 1998, this is how it would have voted in these particular elections, where we had the data.

Id. at 9. Data was available to assess whether districts tended to vote for Democrats or Republicans in past elections. Id. Ms. Meggers testified that, in contrast to past elections, political performance data was used extensively in the 2000 redistricting process. Id. at 17.

The State Senate redistricting plan before this court was approved by the Georgia Senate on August 10, 2001, and by the Georgia House of Representatives on August 17, 2001. The State House redistricting plan was passed by the Georgia House of Representatives on August 29, 2001 and by the Georgia Senate on September 6, 2001. The Congressional redistricting plan was passed by both houses of the Georgia General Assembly on September 28, 2001. No Republicans in either the House or Senate voted for any of these reapportionment plans. (Eric Johnson dep. at 27). All of the African American legislators in the Georgia General Assembly are Democrats, PPFF ¶ 5. With the exception of one African American Representative and one African American Senator who voted against the State House and State Senate redistricting plans, African American legislators voted for the redistricting plans.

Ms. Meggers testified that the goal of the Democratic leadership in the Senate and House was two-fold:

To maintain the number of minority districts that we presently had, but at the same time maintain and increase the number of Democratic seats that they had in the House and the Senate. They knew that they couldn’t just maintain what they had, they actually needed to strengthen those majorities if they were to maintain a majority over the decade. When I say Democratic leadership, you need to understand that the Black Caucus members and the Black Caucus leadership were very involved in that. They are very much a part of the leadership when we talk about this. So, they wanted to maintain those districts, but not waste, is the — I guess the term I heard often, waste their votes.

*42PLEx. 22 at 20-21. One of the reasons given by African American senators for aligning their interests with those of the Democratic Party was that, should the Democratic Party cease to be in the majority in the State House and State Senate, all existing African American chairs of committees would be lost. C. Walker dep. at 94. Senator Robert Brown, an African American from Senate District 26, was Vice Chairman of the Senate Reapportionment Committee overall, and was the chairperson of the subcommittee that did the Senate Plan itself. Pl.Ex. 20 at 23. According to Senator Brown, there are 11 African Americans in the state Senate, and 7 to 8 of that number currently could chair committees. Id. at 18-24. The majority leader of the Senate is an African American, and the chairman of the rules committee, Senator Brown, is also African American. Id.

D. United States Congressional Redistricting

After each decennial census, the United States House of Representatives is reapportioned to reflect population changes in the states. After the 1990 census, the State of Georgia was assigned 11 seats pursuant to that reapportionment. PPFF ¶ 101; Pl.Exs. 8A, 8C. The State of Georgia then had the responsibility to redistrict to reflect those 11 seats. As discussed above, that redistricting was subject to litigation that resulted in a court-ordered redistricting plan. Abrams v. Johnson, 521 U.S. at 82-85, 117 S.Ct. 1925. The court-ordered plan that resulted from that litigation is the benchmark plan before this court. PPFF ¶ 102.

Following the 2000 census, Georgia was apportioned 13 seats in the United States House of Representatives. Id. ¶ 109. A Conference Committee of the Georgia House of Representatives and Senate produced the Congressional redistricting plan submitted to this court for preclearance. Pl.Ex. 20 at 28 (Brown test.). The Conference Committee had six members, three from the Senate and three from the House of Representatives. Of these six, two Senators and one Representative are African American. Id. at 27.

1. The Benchmark Plan

According to the 2000 census results, the percentages of black population (“BPOP”), black voting age population (“BVAP”), and black registration (“BREG”) for each of Georgia’s existing Congressional districts under the benchmark plan are as follows:

% BPOP % BVAPCGa.) 16 b BREG

District 1 31.65 28.97 26.23

District 2 40.85 37.38 35.68

District 3 31.27 28.62 26.69

District 4 50.60 46.24 49.13

District 5 63.57 58.85 60.31

District 6 11.39 10.80 9.23

District 7 18.66 16.88 15.99

District 8 32.66 30.28 27.65

District 9 3.40 3.12 2.53

District 10 39.00 36.12 33.72

District 11 15.01 13.64 12.1

Pl.Exs. 8D, 8E. In the current Congressional plan, there are two districts with over 50% total black population, the Fourth and Fifth Districts, but only one district, the Fifth District, with over 50% BVAP and black voter registration. Id. However, the Fourth District, has over 45% BVAP and black voter registration. Id.

Notwithstanding the fact that the Fifth Congressional District is the only existing majority African American district in terms of voting age population and registered voters, the State of Georgia is currently represented by three African American Congresspersons: Sanford Bishop (Second District), Cynthia McKinney (Fourth District) and John Lewis (Fifth *43District). Pl.Ex. 21 at 2; Tr., 2/4/02, p.m. at 59-61.

Congressman John Lewis currently represents Georgia’s Fifth Congressional District in the United States House of Representatives. Pl.Ex. 21 at 2:5-6. Congressman Lewis has been one of this country’s leading civil rights advocates for the past 50 years, and his actions, along with those of Dr. Martin Luther King, Jr., were instrumental in achieving passage of the Voting Rights Act of 1965. Id. at 3, 12.

Cynthia McKinney, the current African American Congresswoman for the Fourth Congressional District, was first elected in 1994 in the Eleventh Congressional District when the BVAP of that district was 60%. Tr., 2/4/02, p.m. at 61. Based upon the remedial map drawn by the three-judge court, see Abrams v. Johnson, 521 U.S. at 83-85, 117 S.Ct. 1925, Ms. McKinney ran for election in the Fourth Congressional District, and was successful in 1996, 1998, and 2000 when the BVAP of the district was 33%, 39% and 45%, respectively. Pl.Ex. 25, App.; PPFF ¶ 107.

Sanford Bishop, the current African American Congressman for the Second Congressional District, was first elected in a district in which the total BVAP was 52%. Tr., 2/4/02, p.m. at 59-60. Following the redrawing of those district lines by the federal court in 1996, Congressman Bishop won reelection in the Second District when the total BVAP was between 35% and 37%. Pl.Ex. 25, App. In the course of his political career, Congressman Bishop has won reelection to Congress on three separate occasions in a rural majority-white district. PPFF ¶ 108.

By virtue of the reapportionment mandated by the 2000 census results, the State of Georgia’s representation in the United States House of Representatives was increased to 13 seats. Id. ¶ 109. Based upon a total statewide population of 8,186,-453 people, and the assignment of 13 seats in the United State House of Representatives for the State of Georgia, the ideal size of a Congressional district for purposes of adherence to the principle of one person, one vote is 629,727 people. Id. ¶110.

Based upon the population statistics reported in the 2000 census, all of Georgia’s existing 11 congressional districts have populations larger than the ideal district size of 629,727, and are thus out of apportionment. The percentages by which the current districts exceed the ideal district size are as follows:

District 1: 9.92%

District 2: 3.28%

District 3: 24.13%

District 4: 18.33%

District 5: 2.82%

District 6: 49.51%

District 7: 19.44%

District 8: 5.25%

District 9: 29.32%

District 10: 5.16%

District 11: 32.82%

Id. ¶ 111.

2. The Proposed Plan

During its second special session, the Georgia General Assembly enacted Senate Bill 1EX2, which set forth the reapportionment plan for Georgia’s 13 new congressional districts. See Pl.Ex. 9A; PPFF ¶ 112. Senate Bill 1EX2 was adopted by the Georgia State Senate on September 28, 2001, by a vote of 30 to 23. Id. ¶ 114. No member of the Senate Legislative Black Caucus voted against the plan. Id. The bill was adopted by the Georgia House of Representatives on the same day, by a vote of 99 to 59. Id. ¶ 113. No member of the House Legislative Black Caucus voted against the plan. Id. The Governor of Georgia.signed Senate Bill 1EX2 into law on October 1, 2001, as Act No. 2EX11. Id. ¶ 115.

*44Under the proposed Congressional re-distrieting plan, the total population (“TPOP”) and voting age population (“TVAP”) of each district is as follows:

TPOP TVAP

District 1 629.761 456,300

District 2 District 3 629,735 629,748 455,164 464,632

District 4 629,690 472,785

District 5 629,727 492,438

District 6 629,725 455,805

District 7 629,706 444,493

District 8 629,700 457,971

District 9 629.762 467,232

District 10 629,702 463,958

District 11 629,698 465,459

District 12 629,735 470,201

District 13 629,732 450,756

Id. ¶ 116.

Under the proposed Georgia Congressional plan, the percentages of black population (“BPOP”), BVAP, and black registration (“BREG”) for each of the proposed Congressional districts are as follows:

%BPOP %BVAP(Ga.)17 %BREG

District 1 23.21 21.04 18.62

District 2 45.22 41.46 39.99

District 3 40.32 37.55 34.97

District 4 54.69 50.02 51.16

District 5 56.92 52.04 53.36

District 6 7.36 6.87 6.27

District 7 7.43 6.81 6.02

District 8 12.95 12.07 10.37

District 9 14.07 12.99 11.16

District 10 3.65 3.36 2.89

District 11 ■ 29.10 26.36 26.14

District 12 43.19 39.00 39.10

District 13 41.97 38.22 41.57

PLExs. 9C, 9D.

Under the proposed Congressional redistricting plan, there are still two Congressional districts, the Fourth and Fifth, with majority black populations, but the number of Congressional districts with over 50% BVAP (Ga.) and black voter registration has increased from one, the Fifth, to two, the Fourth and Fifth. Id. However, as intervenors emphasize, according to the United States’ calculations of BVAP, there is one district with a majority BVAP — the Fifth District — in both the benchmark and proposed Congressional plans. See Br. of Amicus Curiae/Defendant-Intervenors, Jan. 14, 2002, at 26.

The proposed plan would also create additional districts with significant African American populations: (1) the proposed Second Congressional District has a BVAP in excess of 40% and a black voter registration of nearly 40%; (2) the proposed Thirteenth Congressional District has black voter registration of over 40%; and (3) the Third and Twelfth Congressional Districts have black populations of over 40% and significant BVAP and black voter registrations. PLExs. 9C, 9D. Several African American candidates have announced their intentions to run for the new 12th and 13th Congressional districts. Tate dep. at 107:11-18.

The State presented an analysis of the statewide election returns in four elections between an African American candidate and a white candidate held between 1998 and 2000. This analysis predicted that the voters of the proposed Fifth Congressional District would have supported African American candidates by an estimated 68.97% in a Democratic primary, and by 70.83% - 75.65% in the general elections. PLExs. 9D, 10B.

The State did not introduce expert testimony interpreting the significance of these percentages. Furthermore, the court heard no expert testimony regarding the existence of racially polarized voting patterns in any of the benchmark or proposed Congressional Districts, or on the impact of such patterns on the ability of minority candidates to win election. Intervenor-defendant Patrick L. Jones testified that he “believes” that it is “difficult, if not impossible” for minority candidates of choice to be elected in districts of less than 55% BVAP, and that it will be difficult to *45elect a candidate of choice in the Fifth Congressional District. Int.Ex. 27.

Intervenors have submitted alternative plans, some of which would increase BVAPs in majority-minority districts. See Int.Exs. 20-22. However, the three alternative plans submitted by intervenors create at most two Congressional districts with BVAP majorities. None of the alternative plans place Republican incumbents in the same district.

E. State House Redistricting

The Georgia Constitution mandates that the Georgia House of Representatives consist of not fewer than 180 members apportioned among districts of the State of Georgia. Ga. Const., Art. Ill, § II, ¶ 1(b); O.C.G.A. § 28-2-1; PPFF ¶ 135. Members of the Georgia House of Representatives are elected for two-year terms and serve until the time of the convening of the next General Assembly. Ga. Const., Art. Ill, § II, ¶ V(a); PPFF ¶ 136. Members of the Georgia House of Representatives are elected at the same time as the Governor. Ga. Const., Art. V, § I, ¶ II; PPFF ¶ 137.

1. The Benchmark Plan

The benchmark plan for the Georgia House of Representatives contains 180 single-member districts. PPFF ¶ 140; PI. Exs. HA, 11C. Based upon a total population of 8,186,453 people and the existence of 180 members of the Georgia House of Representatives, the ideal size of a State House district for one person, one vote purposes is 45,480 people. PLEx. 12C at 8; PPFF ¶ 142.

According to the 2000 census population statistics, there are 40 districts in the benchmark plan in which the total non-Hispanic black population is over 50%.18 Id. ¶ 143. In addition, there are 37 districts in the benchmark plan in which the total BVAP is over 50%.19 Id. ¶ 144. This is true whether BVAP is calculated according to the Attorney General’s Guidance or by Georgia’s methodology. Finally, under the benchmark plan, there are 38 districts in which the total black voter registration is over 50%.20 Pl.Ex. 11E.

Georgia’s House districts have traditionally been drawn with a deviation of plus or minus five percent from the ideal district size. PPFF ¶ 146. According to the 2000 census results, only two of the existing House districts with a total black population, total BVAP, or black voter registration over 50%, fall within that traditional deviation requirement. Id.; PLExs. 11D, 11E. All but five of the 37 majority BVAP districts were between -7.23% and - 31.92% in deviation from the ideal district size, indicating they were significantly underpopulated. Id.

2. The Proposed Plan

In its first special session, the Georgia General Assembly passed a redistricting plan for the State House that was not signed by the Governor. James dep. at 89: 4-12; Int.Ex. 31 at 3-4 (Westmoreland *46Decl.). After the passage of the first House plan, Senator Vincent Fort, who is African American, called for a meeting of the Georgia Legislative Black Caucus (“GLBC”). In an August 24, 2001 letter addressed to Representative Carl Von Epps, Chairman of the Caucus, Senator Fort stated:

We are concerned that the GLBC has not been involved in the redistricting process almost at all. This has resulted, among other things, in a legislative plan passing that has diluted majority-minority districts in both the House and the Senate.

Int.Ex. 17.- This letter contained the signatures of six members of the GLBC. Id.; U.S.Ex. 722, 51:23-52:16. Senator Fort either called or spoke to each member of the GLBC whose signature appears upon the letter, and each consented to signing the letter. Id. at 55:22-56:10; USPFF at ¶ 64.

During its second special session, the Georgia General Assembly enacted House Bill 14EX2, which provided for the reapportionment of the Georgia House of Representatives. See Pl.Ex. 12A (identifying plan as “HSEPLN2”); PPFF ¶ 147. House Bill 14EX2 was adopted by the Georgia House of Representatives on August 29, 2001, by a vote of 100 to 72. Id. ¶ 148. The Senate passed the bill on September 6, 2001, by a vote of 29 to 22. Id. Representative Dorothy Pelote and Senator Regina Thomas, both of the Savannah area, were the only African American legislators who voted against the plan. Id. The Governor signed H.B. 14EX2 into law on October 1, 2001, as Act No. 2EX23. Id. ¶ 149.

This proposed House plan contains 180 members allocated to 147 districts. PLEx. 12C; PPFF ¶ 150.124 districts contain one member, 15 districts contain two members, six districts contain three members, and two districts contain four members. Id.

The proposed House plan contains 42 districts in which the total black population is over 50%. PLEx. 12C.21 The proposed House plan contains 39 seats in districts in which the total BVAP, pursuant to the State’s interpretation of the census data, is over 50%.22 PLEx. 12C. When BVAP is calculated pursuant to the Attorney General’s Guidance, the redistricting plan contains 38 House seats in which the BVAP is over 50%. U.S.Resp. to PPFF ¶ 152. The proposed House plan contains 39 districts in which black voter registration is over 50%.23 PLEx. 12D.

Comparing the proposed plan to the benchmark plan, there are two additional districts with black populations of over 50%, and one additional district with black voter registration over 50%. Pl.Exs. 12C, 12D. Either one or two additional seats are created in districts with majority BVAPs, depending on whether the United States or Georgia’s method of calculating BVAP is used. Id.; U.S.Resp. to PPFF ¶ 152.

*473. The Challenged Districts

Intervenors challenge the drawing of seven House districts, 51, 95, 97, 100, 113, 124, and 125, as well as the creation of multi-member districts. See Renewed Mot. to Intervene, Jan. 4, 2002. The alternative House reapportionment plans submitted by intervenors, and drawn by Republican House leader Lynn Westmore-land, place certain Democratic incumbents in the same districts to run against one another. See, e.g., Int.Ex. 31 at 36:19-37:18, 38:17-39:11, 44:4-25, 46:12-16; 48:11-16; 50:1-51:16, 53:2-54:18, 58:15-19, 59:5-60:2. None of Westmoreland’s proposed alternative plans drew districts in which Republican incumbents were drawn within the same district. Id. at 43:3-8; 48:17-21; 51:17-23; 54:19-22; 62:17-22.

Intervenor Roielle Tyra objects to the loss of one majority-minority House district and resulting single district where two minority members are “pitted against each other.” Int.Ex. 26. However, when reviewed as a whole, the proposed House plan creates four new opportunities for African Americans to elect candidates of their choice in open seats in House Districts 48, 59, and 61, and a new opportunity in House District 60. PLExs. 12B, 12D.

a. Proposed House District 51

Proposed House District 51 is a single-member district located wholly within Fulton County, Georgia. Pl.Exs. 12B, 13A. The proposed district embraces territory formerly included in benchmark House District 56, which is also wholly within Fulton County. PLEx. 11B.

In light of the 2000 census results, the ideal size for one of the 180 House seats is 45,480 persons. Thus, the benchmark House District 56 is 4,169 persons, or 9.17%, short of the ideal district size. Pl. Exs. 11D, 13B. Black voter registration levels, as compared to overall voter registration, have declined in benchmark House District 56 over the past three election cycles, from 58.91% in 1996, to 58.74% in 1998, and finally to 55.86% in 2000. Pl. Exs. HE, 13B.

Voters within benchmark House District 56 tend to vote for Democratic Party candidates, as demonstrated by an overall Democratic performance score of 86.92%, as well as the Democratic performance numbers for the individual election years of 1996 (80.25%), 1998 (82.14%) and 2000 (81.50%).24 PLExs. HE, 13B.

In addition, 60.9% of the voters within the benchmark House District 56 voted for Michael Thurmond, an African American, over his white opponent in the 1998 Democratic Party primary runoff election for the open State Labor Commissioner seat. 88.43% of voters in benchmark House District 56 voted for Thurmond in his general election victory over a white Republican opponent. PLExs. HE, 13B. Additionally, voters within the benchmark House District 56 demonstrated electoral support for other African American Democratic candidates running for statewide office, voting in 1998 for Thurbert Baker for Attorney General at a rate of 82.21% and, in 2000, for David Burgess, a candidate for the Public Service Commission, at a rate of 79.23%. PLExs. HE, 13B.

The precincts included in the proposed House District 51 have supported Democratic candidates. According to past election results, the precincts that comprise proposed House District 51 have an overall Democratic performance score of 86.38%. PLEx. 12D. Using 1996 election results, plaintiff has projected Democratic perfor-*48manee numbers of 80.38%, using 1998 election results, 81.82%, and using 2000 election results, 80.89%. Pl.Exs. 12D, 13B. Voters within the proposed district also supported Michael Thurmond in the 1998 primary runoff at a rate of 61.18%, and at 87.97% in his general election contest. Id. Additionally, 81.55% of these voters supported Thurbert Baker, and 78.67% voted for David Burgess. Id.

Proposed House District 51 retains benchmark District 56’s status as a district with a majority of total black population and BVAP, as shown below.

_TPOP25_BPOP_TVAP26 BVAP (Ga.)

Benchmark H.D. 56 41,311 24.801 32,393 17,724 _(60.03%)_(54.72%)

Proposed H.D. 51 43,675 25,162 34,793 18,118 (57.61%)(52.07%)

PLExs. 12C, 13B. Moreover, proposed District 51 retains a majority black voter registration level of 52.68%. PLExs. 12D, 13B. Proposed House District 51 retains this status while making up benchmark District 56’s significant population shortage from the current ideal district size of 45,480. Proposed District 51 is only 1,805 persons, or 3.97%, short of the ideal district size. PLExs. 12C, 13B.

b. Proposed House District 95

Proposed House District 95 is a single-member district that includes all of Hancock, Glascock, Taliaferro and Warren Counties, and parts of Baldwin, McDuffie and Putnam Counties, Georgia. PLExs. 12A, 14A. The proposed district includes territory currently within benchmark House District 120, which encompasses all of Taliaferro, Warren, Glascock and Hancock Counties and parts of McDuffie and Baldwin Counties. PLEx. 11A.

The benchmark House District 120 is 7,056 people, or 15.51%, short of the ideal district size. PLExs. 11D, 14B. Black voter registration numbers have remained relatively steady over the past three election cycles in the benchmark House District 120, with 52.14% black voter registration in 1996, 52.90% in 1998, and 52.07% in 2000. PLExs. HE, 14B.

Voters within benchmark House District 120 tend to vote for Democratic Party candidates as demonstrated by an overall Democratic Performance score of 62.80%, as well as the Democratic performance numbers for the individual election years of 1996 (61.37%), 1998 (66.77%) and 2000 (57.06%). PLExs. HE, 14B. Specifically, 77.32% of the voters within House District 120 voted for Thurmond in the 1998 Democratic Party primary runoff election. In addition, 72.74% of the voters in House District 120 voted for Thurmond in his general election victory. PLExs. 11E, 14B. Voters within House District 120 have also supported other African American Democratic candidates running for statewide office. In 1998 Thurbert Baker received 64.49% of the district’s votes in his race for Attorney General and, in 2000, David Burgess garnered 72.17% of the vote for Public Service Commissioner. Pl. Exs. HE, 14B.

*49Benchmark House District 120 is currently held by Representative Sistie Hudson, a white Democrat. Int.Ex. 31 at 4. In 1996, Representative Hudson faced a primary challenge from an African American opponent, Frederick Favors. Id. at 4-5. Representative Hudson received 51.3% of the vote, while Ms. Favors received 48.7%. Id.

Plaintiff has calculated an overall Democratic performance score of 56.61% for proposed House District 95. In addition, the State projects Democratic performance numbers of 53.04% using 1996 election results, 62.65% using 1998 results, and 51.97% using 2000 results. PLEx. 14B. In looking at the past political performance of the territory contained within proposed House District 95, voters within the proposed district also supported Michael Thurmond at a rate of 75.01% in the 1998 primary runoff, and 67.36% in his general election contest. Additionally, these voters supported Thurbert Baker at a rate of 59.78% and David Burgess at 66.57%. PL Exs. 12D, 14B.

The demographics of benchmark District 120 and Proposed House District 95 are shown below:

[[Image here]]

PLExs. 11D, 12C, 14B. Additionally, 44.06% of registered voters in the proposed District 95 are African American. PLExs. 12D, 14B.

This district also makes up benchmark District 120’s population shortage from the current ideal district size of 45,480. Proposed District 95 is only 890 persons, or 1.96%, short of the ideal district size. Pl. Exs. 12C, 14B.

c. Proposed House District 97

Proposed House District 97 is a single-member district located wholly within Richmond County, Georgia. PLExs. 12A, 15A. House District 97 embraces territory included within benchmark House District 117, which is also wholly contained within Richmond County. PLEx. 11A.

The benchmark House District 117 is 10,918 people, or 24.01%, short of the ideal district size. PLExs. 11D, 15B. Black voter registration levels have increased in the existing House District 117 over the past three election cycles, from 71.80% in 1996, to 75.32% in 1998, and finally to 75.59% in 2000. PLExs. HE, 15B.

Voters within the benchmark House District 117 tend to vote heavily for candidates of the Democratic Party, demonstrated by an overall Democratic Performance score of 79.13%, as well as the Democratic performance numbers for the individual election years of 1996 (76.11%), 1998 (77.73%) and 2000 (79.93%). PLExs. HE, 15B. Specifically, 73.73% of the voters within existing House District 117 voted for Michael Thurmond in the 1998 Democratic Party primary runoff election, and 85.27% voted for him in the general election. PLExs. HE, 15B. Additionally, voters within benchmark House District 117 demonstrated electoral support for other African American Democratic candidates running for statewide office by voting at a rate of 83.09% for Thurbert Baker in 1998, *50and at 85.47% for David Burgess. PLExs. HE, 15B.

Proposed District 97 retains much of the overall Democratic performance of benchmark District 117. Plaintiff has calculated an overall Democratic performance score of 68.59% for this proposed District, and projected Democratic performance numbers of 60.47% using 1996 election results, 64.37% using 1998 results and 63.08% using 2000 election results. Pl.Ex. 15B. In looking at the past political performance of the territory contained within proposed House District 97, 72.92% of the voters in the proposed district supported Michael Thurmond in the 1998 primary runoff and 70.03% supported him in the general election. Additionally, 67.74% of these voters supported Thurbert Baker and 71.11% supported David Burgess. PLExs. 12D, 15B.

The proposed House District 97 also retains District 117’s status as a district in which African American voters comprise a majority of both the total and voting age populations. The following table compares the demographics of the benchmark District 117 and proposed House District 97, using the 2000 census figures:

[[Image here]]

PLExs. 11D, 12C, 15B. Additionally, proposed District 97 retains District 117’s majority black voter registration at a level of 54.53%. PLExs. No 12D, 15B.

Proposed District 97 is only 1,949 persons, or 4.29%, short of the ideal district size. PLExs. 12C, 15B.

d. Proposed House District 100

Proposed House District 100 is a single-member district that contains all of Burke County and part of Richmond County, Georgia. PLExs. 12A, 16A. The proposed district 100 embraces territory included within benchmark House District 116, which also contains all of Burke County and part of Richmond County, Georgia. PLEx. 11A.

Based on the population statistics from the 2000 census, the benchmark House District 116 is 6,161 persons, or 13.55%, short of the ideal district size. PLExs. 11D, 16B. Levels of black voter registration have increased in the existing House District 116 over the past three election cycles, from 52.75% of total voter registration in 1996, to 54.59% in 1998, and to 54.29% in 2000. PLExs. HE, 16B.

Voters within benchmark House District 116 tend to vote heavily for Democratic Party candidates, as demonstrated by an overall Democratic performance score of 62.13%, and the Democratic performance rates for the individual election years of 1996 (60.72%), 1998 (62.88%) and 2000 (59.83%). PLExs. HE, 16B. Specifically, 74.88% of the voters in benchmark House District 116 voted for Michael Thurmond in his 1998 Democratic Party primary runoff election, and 69.96% supported him in the general election. PLExs. HE, 16B. Furthermore, voters in benchmark House District 116 have supported other African American candidates running for statewide office, voting at a rate of 69.05% for Thur-bert Baker in 1998, and 74.72% for David *51Burgess in 2000. PLExs. HE, 16B. Proposed House District 100 retains much of the overall Democratic performance of benchmark district 116, with a projected overall Democratic performance score of 62.56%, and projected Democratic performance rates of 61.13% using 1996 election results, 63.53% using 1998 results, and 60.14% for the 2000 election results. PI. Ex. 16B. In looking at the past political performance of the territory contained within proposed House District 100, 74.98% of voters supported Michael Thurmond in the 1998 primary runoff, and 70.37% supported him in his general election contest. PLEx. 12D. Additionally, 69.48% of these voters supported Thurbert Baker and 74.68% supported David Burgess. PLExs. 12D, 16B.

Like benchmark District 116, proposed House District 100 has a majority African American population and voting age population. The following table compares the demographics of the benchmark District 116 and proposed House District 100, using the 2000 census figures:

[[Image here]]

PLExs. 11D, 12C, 16B. Additionally, proposed House District 100 has a majority black voter registration level of 54.67%. PLExs. 12D, 16B.

Proposed House District 100 is only 1,287 persons, or 2.83%, short of the ideal district size. PLExs. 12C, 16B.

e. Proposed House District 113

Proposed House District 113 is a single-member district located wholly within Muscogee County, Georgia. PLExs. 12A, 17A. The proposed district embraces territory included within benchmark House District 134, which is also wholly contained within Muscogee County. PLEx. 11A.

Based on the population statistics from the 2000 census, the benchmark House District 134 is 14,518 persons, or 31.92%, short of the ideal district size. PLExs. 11D, 17B. Black voter registration levels have increased in the existing House District 134 over the past three election cycles, from 72.87% of total voter registration in 1996, to 74.26% in 1998, and finally to 76.34% in 2000. PLExs. HE, 17B.

Voters within the benchmark House District 134 tended to vote heavily for Democratic Party candidates, as demonstrated by an overall Democratic Performance score of 84.72%, and the Democratic performance rates for the individual election years of 1996 (68.06%), 1998 (87.93%) and 2000 (84.58%). PLExs. HE, 17B. Specifically, 71.70% of the voters in the benchmark House District 134 voted for Michael Thurmond in the 1998 Democratic Party primary runoff election, and 90.11% of voters cast their ballots for him in the general election. PLExs. HE, 17B. Additionally, voters within the benchmark House District 134 demonstrated significant electoral support for other African American Democratic candidates running for statewide office. In 1998, 91.03% of voters supported Thurbert Baker and, in 2000, 88.90% voted for David Burgess. PLExs. HE, 17B.

Proposed House District 113 retains much of the overall Democratic perfor-*52manee of benchmark District 134, with an overall Democratic performance score of 75.04%, and specific projected Democratic performance numbers of 64.81% using 1996 election results, 75.24% using 1998 results, and 73.16% using 2000 election results. Pl.Ex. 17B.

Based on the past political performance of the territory contained in proposed House District 113, 69.38% of voters within the proposed district supported Michael Thurmond in the 1998 primary runoff and 77.38% supported him in his general election contest. Additionally, 79.45% of these voters supported Thurbert Baker, and 78.56% supported David Burgess. Pl.Exs. 12D, 17B.

Like benchmark District 134, proposed District 113 has a majority of African American population and voting age population. The following table compares the demographics of the benchmark District 134 and proposed House District 113, using the 2000 census figures:

[[Image here]]

Pl.Exs. 11D, 12C, 17B. Additionally, proposed House District 113 has a majority black voter registration at a level of 61.88%. PLExs. 12D, 17B.

Proposed House District 113 is only 1,674 persons, or 3.68%, short of the ideal district size. PLExs. 12C, 17B.

Under the proposed plan, two incumbent minority legislators, Representatives Carolyn Hugley and Maretta Taylor, are drawn into the proposed District 113. See Int. Ex. 31 at 8 (Westmoreland Decl.). Inter-venors introduced testimony that these are the only two incumbent Democrats included in any one proposed district in any of the three plans before this court. Id. An open seat was created in proposed House District 109, a majority-white district adjacent to proposed district 113. Id.

f. Proposed House District 124

Proposed House District 124 is a two-member district located wholly within Chatham County, Georgia. PLEx. 12A. Proposed District 124 embraces territory included in benchmark House Districts 148 and 149, which were also wholly contained within Chatham County. PLEx. 11A.

According to population statistics from the 2000 census, benchmark House District 148 was short 10,343 persons, or 22.74%, and benchmark District 149 was 12,815 persons, or 28.18%, short of the ideal district size. PLExs. 11D, 18B. Black voter registration levels have increased in the benchmark House District 148 over the past three election cycles, from 64.19% in 1996, to 65.37% in 1998, and finally to 68.12% in 2000. PLExs. HE, 18B. Black voter registration levels have decreased in benchmark House District 149 over the past three election cycles, from 70.80% of total voter registration in 1996, to 70.46% in 1998, and. finally to 67.96% in 2000. PLExs. HE, 18B.

Voters in benchmark House District 148 tend to vote for Democratic Party candidates, as demonstrated by an overall Democratic performance score of 76.58%, and Democratic performance numbers for indi*53vidual election years of 1996 (72.60%), 1998 (79.26%) and 2000 (74.75%). Pl.Exs. HE, 18B. Specifically, 82.23% of voters in benchmark House District 148 voted for Michael Thurmond in the 1998 Democratic Party primary runoff election, and 83.02% supported him in the general election. Pl. Exs. HE, 18B. Additionally, voters within benchmark House District 148 demonstrated electoral support for other African American Democratic candidates running for statewide office. In 1998, 82.90% of voters supported Thurbert Baker for Attorney General and, in 2000, 82.83% supported David Burgess. PLExs. HE, 18B.

Voters from benchmark House District 149 also tend to support Democratic Party candidates, as demonstrated by an overall Democratic performance score of 81.20%, and Democratic performance numbers for individual election years of 1996 (77.61%), 1998 (81.89%) and 2000 (78.35%). PLExs. HE, 18B. 89.39% of the voters in benchmark House District 149 voted for Michael Thurmond in the primary runoff, and 88.84% supported him in the general election. PLEx. 18B. In 1998, 89.34% of voters in benchmark District 149 voted for Thur-bert Baker, and, in 2000, 85.90% voted for David Burgess. PLExs. HE, 18B.

Proposed House District 124 has an overall Democratic performance score of 67.67%, and projected Democratic performance numbers of 63.16% using 1996 election results, 69.79%, using 1998 results, and 66.46% for the 2000 election results. PLEx. 18B. In looking at the past political performance of the territory contained within proposed House District 124, 81.50% of voters within the proposed district supported Michael Thurmond in the 1998 primary runoff and 75.83% supported him in his general election contest. 76.11% of these voters supported Thurbert Baker and 75.39% supported David Burgess. Pl. Exs. 12D, 18B.

As in benchmark Districts 148 and 149, in proposed District 124 African American voters make up a majority of both the total and voting age populations. The following table compares the demographics of the benchmark House Districts 148 and 149 with proposed House District 124, using the 2000 census figures:

[[Image here]]

PLExs. 11D, 12C, 18B. Additionally, proposed District 124 has a majority black voter registration level of 55.21%. PLExs. 12D, 18B.

The ideal district size for proposed District 124 is 90,960, twice the ideal district size of a single-member district. Thus, proposed District 124 is 2,091 persons, or

4.60%, short of the ideal district size. Pl. Exs. 12C, 18B.

g. Proposed House District 125

Proposed House District 125 is a single-member district located wholly within Chatham, Georgia. PLExs. 12A, 19A. House District 125 embraces territory in-*54eluded within benchmark House District 151, which is also wholly within Chatham County. Pl.Ex. 11A.

Based on population statistics from the 2000 census, the benchmark House District 151 is 11,450 persons, or 25.18%, short of the ideal district size. PLExs. 11D, 19B. Black voter registration levels have increased in benchmark House District 151 over the past three election cycles, from 46.18% in 1996 to 47.93% in 1998, and to 48.40% in 2000. Pl.Exs. 11E, 19B.

Voters in benchmark House District 151 tend to vote for Democratic Party candidates, as demonstrated by an overall Democratic performance score of 64.31%, and the Democratic performance numbers for the individual election years of 1996 (61.47%), 1998 (67.19%) and 2000 (61.39%). PLExs. HE, 19B. Specifically, 75.40% of voters in benchmark House District 151 voted for Michael Thurmond in the 1998 Democratic Party primary runoff election, and 71.67% of voters supported him the general election. PLExs. HE, 19B. Voters in benchmark House District 151 have demonstrated electoral support for other African American Democratic candidates running for statewide office: in 1998, 72.36% voted for Thurbert Baker and, in 2000, 72.78% voted for David Burgess. Pl. Exs. HE, 19B.

Proposed House District 125 has an overall Democratic performance score of 59.40%, and specific projected Democratic performance levels of 56.64% using 1996 election results, 62.75% using 1998 results, and 56.56% using 2000 election results. Pl.Ex. 19B. 70.77% of voters in the proposed district supported Michael Thurmond in the 1998 primary runoff, and 67.13% supported him in his general election contest. Additionally, 67.84% of these voters supported Thurbert Baker, and 68.40% supported David Burgess. PLExs. 12D, 19B.

While Proposed House District 125 has a majority black population, it does not have a majority BVAP. The following table compares the demographics of the benchmark House District 151 and proposed House District 125, using the 2000 census figures:

[[Image here]]

PLExs. 11D, 12C, 19B. Proposed House District 125 has a black voter registration level of 41.67%. PLExs. 12D, 19B. Proposed District 125 is only 836 persons, or 1.84%, short of the ideal district size. Pl. Exs. 12C, 19B.

h. Multi-Member Districts

Some of the districts in Georgia’s proposed House plan are multi-member districts. See PLExs. 12C, 12D. Plaintiffs expert report lists the proposed districts, and the number of seats and BVAP of each district. Pl.Ex. 25. There are six proposed multi-member districts with BVAPs of 50% or higher: District 43 has a BVAP of 65.18%; District 48 has a BVAP of 61.13%; District 49 has a BVAP of 61.86%; District 60 has a BVAP of 59.51%; District 61 has a BVAP of 58.33%; and District 124 *55has a BVAP of 54.14%.27 Pl.Ex. 25, App. 3; Tr., 2/5/02, a.m. at 79-81. According to plaintiffs expert, the probability that minority voters will elect candidates of their choice varies from 80% to near certainty in these multi-member districts. Id.

F. State Senate Reapportionment Plan

The Georgia Constitution mandates that the Georgia Senate consist of 56 Senators elected from single-member districts apportioned among the respective districts of the State of Georgia. Ga. Const., Art. Ill, § II, ¶ 1(a); O.C.G.A. § 28-2-2. Members of the Georgia Senate are elected for two-year terms and serve until the time of the convening of the next General Assembly. Ga. Const., Art. Ill, § II, KV(a). Members of the Georgia Senate are elected at the same time as the Governor. Ga. Const., Art. V, § I, ¶ II.

The 2000 census results show a statewide population of 8,186,453 people. Consequently, the ideal size for the 56 Senate Districts for purposes of one person-one' vote is 146,187 people. PPFF ¶ 407. The State of Georgia’s Senate districts have traditionally been drawn with a deviation of plus or minus five percent from the ideal district size. Id. ¶ 411. According to the 2000 census data, only two of the benchmark Senate districts with a total black population, total BVAP, or black voter registration over 50% were within that traditional deviation requirement. Pl.Exs. ID, IE. All but three of the ten majority BVAP districts deviated from the ideal district size by -14.02% to -26.94%. Id.

On August 10, 2001, the Senate Redistricting Committee leadership called for a final vote on a Senate redistricting plan, SENPLAN2U, shortly after the plan had been distributed to the Senators for review. Ms. Meggers testified that the Redistricting Office did not prepare the demographic and political performance reports for plan SENPLAN2U until the morning of the Friday it was enacted. . U.S.Ex. 702, 76:6-77:22.

The second proposed Senate reapportionment plan became Senate Bill 1EX1 and was adopted by the Georgia State Senate on August 10, 2001, by a vote of 29 to 26. PPFF ¶¶ 412, 413. Senator Regina Thomas of the Savannah area was the only African American Senator who voted against the plan. Id. ¶ 413. The House of Representatives passed Senate Bill 1EX1 on August 17, 2001, by a vote of 101 to 71. Id. Representative Dorothy Pelote, also of the Savannah area, was the only African American representative to vote against the plan. Id. The Governor signed Senate Bill 1EX1 into law on August 24, 2001, as Act No. 1EX6. Id. ¶ 414.

1. Demographics of Benchmark and Proposed Districts

The State of Georgia’s current Senate districting map is the result of a mediated agreement between the State of Georgia and the United States Department of Justice in 1997. Based upon the 2000 census statistics, there are thirteen Senate Districts in the benchmark Senate plan with a total black population of over 50%. PPFF at ¶ 408. These are Senate Districts 2, 10, 12, 15, 22, 26, 35, 36, 38, 39, 43, 44 and 55. Id. The same thirteen Senate Districts have total black voter registration levels of 50% and higher. PPFF at ¶ 410. Twelve of these districts have BVAP of 50% or higher. Id. ¶ 409. These are Senate Districts 2, 10, 12, 15, 22, 26, 35, 36, 38, 39, 43 and 55. Id. Senate District 44, which currently has 52.80% total black population *56and 53.72% black voter registration, has a 49.62% BVAP.28

Like the benchmark plan, the proposed redistricting plan contains 13 Senate Districts with a total black population of over 50%. However, the proposed plan contains only eight Senate Districts in which black voter registration is over 50%. The State contends that the proposed plan also creates 13 Senate Districts in which the BVAP is over 50%. The United States disputes this calculation. According to the United States’ interpretation of the census data, there are only 11 districts with majority-minority BVAPs. When BVAP is calculated in accordance with the Attorney General’s Guidance, proposed Senate Districts 2 and 34 fall below 50% BVAP.

Furthermore, the black population of Senate District 44 would be severely reduced under the proposed plan. Benchmark District 44 has a BVAP of almost 50%. The proposed District 44 has a BVAP of approximately 34%.

The following table compares the demographics of the benchmark Senate Districts with the proposed Senate Districts, using the 2000 census statistics:

[[Image here]]

U.S.Ex. 110; PLEx. 25.

2. Contested Districts

a. Senate District 2

Both the benchmark and proposed Senate District 2 are located within Chatham County, Georgia. Given the ideal district size of 146,187 persons for a Senate district after the 2000 census, the population of benchmark Senate District 2 was 35,629 people, or 24.37%, short of the ideal size. PPFF ¶ 428; see also PLEx. 1-D. The proposed District 2 is 7,217 people, or 4.94%, short of the ideal district size. Id. ¶ 429.

According to 2000 census statistics, the BVAP of the benchmark Senate District 2 *57is approximately 60%. Pl.Ex. 3B. The BVAP of the proposed Senate District 2 is either 50.31%, according to plaintiff, or 49.81%, according to the United States. Pl.Ex. ID; U.S.Ex. 110. The State argues that, because of population changes alone, it was inevitable that the BVAP of District 2 would decrease.

Black registered voters compose 48.42% of the proposed Senate District 2. In contrast to Senate District 2, in at least six of the majority-minority districts, black registration levels are higher than BVAP. The United States suggests that calculations of BVAP for Senate District 2 may include an undetermined number of ineligible voters. Senate District 2, according to Ms. Meg-gers, is home to Savannah State University and the Savannah School of Art and Design. USPFF ¶ 177. Ms. Meggers was unaware of the number of students at the schools, whether the Schools have on-campus housing, or whether all of the students were registered voters. Id. She further testified that a student population of 1,500, which she would not consider “large,” would constitute approximately 1.5% of Senate District 2’s voting age population. Id.

Senator Regina Thomas, an African American, is the current incumbent in Senate District 2. Senate District 2 has been represented by several African American Senators. Without opposition, Roy L. Allen, who is African American, won the Democratic Party nomination and the General Election in 1992. PPFF ¶ 431. Without opposition, Diane H. Johnson, who is African American, won the Democratic Party nomination and the General Election in 1994 and 1996. PPFF ¶¶ 432, 433. In 1998, Senator Johnson was opposed by Regina Thomas in the Democratic primary, but Johnson won by a 297-vote margin of victory. Id. ¶ 434. In 1998, Senator Johnson was re-elected to her third term representing Senate District 2 without opposition in the General Election. Id. ¶ 435.

In 1999, Senator Johnson was indicted and convicted of federal mail fraud. Id. ¶ 436. She resigned from the State Senate on November 12, 1999, and the Governor called a special election to fill the vacant Senate seat. Id. This special election was precleared by the Department of Justice. Id. ¶ 437; (DOJ File No.1999-3631 (Dec. 29, 1999)). The following candidates qualified to run in the special nonpartisan election held on December 21, 1999, to succeed Senator Johnson: Pro-Life Anderson (white man), Dana F. Braun (white man), Willie Brown (African American man), Bob Bryant (African American man), Charles Gordon (African American man), and Regina Thomas (African American woman). Id. Regina Thomas won that election. Id.

Senator Thomas testified that she was a member of the Senate Reapportionment Committee. However, she testified that her role in the redistricting process was “[jjust to be on the committee” and she did not do anything more than attend every meeting. U.S.Ex. 704, 38:25-39:3, 40:12-24. She voted against the redistricting plan and has expressed her displeasure with the re-drawing of her district. Id. She believes that she will have difficulty being re-elected and that African American voters will have fewer opportunities to elect candidates of choice in the re-drawn Senate District 2. Plaintiff attempts to discredit Senator Thomas as being “surprisingly unfamiliar with the demographics of her district” because she indicated at her deposition testimony that her district had 49% BVAP, and not 50%. PPFF at ¶ 421 (citing Senator Thomas Dep. at 146^17). However, as discussed above, using the method of calculating BVAP mandated by the United States’ Guidance, proposed Senate District 2 has a BVAP of less than 50%.

*58In a 1999 runoff election, Senator Thomas defeated her white opponent, Dana Braun, by slightly more than 70 votes,29 when benchmark Senate District 2 had a black registration level of more than 60%. PPFF ¶ 440. When Thomas ran in November 2000, she won reeleetion in the General Election by a vote of 22,723 (77.8%) to 6,494 (22.2%). Id. ¶ 441. Her opponent in that general election was a white Republican, who apparently did not have the support of the Republican Party. See U.S.Ex. 727, 86:19-88:13 (Senate minority leader for Republican Party describing Thomas’ opponent as a “nut case” and “flaky, half-crazed Republican that runs for office now and then”); see also U.S.Ex. 704, 32:19-23, 187:23-188:2, 189:19-25 (Thomas testified that opponent was a pizza delivery man, not well known in the community, and had been arrested for impersonating a police officer and carrying concealed weapons). Although Senator Thomas won that 2000 election by a substantial majority, she did not receive a majority of the white vote. U.S.Ex. 601.

The United States presented testimony from eleven witnesses from Chatham County about proposed Senate District 2, including two State Senators, four African American City Council Aldermen, two Chatham County Commissioners, and three members of the Executive Committee of the Savannah Branch of the NAACP. Almost all of the witnesses testified to the existence of racially polarized voting in Senate District 2. For example, Dr. Prince Jackson, an Executive Committee member of the local NAACP, testified that it is his “belief that African-American voters typically vote for African American candidates, and white voters typically vote for white candidates.” U.S.Ex. 503 ¶ 8 (Jackson deck).

In addition, Richard Shinhoster, an Executive Committee member of the local NAACP, testified that a slim majority of BVAP may not be sufficient to elect a candidate of choice if voter turnout rates are low; the African American community “cannot always be assured that a black can be elected when the majority — when the ratio is so close.... ” Shinhoster dep. at 16. Shinhoster also claims that an increase in BVAP in proposed Senate District 2 is needed because “voting pattern[s] in Savannah and Chatham County are polarized along racial lines.” U.S.Ex. 509 ¶ 8 (Shinhoster deck). On the other hand, on cross-examination, Savannah City Aider-man Helen Johnson testified that “people don’t really vote because of racial issues or because their particular race right in the city elections.” Johnson dep. at 41.

Several of the lay witnesses from Senate District 2 presented by the United States testified that the addition of the Tybee Island, Isle of Hope, and Thunderbolt areas in the proposed Senate District 2 “will make it more difficult for African American voters to elect their chosen candidates.” U.S.Ex. 509 ¶ 11 (Shinhoster deck); U.S.Ex. 506 ¶ 11 (D. Jones deck).

Despite the existence of racial polarization, these witnesses also testified that African American candidates of choice can be elected to office when they are professional and well-respected by the white community. For example, some of the United States’ witnesses stated that they considered Regina Thomas to be a strong African American candidate and very popular in her Senate District. Jackson dep. at 62-63; Shinhoster dep. at 26-28, 68. Alderman Gwendolyn Goodman testified that, despite a tendency of people to vote according to race, when an African American candidate is known as a professional and respected by the white community, *59many white voters will “go with what is right.” Goodman dep. at 29-80.

The United States’ expert, Dr. Richard Engstrom, analyzed local election data from the benchmark Senate District 2, including data from elections in municipalities and counties within benchmark Senate District 2. The specific results of his assessment of racially polarized voting are found in Part 2G of this Memorandum Opinion.

b. Senate District 12

Both the benchmark and proposed Senate District 12 are located in southwest Georgia. The benchmark Senate District 12 is 25,982, or 17.8%, below the ideal district size. According to 2000 census data, the benchmark Senate District 12 has a BVAP of 55.4% (Ga.) or 54.94% (U.S.). Under the Senate plan before the court, Senate District 12 will have a BVAP of 50.66% (Ga.) or 50.22% (U.S.). Plaintiff argues that some drop in BVAP was inevitable. PLEx. 4B; PPFF ¶ 474. However, several majority-black precincts were removed from benchmark Senate District 12 in drawing the proposed district. U.S.Ex. 118.

In addition to reducing BVAP, the proposed Senate District 12 reduces the black registration as a percentage of overall registration from approximately 52.5% to 47.5%. In 1996, 48.16% of the registered voters in the benchmark Senate District 12 were African American. PPFF ¶ 483. That number grew to 50.69% in 1998 and to 52.48% in 2000. Id. Using the proposed Senate District 12 lines, the percentage of black registered voters would have been 44.86% in 1996, 46.57% in 1998 and 47.56% in 2000. Id.; Pl.Ex. 4B.

The percentage of registered voters who are black, 47.46%, is less that the BVAP of proposed Senate District 12, either 50.66% (Ga.), or 50.22% (U.S.). PPFF ¶487. In contrast to the proposed Senate District 12, black registration levels are higher than BVAP in at least six of the other proposed majority-minority districts. The United States suggests that the BVAP in proposed Senate District 12 may include an undetermined number of ineligible voters. Senate District 12 includes two or three small colleges in Albany, and a state prison in Calhoun County. U.S.Ex. 733, 124:22-23. Meggers did not know the enrollment of the colleges, but testified that the prison inmate population was approxi.mately 1,100. Id. Furthermore, John White testified to his belief that a significant number of African American felons and ex-felons live in the district and cannot vote under state law, thus reducing the number of eligible voting age African Americans in the existing and proposed district. U.S.Ex. 513 ¶ 27.

The past voting performance of precincts contained within proposed Senate District 12 in certain past statewide elections shows that 58.57% of voters in proposed Senate District 12 supported Michael Thurmond in his primary run-off, while 65.89% of voters supported him in the general élection. PLEx. 4B. In 1998, 64.53% of the voters in proposed Senate District 12 voted for Thurbert Baker for Attorney General and, in 2000, 65.83% voted for David Burgess for Public Service Commissioner. PLEx. 4B.

In the 1992 Democratic Primary, Mark Taylor, a white man and current Georgia Lieutenant Governor, defeated Charles Sherrod, an African American man, to obtain the party nomination for Senate District 12. PPFF ¶ 490. Mark Taylor won election to Senate District 12 in the 1992 General Election without opposition. Id. In 1994, Taylor was unopposed in the Democratic Primary and the General Election and won re-election. Id. Mark Taylor testified that, when first elected to the *60Senate, he had the political endorsement of the African American community. Taylor dep. at 20. In 1996, John White, an African American candidate, opposed Taylor in the Democratic Party primary. PPFF ¶ 491. Taylor defeated White by a margin of 15,043 (61.5%) to 9,406 (38.5%) votes. Id. Taylor went on to win the general election without opposition. Id.

Taylor ran for Lieutenant Governor in 1998, and the Senate District 12 seat became open. Id. ¶ 492. White ran and lost narrowly to a local white attorney, Michael Meyer von Bremen, in the Democratic Party primary. Id. Meyer von Bremen won the primary by 51.4% of the votes. Id. Engstrom’s expert report demonstrates that in that Democratic Party primary, virtually all the black voters voted for White, and almost none of the white voters voted for White. U.S.Ex. 601, Table 2.

Meyer von Bremen subsequently won the general election in 1998, won re-election in 2000, and is the current incumbent Senator in District 12. Id. A Clark University study found Meyer von Bremen to be supportive of minority issues. Pl.Ex. 23 at 9-10. However, both experts for the plaintiff and the United States have found that Senator Meyer von Bremen is not the candidate of choice of African American voters. Pl.Ex. 25, App. 2; U.S.Ex. 601, Table 2. Furthermore, several of the United States’ witnesses testified that Senator Meyer von Bremen does not represent the interests of the African American community. U.S.Ex. 514 ¶ 10 (Wright deck); U.S.Ex. 512 ¶ 7 (D.Williams).

In addition to the statistical evidence suggesting the existence of racially polarized voting in Senate District 12, the United States’ witnesses testified about their experiences of racially polarized voting. Describing his 1992 defeat to Mark Taylor, a white candidate, Mr. Charles Sherrod stated that “Mr. Taylor won with about two-thirds of the votes. I did very badly in the white community.” U.S.Ex. 510 ¶ 6 (Sherrod deck). He explains that “whites would vote for my white opponent.... ” Sherrod admits in his deposition, however, that rural black voters also voted for Taylor and that Taylor’s financial resources greatly aided his success. Sherrod dep. at 58-59. Sherrod acknowledges that White, in 1996, ran a more competitive campaign against Taylor than he did in 1992; White had better financing, better organization, as well as popularity as “an announcer on the TV.” Id.

Sherrod states in his affidavit that “most white voters in Southwest Georgia simply will not vote for black Senate Candidates.” U.S.Ex. 510 ¶ 11. He bases his opinion on the assumption that “whites get more support from black voters than black[s]” get from white voters, but indicated that if the reverse were true, it would change his opinion. Sherrod dep. at 94-96.

Albany City Commissioner Arthur Williams testified: “I believe Mr. White lost [the 1996 Senate District 12 primary election] because he could not compete with Mr. Taylor’s money and because white voters responded to Mr. Taylor’s use of the race card.” U.S.Ex. 511. Williams contends that Taylor played the “race card” during his campaign by sending out literature that compared White to Williams and urged citizens to vote in the Democratic Party primary against White, and to not vote in the Republican Party primary. Id. ¶¶ 6, 8. Plaintiff suggests that the comparison of White to Williams in Taylor’s campaign literature was not a racial appeal, but rather is explained by the fact that White and Williams had similar positions on issues concerning municipal contracts with minority-owned businesses and affirmative action in city employment. *61PPFF ¶ 513 (citing U.S.Ex. 518 ¶ 9 (White decl.); A. Williams dep. at 77-78).

The State argues that White’s defeat at the polls was not due to racially polarized voting, but was rather the result of negative articles in the local and statewide press. One article published in the Atlanta Constitution on June 28, 1990, stated: “There is something appealing about the naked self-interest of State Representative John "White’s approach to public service. His blatant embrace of all that is bad about politics ought to be offensive to any voter, but hpw can you get angry at such candid greed?” White dep. at 28-29, Exhibit 1. White admits that such articles might have affected white voters’ decisions to vote for him or his opponent. White dep. at 6-7,18.

Wdiite stated that he may not be a strong candidate among white voters. White dep. at 29, 33. He testified that three African American candidates were more successful at getting white votes than was he, but explained that African American candidates were able to garner white votes where they had the support of white political figures. WTiite dep. at 34-35, 38, 123-24. White believes that he lost to Meyer von Bremen because “some black people did not turn out because they figured [White] was a sure bet to win.” U.S.Ex. 514 ¶ 15 (White deck).

Finally, several of the United States’ witnesses believed that it would be close to impossible for an African American candidate to be elected in the proposed Senate District 12. U.S.Ex. 514 ¶4 (Wright deck); U.S.Ex. 512 ¶ 4 (D. Williams deck). Some of these witnesses, however, stated that white voters have supported some African American candidates such as an African American Superior Court Judge, Congressman Sanford Bishop and Michael Thurmond. Wright dep. at 25-26, 83-84.

The United States’ expert analyzed local election data from the benchmark Senate District 12. The specific results of his assessment of racially polarized voting are found in Part 2G of this Memorandum Opinion.

c. Senate District 26

Both the benchmark and proposed Senate District 26 are substantially located within Bibb County, Georgia. According to the 2000 census data, the benchmark Senate District 26 is underpopulated by 42,119 people, or is 28.8% short of the ideal district size. The benchmark Senate District 26 has a BVAP of 62.45% (Ga.) or 61.93% (U.S.). Proposed Senate District 26 has a BVAP of 50.80% (Ga.) or 50.39% (U.S.).

The percentage of black registered voters in the benchmark District 26 is 62.79%. PPFF ¶ 529. That percentage decreases to 48.27% in the proposed District 26. Id. The number of black registered voters in the benchmark Senate District 26 grew from 46.51% to 48.27% from 1998 to the year 2000. Id. ¶ 531. Levels of black voter registration in the precincts comprising proposed Senate District 26 increased from 46.5% of total voter registration in 1998, to 48.3% in the year 2000. PbEx. 5B.

In contrast to proposed Senate District 26, black registration levels are higher than BVAP in at least six of the proposed majority-minority districts. The United States suggests that BVAP in Senate District 26 includes an undetermined number of ineligible voters residing in the district. USPFF ¶ 382; U.S.Ex. 733,126:2-22. Ms. Meggers testified that Mercer University is located in the proposed Senate District 26, but that she does not know the full-time student enrollment of the university. Id. Senator Brown also testified that he did not consider ineligible prison popula*62tions in drafting the proposed Senate District 26. USPFF ¶ 38B.

According to precinct returns for past elections, 71.67% of voters in the precincts in the proposed Senate District 26 voted 71.67% for Michael Thurmond in his 1998 primary runoff and 68.58% voted for Mr. Thurmond in his 1998 General Election race. Pl.Exs. 2D, 7B. In 1998, 66.92% of voters in those same precincts voted for Thurbert Baker in his Attorney General race, and 68.91% voted for David Burgess in his 2000 Public Service Commissioner race. Pl.Exs. 2D, 7B.

Senator Brown is the current incumbent in Senate District 26. Brown was originally elected in a special election in 1991, defeating Robert Reichert, a white man, by a margin of 7,403 (56.71%) to 5,651 (43.29%) votes. PPFF ¶ 535. Senate District 26 at that time had a total black population of 46,623 (58.0%), and BVAP of 31,350 (52.79%). 47.4% of registered voters at that time were African American. Id.

In the 1992, 1994 and 1996 Democratic primaries and in the general elections, Brown was nominated and elected without opposition. Id. In the 2000 Democratic Party primary election, Brown had no opposition. Id. In the 2000 general election, Brown defeated Greg Williams, a Republican African American candidate, by a margin of 21,453 to 5,491 votes. Id.

Witnesses for the United States stated that, in their opinion, in Senate District 26, “most white voters vote for white candidates and most black voters tend to vote for black candidates.” U.S.Ex. 515 ¶ 6 (Abrams deck); Abrams dep. at 70; U.S.Ex. 517 ¶¶ 8, 9 (Bivins decl.); U.S.Ex. 519 ¶ 8 (Hart deck).

African American candidates have been successful at attracting some white votes in Bibb County. One of the United States’ witnesses, Albert J. Abrams, ran against a white opponent and won a seat on the Bibb County Board of Education in 1998, when the County’s BVAP was 43% and the percentage of black registered voters was 39.86%. Abrams dep. at 13:14, 11:13. Abrams acknowledged that he benefited from white crossover voting. Id. at 21. Samuel F. Hart, a District Commissioner and vice-chair of Bibb County Board of Commissioners, testified that he received a substantial portion of the white vote. Hart dep. at 24. Board of Education Member William Barnes also testified that he believed Senator Brown has been successful in winning white votes. Barnes dep. at 54-55.

Several of the United States’ witnesses testified that they did not believe that African American voters would be able to elect a candidate of choice other than the current incumbent, Senator Brown. See, e.g., U.S.Ex. 517 ¶ 10 (Bivins deck); U.S.Ex. 519 ¶ 9 (Hart deck); U.S.Ex. 725, 36:15-37:3. However, the witnesses also affirmed the existence of other viable African American candidates. Abrams dep. at 45, 61; Hart dep. at 44-45.

The United States’ expert analyzed local election data from the benchmark Senate District 26. The specific results of his assessment of racially polarized voting are found in Part 2G of this Memorandum Opinion.

d. Senate District 15

The newly drawn Senate District 15 is wholly contained within Muscogee County, Georgia. PkExs. 2A, 5A. The benchmark Senate District 15 encompassed all of Chattahoochee County and part of Musco-gee County. PkExs. 1A, 5A.

The benchmark Senate District 15 is 39,385 people, or 26.94%, short of the ideal district size. PPFF ¶ 562; PkExs. ID, 5B. Proposed Senate District 15 is only 6,821 *63persons, or 4.67%, short of the ideal district size. PPFF ¶469; PLExs. 2C, 5B.

Benchmark Senate District 15 has a total black population percentage of 65.75%, and BVAP of 62.05Ga.) or 60.93U.S.). PPFF ¶ 561. The proposed Senate District reduces the percentage of BVAP to 50.87% (Ga.) or 50.05% (U.S.). PPFF ¶ 568.

The percentage of black registered voters is reduced from 72.69% in the benchmark District 15 to 50.25% in the proposed District 15. Id. ¶¶ 563, 570. Black voter registration numbers have increased in the benchmark Senate District 15 over the past three election cycles, from 69.19% in 1996 to 70.79% in 1998 and finally to 73.00% in 2000. Id. ¶ 563; Pl.Exs. IE, 5B.

Voters within the benchmark Senate District 15 tend to vote heavily for Democratic Party candidates, as demonstrated by an overall Democratic performance score of 81.59%, as well as the Democratic performance levels for the individual election years of 1996 (71.23%), 1998 (82.93%) and 2000 (81.79%). PLExs. IE, 5B. 71.06% of the voters within the benchmark Senate District 15 voted for Michael Thurmond in the 1998 Democratic Party primary runoff election, and 87.59% supported him in the general election. PLExs. IE, 5B. Voters in benchmark Senate District 15 demonstrated electoral support for other African American Democratic candidates, voting for Thurbert Baker in 1998 at a rate of 88.95%, with 79.23% voting for David Burgess in 2000. PLExs. IE, 5B.

Proposed Senate District 15 retains an overall Democratic performance score of 65.91%, with projected Democratic performance levels of 61.99% using 1996 election results, 67.57% using 1998 results, and 63.44% using 2000 election results. Pl. Exs. 2D, 5B. 66.41% of voters in the precincts comprising the proposed district supported Michael Thurmond in the 1998 primary runoff, and 68.97% voted for him in the general election. PPFF ¶ 567. Additionally, 71.18% of these voters supported Thurbert Baker, and 70.34% voted for David Burgess. PLExs. 2D, 5B.

The current incumbent of Senate District 15 is Senator Ed Harbison, an African American man. PPFF ¶ 571. In the 1992 Democratic Party primary for Senate District 15, Harbison defeated Joseph Wiley, an African American man, for the party nomination. Id. Harbison defeated William Wright, his Republican African American opponent, in the 1992 General Election. Id. ¶ 572. Harbison had no opposition in the 1994, 1996, 1998 and 2000 Democratic primaries and general elections. Id.

e. Senate District 22

The newly drawn Senate District 22 is wholly contained within Richmond County, Georgia. PPFF ¶ 574; PLExs. 2A, 6A. Benchmark Senate District 22 is also wholly contained within ' Richmond County. PPFF ¶ 575; PLExs. 1A, 6A.

The benchmark Senate District 22 is underpopulated from the ideal district size by 37,675 people or 25.77%. PPFF ¶ 580; PLExs. ID, 6B. The proposed Senate District 22 makes up for this shortage, and is only 7079 people, or 4.85%, below the ideal size.

The benchmark Senate District 22 has a total black population of 66.84% (Ga.). PPFF ¶ 579. The proposed Senate District 22 has a total black population of 54.71%. Id. ¶ 586. The proposed Senate District 22 would experience a decrease in BVAP from 63.51% (Ga.) or 62.65% (U.S.), to 51.15% (Ga.) or 50.76% (U.S.). Id. ¶¶ 576, 586. The percentage of black registered voters would also fall from 64.07% to 49.44%. Id. ¶¶ 581, 587.

*64Voters within the benchmark Senate District 22 have supported Democratic Party candidates, as demonstrated by an overall Democratic performance score of 72.46%, and Democratic performance numbers for the individual election years of 1996 (71.02%), 1998 (71.34%), and 2000 (71.46%). Pl.Exs. IE, 6B. Specifically, 67.85% of the voters within the existing Senate District 22 voted for Michael Thurmond in the 1998 Democratic Party primary runoff election, and 76.64% supported him in the general election. PI. Exs. IE, 6B. Additionally, 73.97% of voters in the benchmark Senate District 22 voted for Thurbert Baker in 1998, and 77.41% voted for David Burgess in 2000. Pl.Exs. IE, 6B.

The proposed Senate District 22 has an overall Democratic performance score of 62.28%, and projected Democratic performance numbers of 60.98% using 1996 election results, 63.19% using 1998 results, and 59.98% for the 2000 election results. PPFF ¶ 584; Pl.Exs. 2D, 6B. 65.09% of voters in precincts comprising the proposed district supported Michael Thurmond in the 1998 primary runoff, and 66.91% voted for him in the general election. PPFF ¶ 585. Additionally, these voters supported Thurbert Baker by 63.87% and David Burgess by 69.05%. Id.; Pl.Exs. 2D, 6B.

Senator Charles Walker is the current incumbent of State Senate District 22 and is the Majority Leader in the Senate. Since his election to the Senate in 1990, Mr. Walker has faced only one white primary opponent, David Moretz. USPP ¶ 588. He has faced Republican opposition only one time, by a white man, D.L. Johnson, in the 2000 general election. Id. Senator Walker believes that he can win reelection, but noted that his successor would have to be “well-financed, relatively trained and had some previous political experience.” Walker Dep. 81:19-25-82:3.

G. Expert Testimony on Methodologies of Predicting Voting Patterns

The State of Georgia relied on a single expert, Dr. David Epstein, to attempt to meet its burden of proof on the State House, State Senate and United States Congressional redistricting plans. Epstein relies on a single methodology, probit analysis, for his conclusions as to all three plans. The United States' expert, Dr. Richard Engstrom, relied on three variations of regression analysis, while the in-tervenors’ expert, Dr. Jonathan Katz, limited the scope of his report to a critique of Epstein’s probit analysis. The court is struck by the narrow scope of Epstein and Engstrom’s reports. Epstein’s report addresses the limited question of what percentage of BVAP will produce a 50% chance that an African American candidate of choice will win election. Engstrom’s report attempts to ascertain the degree of racially polarized voting in existing Senate Districts 2, 12 and 26. Neither report focuses its inquiry on the question of whether the proposed redistricting plans are retrogressive.

1. Probit Analysis

Plaintiffs expert report describes probit analysis as a standard statistical method for determining the likelihood of an event that has two possible outcomes. Epstein describes the method as a fitting tool to allow political scientists to determine the likelihood that an African American candidate of choice will win. There are only two possible outcomes with which the analysis is concerned — the candidate of choice wins or loses. Epstein’s analysis is designed to predict the “point of equal opportunity,” or the point at which the demographics of a district will result in a 50% *65chance that an African American candidate of choice will win election. The demographics of the district are represented by the percentage of BVAP, as calculated by Georgia. While it appears true that probit analysis is a standard statistical technique, no court has relied on such an analysis in reviewing a reapportionment plan.

Epstein’s database includes 1,258 elections, all elections in the relevant legislative offices held since 1996. The database identifies the district in which an election was held, the district’s BVAP at the time of the election, the race of the incumbent or whether the election was an open seat or special election, the race of the winner, and whether the winner was an African American candidate of choice.

Epstein proposes that an African American “candidate of choice” must necessarily be determined by some element other than race. Epstein defines a candidate of choice as any African American candidate who won election, or any white candidate who won election in a district with a majority BVAP and who received a majority of the African American votes cast.30 Thus, the database of elections does not include any white candidates who received the majority of African American votes in districts with less than 50% BVAP, whether they won or lost. Tr., 2/4/02, p.m. at 44. Intervenors’ expert, Dr. Katz contends that Epstein’s definition of a candidate of choice is arbitrary. He suggests that a “more natural definition of candidate of choice would be to calculate district by district, using ecological inference techniques, which candidate receives the majority vote of Blacks.” Int.Ex. 25 at 9.

Epstein’s database includes election data from Congressional, Senate and House races. He predicts one “equal opportunity number” from this data, and does not perform separate analyses for the Congressional, Senate and House elections. In his report, he states that “as a matter of political and voting behavior, voting patterns on legislative races are generally similar from one body to another.” Pl.Ex. 25 at 14. Epstein provides no explanation or justification for this conclusion other than his opinion that the more elections are in the database, the better. While a footnote to Epstein’s statement suggests that racial voting patterns throughout Georgia are “sufficiently similar” to warrant the combination of election results in the database, on cross-examination Epstein clarified that this footnote referred to his conclusion that there was no need to perform regional analyses. Id. 14, n. 11. Epstein testified that he reached this conclusion by comparing estimates of white crossover vote in the current Senate districts in three 1998 statewide races.

Epstein testified that, before he performed his probit analysis, he first determined that it was permissible to treat the State as a unified whole by evaluating white crossover voting.31 He uses the King methodology of ecological inference *66to analyze the precinct level voting patterns, and to estimate the percentage of white voters who “crossed over” to vote for African American candidates in each of the existing Senate Districts. U.S.Ex. 122. These estimates are based on data from three 1998 statewide general elections. Id. The average estimated white crossover voting in those districts ranges from 24.73% to 57.39%. Id. Epstein testified that he did not consider this range to be statistically significant because the crossover estimates were regularly distributed, and all but two “outliers” fell within the expected range. Epstein further testified that, after his first deposition, he expanded his analysis to look at black crossover voting and black registration levels as indicators of minority voter mobilization. Tr., 2/4/02, p.m. at 68-69. These additional analyses are not part of the record before the court. However, on redirect, Epstein testified that he got the “exact same result” when he ran these analyses, finding that black crossover and black registration levels were in a “tight distribution,” with “very few outliers.” Tr., 2/5/02, a.m. at 98.

To reach his conclusions regarding the non-retrogressive effect of the proposed plans, Epstein relies only upon an analysis of open seat elections. The database includes only 158 open-seat elections. Epstein finds that the point of equal opportunity is 44.3% BVAP, which means that “there’s a 50-50 chance of electing a candidate of choice” in a district with an open seat and with 44.3% BVAP. Tr., 2/5/02, a.m. at 29; Pl.Ex. 25 at 17. Epstein explains this result by noting that over 90% of African American voters support Democrats, while over 60% of white voters support Republicans. Id. at 17, n. 14. Thus, an African American supported candidate would usually be able to win a primary in a 44% BVAP district because African Americans would form the majority of Democratic voters, and the candidate would receive enough white Democratic crossover votes to win the district overall.

Epstein also considers the races of African American incumbents, but finds that a probit analysis is inappropriate for these races. Of the recent 200 races with an African American incumbent, only once was the African American incumbent defeated. Hildred Shumake lost in 1992; however, at the time of the election, Shu-make was under indictment on extortion charges and suspended from the State Senate. African American incumbents have run in districts with as low as 35% BVAP and have won. Therefore, to assess the likelihood of an African American incumbent losing to a white candidate who was not a candidate of choice, Epstein combined the open seat data with the African American incumbent elections. This yielded an equal opportunity point of 37.4% BVAP.

The point at which an African American candidate has a 50% or greater chance of unseating a white incumbent occurs at 56.50% BVAP. Senate District 12 is currently represented by Senator Meyer Van Burén, a white man. The existing district has a 55.25% BVAP, which would be reduced to 50.66% BVAP under the proposed plan. Epstein testified that, according to his calculations, an African American candidate of choice would have less than a 50% chance of being elected in a district with 55.25% BVAP and a white incumbent; the estimated probability of an African American candidate of choice being elected in a district with a white incumbent and 50.66 BVAP falls to approximately 29%-30%.

On the first day of trial, Epstein produced an “S curve,” which had previously not been provided to the court or to opposing counsel. PLEx. 109. He testified that the curve was not a necessary part of his *67analysis. Rather, he characterized it as a “visual aid” to explain the correlation of a district’s BVAP with the probability that a minority preferred candidate would be elected. Tr., 2/5/02, a.m. at 15-16. The illustration is an “S curve or a probability curve representing the results of the pro-bit analysis done for the open-seat races in [Epstein’s] database.” Id. at 18. The curve plots the probabilities that an African American candidate of choice will be elected at a given level of BVAP. This calculation is based on a formula that assumes normal probability distribution of the existing data points. Id. at 23.

Epstein testified that, in a district with 50% BVAP, there is a 75% probability that an African American candidate of choice will be elected. Thus, in districts with BVAPs between 44% and 50%, Epstein’s calculations show that the probability that an African American candidate of choice ■will be elected increases from 50% to 75%, representing a steep increase in the curve. Tr., 2/5/02, a.m. at 29. Epstein further testified that a “cumulative, normal curve[ ],” “by definition,” “is steepest right at its midpoint.” Id. at 30.

The proposed Senate plan includes six districts, in which the BVAP is between 50.3% and 51.5%. Estimates for these districts, therefore, would be in the “steepest” portion of the curve. Given the absence of tick marks on the axes or plotted data points, nothing in the record permits the court to know what corresponding probabilities Epstein would assign to these districts. However, the probabilities would presumably all be higher than 75%, the probability Epstein correlates with 50% BVAP.

There are very few elections in Epstein’s database that fall along the steep part of the S curve. In only two of the 158 open-seat elections does Epstein indicate that an African American candidate of choice was elected in a district with less than 53% BVAP. Pl.Ex. 25, app. II (Cong. Dist. 4; House Dist. 89). Of the 158 open-seat election data included in the database, only thirty-six of them represent elections in districts with BVAPs of 30%-70%. According to Epstein’s database, eight elections took place in districts with BVAPs of 31%-36%, only one of which resulted in the election of a candidate of choice. After having served as a Congressional Representative, Cynthia McKinney was elected in a redrawn Congressional District that had 33% BVAP.

Epstein’s open seat database contains ten elections in districts with BVAPs of 49%-54%: five were won by candidates of choice, four were not won by candidates of choice and one was indeterminate.32 There are 18 districts with BVAPs of 56%-68%, all of which elected candidates of choice.

Epstein admits that it would be preferable to have data points in the steepest region, but states that political scientists often do not have all the data that they would desire. Tr., 2/4/02, p.m. at 31. Georgia, in its Post-Trial Brief, argues that the absence of such data points can not reflect poorly on its case and, in fact, the absence of such points is the result of previous unconstitutional, race-based poli*68cies of the Department of Justice. Post-Tr.Br. at 9; PPFF ¶ 294.

Intervenors’ expert, Dr. Katz, suggests that Epstein should have identified a margin of error to be applied to his estimates. Int.Ex. 25 at 9. Epstein states that his data are normally distributed, and that the statistical accuracy of the probit analysis is demonstrated by the fact that the curve represents over 80% of his data. However, the United States suggests that, due to the large number of elections in the open seat database in districts with high and low percentages of BVAP, an explanation of 80% of the data does not prove the accuracy of Epstein’s estimates at the steepest part of the curve.

The United States' argues that probit analysis is extremely sensitive to even single data points. In cross-examination, defendants focused on Epstein’s coding of Cynthia McKinney’s victory in the Fourth Congressional District as a race for an open seat. McKinney’s district was redrawn in 1996, and the proposed district in which she ran retained approximately a third of her constituents.33 Epstein testified that, had he taken McKinney’s race out of his “open-seat” database, his point of equal opportunity would rise by approximately two points. Tr., 2/5/02, a.m. at 43. Thus, the removal of one race would cause his estimation of the point at which an African American candidate of choice has a 50% chance of winning election would rise from 44% to 46%.

Epstein’s analysis of whether the reapportionment plans will retrogress consists of comparing the number of majority BVAP districts under the benchmark and proposed plans, and comparing the number of districts under the benchmark and proposed plans that have BVAPs greater than 44.3%. His results are represented in the following table.

[[Image here]]

Pl.Ex. 25 at 18.

Epstein, without explanation, used the “equal opportunity” point for open seat districts to do this comparison. According to the results of his analysis, a redrawn district with a white incumbent would need to have 56.5% BVAP in order for an African American candidate of choice to have a 50% chance of being elected.

There are six districts in the proposed senate plan where BVAP is between 50.3% and 51.5%. Epstein does not consider the effect of reducing BVAPs in majority-minority districts to bare majorities.

Intervenors’ expert also notes that Epstein’s data reflects only elections in single-member districts. Int.Ex. 25 at 11. Epstein does not explain whether his analysis is applicable to the proposed multi-member districts created in the State House. Id. In the proposed House plan, there are 124 single-member districts, 15 districts with two members, 6 districts with 3 members and 2 districts with four members. Epstein merely counts the number of seats in the proposed multi-member districts for purposes of his comparison of the number majority-minority seats under the proposed and benchmark plans.

*692. Engstrom’s Use of Ecological Regression Analysis to Assess Racially Polarized Voting

The Supreme Court has relied on regression analysis to assess the severity of racial bloc voting and whether existing voting patterns would prohibit a minority population from electing candidates of choice. See, e.g., Thornburg v. Gingles, 478 U.S. 30, 52-53, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). This regression analysis is also referred to as “ecological inference” methodology, and allows political scientists to infer voting behavior from aggregate information. Engstrom employs three types of regression analysis: ecological inference, King’s ecological inference34 and homogeneous precinct analysis.35

The United States’ expert report, submitted by Engstrom, clearly describes racially polarized voting patterns in Senate Districts 2,12 and 26. U.S.Ex. 601. However, in the report submitted with the government’s direct testimony, Engstrom does not attempt to predict the effect of this polarized voting on the ability of minority voters to elect candidates of their choice under the proposed redistricting plans.

Engstrom’s report analyzes the extent to which the candidate preferences of African American and non-African American voters within Georgia State Senate districts 2, 12 and 26 have differed in recent elections, in which they were presented with a choice between African American and non-African American candidates. Id. at 2. Engstrom reviews voting statistics from five elections for State Senate seats. These include, in Senate District 2, the 1999 special election, the 2000 special runoff election, and the 2000 general election; and, in Senate District 12, the 1996 and 1998 Democratic primary elections. Id. at 2-3. In addition, Engstrom reviews voting patterns in the three senate districts in seven biracial statewide elections. Id. at 3. These elections include the 1998 democratic primary for Insurance Commissioner and Labor Commissioner, the 1998 runoff Democratic primary for Labor Commissioner, the 1998 general election for Insurance and Labor Commissioners, as well as for Attorney General, and the 2000 Democratic primary for Public Service Commissioner. Id. Engstrom also analyzes data from three biracial eounty-*70wide elections from the predominant counties within each of the Senate districts, and from four biracial citywide elections from the largest cities within these counties. Id. Specifically, Engstrom analyzed Savannah city elections for Senate District 2, Bibb County and Macon City elections for Senate District 26, and elections from Albany City, Doughtery County and the existing House Districts 161 and 162 for Senate District 12. Id. at Tables 1-3.

Engstrom’s analysis of the elections involving Regina Thomas in Senate District 2 demonstrates the effect of racially polarized voting and the minimal amount of so-called “white crossover.” In the 1999 special election, there were four African American candidates, and two white candidates. While 80%36 of African Americans voted for one of the four African American candidates, only 20.4% of non-African Americans voted for one of the four candidates. Id. at Table 1. In the special runoff between Regina Thomas, an African American candidate, and a white candidate, Thomas received 78.8% of the African American vote, while she received an estimated 8.9% of the non-African American vote. Id. In the 2000 election, Thomas is estimated to have received 99.2% of the African American vote and 43.6% of the non-African American vote. Id.

Senate District 12 currently has 55% BVAP and 52.5% African American voter registration in November 2000. Id. at 8. John White, an African American candidate, has twice won and lost against a white opponent in the Democratic primary. Id. at 8-9. He lost in 1996, with approximately 65% of African Americans supporting him, and approximately 10% of non-African Americans supporting him. Id. at Table 2. In 1998, he received 90% of African American votes and 17.5% of non-African Americans votes.- Id. He lost, winning .only 48.6% of all votes. Id. at 9.

In Senate District 26, no bi-racial elections have occurred since the 1996 redis-trieting. Id. at 10. Bibb County, which appears to be within the proposed Senate District 26, had bi-racial elections for its Board of Education, and for the Democratic primaries for District Attorney and County Commissioner. Id. at Table 3. In the two Board of Education races, the African American candidates received 99.5% and 99.3% of the African American vote, and 34.2% of the non-African American vote. Id. In the primary for District Attorney, the African American candidate received 68.4% of the African American vote, and 16.6% of the non-African American vote. Id. In the Democratic primary for Chair of the Bibb County Commission, an African American candidate received 64.8% of African American votes and 2.7% of non-African American votes. Id.

In none of the municipal or county elections analyzed by Engstrom did the preferences of African American voters and non-African American voters coincide. Engstrom concludes that his analysis of the three districts in question “reveal a pattern of racially polarized voting in these areas.” Id. at 11.

Engstrom attempts to simulate the voting patterns of the proposed Senate districts. In order to do this, he considers precinct-level information from seven statewide races between African American and white candidates, “reaggregating” the data to reflect the contours of the three proposed districts. Id. In the seven hypothetical elections in the reaggregated dis*71tricts, an African American candidate lost only once in one reaggregated district. Id. at Table 4. On cross-examination, Eng-strom testified that there would be a “very good chance” that, based on the voting patterns revealed by his reaggregation analysis, African American candidates would win election in the reconstituted districts. Tr., 2/5/02, p.m. at 88. However, he concludes that his reaggregation results were “not a good indication that these proposed districts will provide African American voters with a realistic opportunity to elect representatives of their choice, given that the level of crossover voting tends to be considerably higher in these elections than in the senate and other elections involving local candidates.” U.S.Ex. 601 at 12. Engstrom emphasizes that African American candidates consistently received less crossover voting in local election than in statewide elections. Id.

Engstrom further justifies his conclusion that the local election analysis was more probative than his reaggregation analysis by describing a second reaggregation analysis he performed. When Engstrom reaggregated the statewide election data to reflect the existing Senate Districts, he found that African American candidates were supported at similar levels as those reflected in Table 4 of his report. Tr., 2/5/02, p.m. at 91. However, he noted that, in the existing Senate District 2, Ms. Thomas barely won her runoff election in 1999, by some 73 votes, and that, in the existing District 12, Mr. White had not succeeded in 1996 or 1998. Id. Thus, he concluded that reality did not reflect the favorable prediction of the reaggregation analysis.

III. Conclusions of Law

The court has jurisdiction to hear this case pursuant to 42 U.S.C. § 1973c; 28 U.S.C. § 1346(a)(2); 28 U.S.C. § 2201. This court is properly convened as a three-judge court pursuant to 42 U.S.C. § 1973c and 28 U.S.C. § 2284. None of the parties contests the applicability of Section 5 to this matter.

The State of Georgia is a covered jurisdiction as defined under Section 5 of the Voting Rights Act of 1965, as amended. 42 U.S.C. § 1973c; 28 C.F.R. § 51 (Appendix). Section 5 applies to any changes in voting processes in Georgia, and mandates that the State receive pre-clearance prior to instituting any such changes. Georgia v. United States, 411 U.S. 526, 527-28, 93 S.Ct. 1702, 36 L.Ed.2d 472 (1973). The reapportionment of seats in Georgia’s Congressional delegation and General Assembly seats for the House and Senate are changes in voting procedures covered by the Voting Rights Act and require preclearance.

In an action for declaratory judgment pursuant to Section 5, the plaintiff has the burden of proving by a preponderance of the evidence the absence of both discriminatory effect and discriminatory purpose in the reapportionment of its legislative districts. City of Pleasant Grove v. United States, 479 U.S. 462, 469, 107 S.Ct. 794, 93 L.Ed.2d 866 (1987).

A. Court’s Obligations in Action for Declaratory Judgment

The State of Georgia made the strategic decision to institute an action in this court for declaratory judgment and not to seek administrative preelearance from the Department of Justice. The vast majority of changes in voting procedures subject to Section 5 are submitted to the Department of Justice for preclearance. The majority of cases that come to this court represent plans or changes to which the Department of Justice has objected. Here, Georgia’s redistricting plans have been submitted in the first instance to this court for review.

*72When the plans were submitted for judicial review, Georgia did not know whether the Attorney General would object to the plans and, if he did, which plans and which districts would be considered problematic. The Attorney General eventually identified only the Senate redistricting plan as objectionable, and, in particular, proposed Senate Districts 2,12 and 26.

This case presents a unique circumstance wherein the Attorney General has not objected to two of three redistricting plans proposed by the State of Georgia, and yet the State has come to this court seeking judicial — and not administrative— preclearance of all three plans. The United States and intervenors argue that the United States’ failure to object to the two plans does not justify entry of declaratory judgment because the State retains the burden of proof and because the objections of intervenors preclude entry of a declaratory judgment. The State argues that the failure of the United States to object to the Congressional redistricting plan and the State House plan relieves this court of any obligation to make findings with respect to those plans. The court does not agree. The State chose not to seek administrative preclearance. In asking this court to enter a declaratory judgment as to all three plans, it imposes on this court an affirmative duty to inquire whether the plans have the effect or purpose of denying or abridging the right to vote on account of race or color. Furthermore, the State assumes the burden of demonstrating by a preponderance of the evidence that such a declaratory judgment is warranted.

Plaintiff cites Morris v. Gressette as evidence that Congress intended that courts review requests for declaratory judgment only where the Attorney General objects to such an entry of judgment. 432 U.S. 491, 97 S.Ct. 2411, 53 L.Ed.2d 506 (1977). In Morris, the Supreme Court characterized the difference between the preclearance avenues as one based on the distinction between affirmative action and silent acquiescence by the Attorney General:

This [administrative preclearance] method of compliance under § 5 is unlike the [declaratory judgment preclearance method] in that implementation of changes in voting laws is not conditioned on an affirmative statement by the Attorney General that the change is without discriminatory purpose or effect. To the contrary, compliance with § 5 is measured solely by the absence, for whatever reason, of a timely objection on the part of the Attorney General.

Id. at 502, 97 S.Ct. 2411. Yet, Section 5 clearly presents two different methods of preclearance. The State cannot point to the Attorney General’s apparent acquiescence — a circumstance relevant under the statute when administrative preclearance is sought — to justify a grant of judicial preclearance. Had the Attorney General administratively precleared the United States Congressional redistricting plan and State House redistricting plan, the current matter may well have been mooted. See City of Dallas v. United States, 482 F.Supp. 183 (D.D.C.1979) (three-judge court) (holding that Section 5 did not provide a forum for intervenors to challenge a voting plan after the Attorney General had approved a revised voting plan).

However, the idea that Congress intended this three-judge court to be a rubber stamp is simply untenable. In City of Richmond v. United States, the Supreme Court reviewed a decision of this court that had engaged in a Section 5 inquiry where the United States and the City of Richmond had entered a consent decree, but citizen intervenors objected to the entry of the consent decree. 422 U.S. 358, 366, 95 S.Ct. 2296, 45 L.Ed.2d 245 (1975). If the State were correct that this Court’s *73jurisdiction is stripped when the United States fails to object to a plan submitted for judicial preclearance, the three-judge court in City of Richmond would have been without jurisdiction to consider inter-venors’ claims. While the Supreme Court’s decision does not comment on this issue, neither did the Court question the district court’s exercise of jurisdiction and, in fact, remanded the case to the district court for further proceedings. Id. at 378, 95 S.Ct. 2296.

Similarly, we reject the State’s argument that this court’s review is limited only to those districts challenged by the United States, and should not encompass the redistricting plans in their entirety. In a declaratory judgment action brought pursuant to Section 5, the court’s review necessarily extends to the entire proposed plan. Refusing to preclear only the specific districts to which defendants object would nevertheless require the State to rework its entire Senate plan. Moreover, Georgia has presented no legal authority that would limit the Section 5 inquiry to those districts challenged by the Attorney General as retrogressive. Indeed, the very structure of the declaratory judgment procedure, under which the court, and not the Attorney General, is vested with the final authority to approve or disapprove the proposed change as a whole, argues conclusively against the State’s suggestion.

B. Legal Standard for Assessing Retrogressive Effect

The legal standard for reviewing redistricting plans submitted for Section 5 preclearance has been defined in a deceptively simple manner. In Beer, the Supreme Court held that a reapportionment plan must not “lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Beer, 425 U.S. at 141, 96 S.Ct. 1357. In other words, the new apportionment plan must not have the effect or purpose of providing minority voters with less opportunity to elect candidates of choice than did the previous plan. Id.

All parties in this action make liberal use of the word “retrogression” in their arguments, but shy away from any attempt to define retrogression. The court can not so easily avoid this task.

Section 5 does not focus on the discriminatory impact of a plan, an inquiry which would measure the plan against an ideal, non-discriminatory plan. See Bossier II, 528 U.S. at 328, 120 S.Ct. 866; cf. City of Lockhart v. United States, 460 U.S. 125, 134, 103 S.Ct. 998, 74 L.Ed.2d 863 (1983) (finding that a change that did “not increase the degree of discrimination against blacks ... [was] entitled to § 5 preclearance.”). Rather than an objective ideal, the Supreme Court has explained that retrogression is to be measured with respect to the status quo, reflecting Congress’ intent to “freeze” the existing procedures if a proposed change would retrogress. Bossier I, 520 U.S. at 477, 117 S.Ct. 1491. Preclearance must be denied if a proposed change “ ‘abridges the right to vote’ relative to the status quo.” Bossier II, 528 U.S. at 334, 120 S.Ct. 866; see Holder v. Hall, 512 U.S. 874, 883-84, 114 S.Ct. 2581, 129 L.Ed.2d 687 (1994) (“The baseline for comparison is present by definition; it is the existing status.”); City of Lockhart v. United States, 460 U.S. 125, 132-33, 103 S.Ct. 998, 74 L.Ed.2d 863 (1983) (proper comparison is between new system and system actually in effect prior to adoption of new charter, regardless of what state law might have required); Texas v. United States, 785 F.Supp. 201, 203 (D.D.C.1992) (preclearance involves a comparative inquiry into whether the change in voting procedures is retrogressive when compared with the plan that would other*74wise be in force and effect). In the context of a vote-dilution claim, the Court has held that preclearanee, which rests on a finding of non-retrogression, “is nothing more than a determination that the voting change is no more dilutive than what it replaces.” Bossier II, 528 U.S. at 335, 120 S.Ct. 866.

The Supreme Court has never comprehensively defined “retrogression,” nor has it engaged in any detailed discussion of what constitutes an “effective exercise of the electoral franchise” by minority voters. But see Holder v. Hall, 512 U.S. 874, 895-903, 114 S.Ct. 2581, 129 L.Ed.2d 687 (Thomas, J., concurring) (stating that the Court has tacitly selected the number of elected officials as its indicator of electoral strength). Section 5 cases have focused almost exclusively on evaluating whether a proposed change would leave minority voters in a “worse” position than under the existing plan. See Bossier I, 520 U.S. at 487, 117 S.Ct. 1491. The Court has clearly held that compliance with Section 5, and avoidance of retrogression, does not require jurisdictions to improve or strengthen the voting power of minorities. Bush v. Vera, 517 U.S. 952, 983, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996). Nor does Section 5 require that redistricting plans ensure victory for minority preferred candidates. Rather, it is a mandate that “the minority’s opportunity to elect representatives of its choice not be diminished, directly or indirectly, by the State’s actions.” Bush, 517 U.S. at 983, 116 S.Ct. 1941 (emphasis added); accord Hale County v. United States, 496 F.Supp. 1206, 1216 (D.D.C.1980) (principle of nonretrogression requires that covered jurisdictions demonstrate that minority voters maintain “at least as much electoral power as they possessed under” the existing system, and that proposed change will not result in a “retreat from the potential advantage previously held by black voters....”) (emphasis added).

While slightly more detailed definitions of retrogression are found in decisions by three-judge panels of this court, they simply extend the underlying rationale that preclearance must be withheld from a plan that would diminish minority voting strength—directly or indirectly. In Texas v. United States, the court explained that “[t]his rule mandates that preclearance be denied under the ‘effects’ prong of Section 5 if a new system places minority voters in a weaker position than the existing system.” 866 F.Supp. 20, 27 (D.D.C.1994); accord Arizona v. Reno, 887 F.Supp. 318, 320 (D.D.C.1995) (“any change which would place a protected minority group in a position worse than its position [under the benchmark plan] does not merit clearance”); New York v. United States, 874 F.Supp. 394, 397 (D.D.C.1994) (“[i]f the position of minority voters is no worse under the new scheme than it was under the old scheme, then the proposed change is entitled to preclearance under section 5”).

While courts have frequently considered the number of “majority-minority” districts as indicative of minority voting strength, the parties in this matter apparently agree that Section 5 is not an absolute mandate for maintenance of such districts. This agreement is entirely proper. It is true that in United Jewish Organizations of Williamsburgh, Inc. v. Carey, a plurality of the Supreme Court construed the retrogression standard to mean that “had there been districts with black majorities under the previous law, and had [the covered jurisdiction] in fact decreased the number of majority black districts, it would have had to modify its plan in order to implement its reapportionment by carving out a large enough black majority in however many additional districts would *75be necessary to satisfy the Beer test.” 430 U.S. 144, 159-60, 97 S.Ct. 996, 51 L.Ed.2d 229 (1977) (White, J., joined by Stevens, Brennan, and Blackmun, JJ.); accord Ketchum v. Byrne, 740 F.2d 1398, 1402 n. 2 (7th Cir.1984) (defining “retrogression” as “a decease in the new districting plan or other voting scheme in the absolute number of representatives which a minority group has a fair chance to elect”). This statement in United Jewish Organizations, however, was pure dicta: the plurality’s holding was simply that Section 5 authorizes some consideration of race in drawing district lines. 430 U.S. at 161, 97 S.Ct. 996. Moreover, the Court’s subsequent cases have analyzed the issue of the creation and maintenance of majority-minority districts in the broader context of assessing minority voting strength in a given jurisdiction.

Indeed, in the context of Section 2 cases, the Court has observed that majority-minority districts do not inherently increase or decrease minority voting strength, but rather can have “either effect or neither.” Voinovich v. Quilter, 507 U.S. 146, 154-55, 113 S.Ct. 1149, 122 L.Ed.2d 500 (1993). Breaking apart a majority-minority district and dispersing minority voters into neighboring districts can have different consequences in different contexts. On the one hand, it can diminish minority voters’ power by “fragmenting [them] among several districts where a bloc-voting majority can routinely outvote them.” De Grandy, 512 U.S. at 1007, 114 S.Ct. 2647. On the other hand, such dispersal can actually increase electoral opportunity if it eliminates “packing” whereby the minority voters are crammed into a small number of “safe” districts and deprived of an ability to influence a greater number of elections. Id.; cf. Holder v. Hall, 512 U.S. 874, 898-903, 114 S.Ct. 2581, 129 L.Ed.2d 687 (1994) (Thomas, J., concurring) (suggesting that Section 2 case law is inconsistent because some courts emphasize the need for majority-minority districts to equalize voting power while others have found that such districts have the effect of isolating minority voters and limiting their electoral strength).

Thus, in an area where voting patterns are not polarized according to race, distributing African American voters out of a single district in which they were a majority and creating substantial minorities in a larger number of districts may increase the voters’ ability to elect candidates of choice. See Hays v. Louisiana, 936 F.Supp. 360, 364 n. 17 (W.D.La.1996) (noting that one plan “with its one black majority and three influence districts, empowers more black voters statewide than does” a plan with two black-majority districts and five “bleached” districts in which minority influence was reduced in order to create the second black-majority district); cf. Shaw v. Hunt, 517 U.S. 899, 947 n. 21, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996) (Stevens, J., dissenting) (“Of course, a State that unfairly ‘packs’ African-American voters into a limited number of districts may be subject to a § 2 challenge on the ground that it has failed to create so-called ‘influence districts’.... ”).

As such, the Court has suggested that the propriety — and even the legality — of majority-minority districts depends on a careful analysis of the facts and circumstances surrounding their use, and particularly of the nature of “society’s racial and ethnic cleavages.” De Grandy, 512 U.S. at 1020, 114 S.Ct. 2647. Therefore, in asking whether the elimination of a majority-minority district — or the reduction of African American population in such a district — is actually retrogressive, a court must take account of the fact that “there are communities in which minority citizens are able to form coalitions with voters from other racial and ethnic groups” and thus have less *76need to “be a majority within a single district in order to elect candidates of their choice.” Id.

In a Section 5 case, this court’s analysis — while limited to the question of retrogression — is fact-intensive and must carefully scrutinize the context in which the proposed voting changes will occur. In particular, the level of racially polarized voting, or the degree to which there is a correlation “between the race of a voter and the way in which the voter votes,” Thornburg v. Gingles, 478 U.S. 30, 53 & n. 21, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986), sheds light on whether a decrease in districts’ minority populations will produce an impermissibly retrogressive effect. “Unpacking” African American districts may have positive or negative consequences for the statewide electoral strength of African American voters. To the extent that voting patterns suggest that minority voters are in a better position to join forces with other segments of the population to elect minority preferred candidates, a decrease in a district’s BVAP may have little or no effect on minority voting strength. De Grandy, 512 U.S. at 1020, 114 S.Ct. 2647. In such circumstances, dilution of minority voting age population may have no retrogressive effect. However, if racially polarized voting persists in an area and its electoral history demonstrates that minority voters’ preferences diverge greatly from those of non-minority voters, a decrease in BVAP may translate into a lessening of minority voting strength.

In scrutinizing the plans presented for preclearance, we must therefore consider whether the State has met its burden of demonstrating that the dispersion of African American voting age population throughout .the districts is not so affected by racial bloc voting that it will have a negative impact on the opportunities available to Georgia’s African American voters to make their collective voices heard.

Before beginning this analysis, we must first address the State’s contention that the retrogression inquiry is limited to determining whether reapportioned districts provide minority voters with an “equal opportunity” to elect minority candidates. Georgia contends that because its plan preserves for black voters a reasonable— or equal — chance to elect candidates of choice in the three districts at issue, the State has satisfied Section 5. Georgia thus asks us to apply a Section 2 test to the proposed plans. See, e.g., Quilter, 507 U.S. at 155, 113 S.Ct. 1149 (“Only if the apportionment scheme has the effect of denying a protected class the equal opportunity to elect its candidate of choice does it violate § 2.”). The State’s implicit argument is that retrogression cannot exist where its proposed plan satisfies Section 2.37 We disagree.

The Supreme Court has clearly articulated that Section 2 and Section 5 have separate functions in the scheme of the Voting Rights Act. Holder, 512 U.S. at 883-84, 114 S.Ct. 2581 (Section 5 and Section 2 differ in “structure, purpose, and application”); Morse v. Republican Party of Virginia, 517 U.S. 186, 209-10, 116 S.Ct. 1186, 134 L.Ed.2d 347 (1996) (noting that Section 5 may cover a broader range of voting procedures than does Section 2).

These distinct roles are highlighted by the different inquiries mandated by Section 2 and Section 5. Under Section 5, “[i]f the change ‘abridges the right to vote’ *77relative to the status quo, preclearance is denied....” Bossier II, 528 U.S. at 334, 120 S.Ct. 866. The comparison here is not to an abstract ideal such as equality or proportionality, as in a Section 2 case, but rather to a concrete, existing plan. See Hall, 512 U.S. at 883-84, 114 S.Ct. 2581 (finding no Section 2 violation because the Court could not identify the “ideal” benchmark). The relativistic inquiry prescribed by the Court means that the nature of the existing benchmark plan will set the terms of the retrogression analysis. The fact that a weak status quo may make preclearance easy does not mean that a stronger status quo should not make preclearance more difficult. The Supreme Court’s distinction between the types of benchmarks implicated by Section 5 and Section 2 compels this court to conclude that a plan that diminishes existing minority strength, but which is not discriminatory, should not be precleared.

That the court rejects the State’s position that a Section 5 violation can not exist without a violation of Section 2 does not mean that the question of whether its plans are in fact non-diseriminatory is irrelevant to the retrogression inquiry. See Bossier I, 520 U.S. at 486-88, 117 S.Ct. 1491. Evidence that a plan satisfies Section 2 by preserving reasonably good opportunities for African American voters to elect candidates of choice may bear on whether there has been impermissible retrogression under Section 5. Undoubtedly, a change that has this effect is less likely to be marked by retrogressive intent. Yet, if existing opportunities of minority voters to exercise their franchise are robust, a proposed plan that leaves those voters with merely a “reasonable” or “fair” chance of electing a candidate of choice may constitute retrogression in overall minority voting strength.

Accordingly, the court’s inquiry in this case will concentrate on whether the State’s proposed reapportionment plans will diminish African American voters’ opportunity to exert electoral power at the polls. While the parameters of the court’s investigation are clear, and are grounded in settled law, this case presents a unique factual context. Georgia’s State House and State Senate reapportionment plans were drafted to bolster support for the Democratic Party, in part by “unpacking” predominantly African American districts. The State Senate plan would redraw four districts with existing BVAPs of 55.43% to 62.45% such that they would have bare majorities of BVAP, ranging from 50.31 to 50.87%. The slim nature of these majorities is cause for concern when considered in conjunction with levels of black voter registration in the districts. In five of the “majority-minority” districts created under the reapportionment plan, percentages of black voter registration range from 47.46 to 49.44%.38 Under the benchmark plan, African American registration levels in these same districts ranges from 52.48% to 64.07%.

The appropriate analysis of these changes focuses on whether they are likely to diminish minority voters’ ability to effectively exercise their franchise. The mere fact of dilution, the spreading out of minority voters, is not unlawful in the Section 5 context, at least to the extent that it does not lead to a palpable decrease in minority voting strength. See Bossier Parish II, 528 U.S. at 327, 120 S.Ct. 866 (finding that an intent to dilute in violation of Section 2 is not necessarily an intent to retrogress in violation of Section 5). In *78particular, if voting patterns are not marked by racially polarized voting or other barriers to the effective exercise of minority voters’ franchise, dilution may have little or no effect on the ability of those voters to elect preferred candidates.

Accordingly, contrary to the fears expressed by plaintiff, the Voting Rights Act allows states to adopt plans that move minorities out of districts in which they formerly constituted a majority of the voting population, provided that racial divisions have healed to the point that numerical reductions will not necessarily translate into reductions in electoral power. In contrast, the more that white voters prove unwilling to cast their ballots for candidates preferred by minorities, and the more that different races decline to support the same candidates, the greater the negative impact of decreasing the percentage of minority voters will be on the electoral strength of those who are left behind.

In sum, then, just as cross-racial coalition-building — the opportunity “to pull, haul, and trade to find common political ground,” De Grandy, 512 U.S. at 1020, 114 S.Ct. 2647—can allow smaller numbers to extend great influence, so too can its antitheses, racial-bloc voting, allow a drop in numbers to become a retrenchment of power. Thus, the more we find evidence of racial polarization in the disputed Senate districts, the more we are persuaded that Georgia has failed to meet its burden of proving that the reductions in African American populations in those districts and in other majority-minority districts has not lessened the ability of its African American voters to effectively exercise their collective right to vote.

C. Effect of Georgia’s Reapportionment Plans

In considering each of the three redistricting plans submitted for review, the court must determine whether the State has proven that the plans will not have a retrogressive effect on minority voting strength.

Measuring Minority Voting Strength

In large part, the retrogression inquiry looks to the plan’s effect on minority voting strength by considering the number of potential African American voters in the existing and proposed districts. The 2000 census data — the very data that gives rise to the State’s obligation to conduct reapportionment — presents a unique problem, and one which the court must address as a preliminary matter. For the first time, the census permitted respondents to aver that they were of more than one race. Thus, in addition to responses that list only “black,” the census data includes data on individuals who identified as “black” and other races. Ironically, the opportunity to more specifically identify one’s race has given rise to a controversy regarding how best to identify a “black” voter. The United States relies on the Department of Justice Guidance issued in January, 2001, notifying covered jurisdictions that the United States would consider only black and black/white responses to be “black voters” for purposes of preclearance. Yet, Georgia’s redistricting office calculated “black” population as including those individuals who responded to the census by identifying themselves as black or black and any other race. While Georgia’s political science expert suggests that the difference is inconsequential, the two methods of counting black voters lead to divergent conclusions about the number of majority-minority districts in the proposed plans.

In fight of the dispute regarding the census data, the United States urges the court to consider black registration data as further indication of the African American voting population of the proposed Senate Districts. This court is thus faced with *79three numbers — two calculations of BVAP and registration data — all of which present different pictures of the proposed plan’s effect on minority voters’ exercise of their electoral franchise.

Courts have consistently relied on percentages of BVAP to consider whether minority voters’ ability to exercise their franchise has been affected by a voting procedure change. For example, in United Jewish Organizations, the plurality opinion expressed a preference for voting age population statistics. 430 U.S. at 164 n. 23, 97 S.Ct. 996 (opinion of White, J.). Current voting age population data is probative because it indicates the electoral potential of the minority community. However, as illustrated by the United States’ cross-examination of Epstein, voting age population may not reflect the population of eligible voters, as it may include college and institutional populations.

On the other hand, the Court has cautioned against reliance on voter registration data, which may reflect effects of prior discrimination or levels of apathy in the community. See City of Rome, 446 U.S. 156, 186, n. 22, 100 S.Ct. 1548, 64 L.Ed.2d 119 (1980) (noting that courts had found that black population figures, when “coupled with” BVAP data, “provides more probative evidence ... than does voter registration data, which may perpetuate the effects of prior discrimination in the registration of voters, ... or reflect a belief among the Negro population that it cannot elect a candidate of its choice-”). Epstein also testified that BVAP is a better indicator of minority voting strength than registration numbers. Tr., 2/5/02, p.m. at 49 (“Registration is very volatile. If you have a good voter registration drive, if you have different types of candidates running for office, any number of individual factors can influence black registration as well as, of course, the normal population movements in and out of the district.”); ef. U.S.Ex. 503 ¶ 6; U.S.Ex. 716, 11:25-13:11; 21:9-22:3; 25:19-26:5; U.S.Ex. 719, 66:3-21 (Senate District 2 residents testified that, in their experience, black voter turnout is lower than the percentage of black registered voters).

The United States urges this court to refrain from choosing one measurement, and rather to evaluate the entire picture, taking into account the different measurements of eligible African American voting population. The court agrees that such an approach is both prudent and in keeping with controlling precedent. In Johnson v. De Grandy, the Supreme Court rejected the idea that one “type” of number is necessarily more probative than all other numbers: “The legal standard is not total population, voting age population, voting age citizen population or registration, but the ability to elect. The Supreme Court repeatedly has declined to elevate any of these factual measures to a ‘magic parameter.’ ” 512 U.S. 997, n. 14, 114 S.Ct. 2647, 129 L.Ed.2d 775 (1994). In light of Supreme Court precedent, we find that BVAP may be a more appropriate number to consider in determining whether a district is properly characterized as a “majority-minority” district. However, in its review of the proposed plan’s predicted effect on black voting strength, the court will consider all the record information, including total black population, black registration numbers and both BVAP numbers.

Such an approach also allows the court to avoid the troublesome and inevitably controversial question of who constitutes a “black” voter for purposes of the Voting Rights Act. The court is reluctant to enter the fray of the parties’ dispute concerning the proper counting of mixed-race responses to the census. On one hand, we note *80that the Attorney General published a Guidance that gave advance notice of how it would calculate BVAP for the purposes of the Voting Rights Act. On the other hand, Georgia’s decision to include all individuals who identified themselves as black, whether as their only race or in combination with any other race, is not inherently unreasonable. The State’s decision is further supported by Dr. Harrison’s expert report, which was unchallenged by the defendants. Pl.Ex. 26. The court does not find that the United States’ or Georgia’s calculation of BVAP is more or less probative. Rather, the court will consider both numbers in considering the overall effect of the proposed redistricting plans.

1. State Senate Plan

In evaluating the evidence of the Senate ' redistricting plan’s purpose and effect, the court considers a wide range of factors that contribute to minority voting strength. First, the court reviews the evidence presented by plaintiffs expert, and, in particular, his statistical analysis predicting a correlation between BVAP and the opportunity of African American voters to elect candidates of choice. The court also considers the State’s argument that the reapportionment plans were necessitated by the underpopulation of the districts in question and the State’s obligation to abide by the constitutional principle of one person, one vote. The court reviews the — albeit sparse — evidence of racially polarized voting and the degree of white crossover voting in Senate Districts 2, 12 and 26. Finally, the court reviews the lay testimony proffered by the parties on the question of the proposed plans’ effect on minority voters’ ability to exercise their electoral franchise.

a. Probit Analysis and Decreases in Senate Districts’ BVAP

The State of Georgia relies on Epstein’s report, together with the statistical data, lay testimony and reaggregation analysis performed by Engstrom, to meet its burden of demonstrating that the proposed Senate redistricting plan would not have a retrogressive effect. Two significant and novel issues are presented by Georgia’s rebanee on Epstein’s report. First, we must determine what, if any, relevance Epstein’s results have for our analysis of whether the proposed plans are retrogressive. In other words, if the court is persuaded that a district with 44.3% BVAP has a 50% chance of electing minority voters’ candidate of choice, in what way is the calculation of this point of “equal opportunity” relevant to retrogression? A second, related, question is whether a decrease in the probability of electing a candidate of choice, without more, constitutes a retrogressive effect within the meaning of Beer.

Epstein’s report demonstrates that there is a correlation between a district’s BVAP and the bkebhood that a candidate of choice wib be elected in that district. None of the parties have disputed the existence of this correlation. It is the significance of this correlation that is disputed. Georgia argues that identification of the point at which African American and non-African American voters have “equal” chances of electing preferred candidates wib permit the court to determine if African American voters will retain their voting strength under the proposed plans. Georgia urges this court to take Epstein’s magic number of a “point of equal opportunity” at 44.3% BVAP and apply it to each of the plans — and, indeed, each of the electoral districts — currently before the court. This approach is untenable.

The court rejects the notion that the “point of equal opportunity” is in any way dispositive of the Section 5 inquiry. The Supreme Court has repeatedly held that, *81while a Section 2 suit compares the change in voting procedures to an ideal, fair benchmark, Section 5 actions must compare the proposed plan to the existing opportunities to elect candidates of choice. Hall, 512 U.S. at 883-84, 114 S.Ct. 2581. Thus, as already discussed, our analysis must focus, not on the level of BVAP that will ensure a “fair” or “equal” opportunity to elect preferred candidates, but on whether the proposed changes would decrease minority voters’ opportunities to elect candidates of choice.

Epstein analyzes retrogression in two ways. First, he compares the number of majority-minority districts under the existing and proposed senate plans. He presents no expert opinion as to the statistical significance of this comparison, and relies on no underlying analysis for this comparison. He has merely placed the census data in a tabular form.

The second analysis of retrogression is based on Epstein’s “equal opportunity” point, and simply counts the number of districts with BVAPs higher than 44.3% under the existing arid proposed plans. This comparison is problematic for the same reasons that Epstein’s “equal opportunity” point is uninformative; it tells the court nothing about the relative increase or decreases in minority voting power.

Furthermore, the “equal opportunity” comparison assumes that all seats in the proposed Senate Districts should be analyzed as open seats. Yet, Epstein’s report concludes that the point of equal opportunity was much higher where a white incumbent was in office — 56.5% BVAP. Senator Michael Meyer von Bremen, a white man, currently occupies the seat for Senate District 12. Over 72% of proposed Senate District 12 would be comprised of the benchmark district. U.S.Ex. 112. Given Epstein’s own explanation of his decision to “code” Senator McKinney as running for an open seat because he was “following the rule that open seat as having less than 50 percent of [one’s] former constituents,” see Tr., 2/4/02, p.m. at 62, the court does not understand how Epstein could fail to consider Senator Meyer von Bremen to be a white incumbent. Thus, Epstein’s “retrogression” analysis tells the court little or nothing about actual minority voting strength in the “counted” districts. By simplifying the nature of the retrogression inquiry as he has, Epstein has rendered his analysis all but irrelevant.

We do not suggest that a probit analysis may never be a valuable tool for examining retrogression, but merely that the one offered by the State in this case is entirely inadequate to that task. Presumably, pro-bit analysis could evaluate the increases and deceases in BVAP in each proposed district and assess the statewide increase or decrease in the probability that minority preferred candidates will be elected under the proposed plan. Here, however, Epstein made no attempt to address the central issue before the court: whether the State’s proposal is retrogressive. He failed even to identify the decreases in BVAP that would occur under the proposed plan, and certainly did not identify corresponding reductions in the electability of African American candidates of choice. The paucity of information in Epstein’s report thus leaves us unable to use his analysis to assess the expected change in African American voting strength statewide that will be brought by the proposed Senate plan.

While questioning the relevance of Epstein’s testimony regarding the “equal opportunity point,” and noting the severe shortcomings of the probit analysis identified by the defendants, the court nevertheless takes note of Epstein’s uncontradicted predicted correlations between BVAP and *82the likelihood of electing a candidate of choice. These correlations are plotted on his “S curve,” and were discussed in his cross-examination and redirect testimony. As reflected in the table below, according to Georgia’s calculations of BVAP, only one of the existing majority-minority districts would not experience a decrease in BVAP under the proposed plan. Furthermore, District 44, which currently has a majority black population and majority of black registered voters, but has 49.21% BVAP, would also experience a substantial decrease in its BVAP. The BVAP of District 34, the plan’s proposed new majority-minority seat, would increase. Of the remaining twelve districts in which the BVAP decreases, reductions of over 10% are present in nine districts. The following table shows the effect of the proposed Senate plan on the BVAPs of benchmark and proposed majority-minority districts.

[[Image here]]

[[Image here]]

U.S.Ex. 118; Pl.Ex. 25, app. III.

Epstein testified that a drop in a district’s BVAP would result in a diminished likelihood of success for African American preferred candidates. According to Epstein’s “S curve” and his testimony, the impact of a decrease in percentages of BVAP will vary depending on where it occurs in the curve, or, in other words, in relation to the actual level of a district’s BVAP in a district. Thus, a 5.7% decrease in a district’s BVAP from 50% to 44.3% would, according to Epstein’s analysis, result in a 25% decline in the likelihood that a candidate of choice would be elected. However, a similar decline in BVAP would not have the same effect if the overall BVAPs were higher. Thus, in Senate District 10, where the proposed plan would decrease BVAP by 6.52%, reducing BVAP from 69.72% to 63.42%, Epstein’s “S curve” indicates that this would reduce only slightly the probability of electing an African American preferred candidate. Unfortunately, it is impossible to read Epstein’s curve in any accurate fashion, and his expert report provides no additional guidance or information.39 Nevertheless, the court does find Epstein’s testimony relevant insofar as it suggests that decreases in BVAP within the ranges proposed in the contested Senate districts may have a significant (if inadequately quantified) negative impact on the likelihood that African American voters will be able to elect their candidates of choice.

b. Underpopulation of Existing Senate Districts

The benchmark plan for Section 5 purposes is the last legally enforceable plan, *83which in this case is Georgia’s existing Senate plan. Abrams v. Johnson, 521 U.S. 74, 97, 117 S.Ct. 1925, 1939, 138 L.Ed.2d 285 (1997). Georgia argues that the demographics of the existing plan do not represent a fair basis for comparison because, in light of the 2000 census, the districts at issue are all underpopulated, and thus in violation of the principles of one person-one vote. See Wesberry v. Sanders, 376 U.S. 1, 84 S.Ct. 526, 11 L.Ed.2d 481 (1964) (holding unconstitutional gross disparities of population across Georgia congressional districts and the resulting dilution of voting rights of residents of the more populous districts).

The State contends that, where population had to be added to districts, it was inevitable that their BVAP levels would decrease. See, e.g., PPFF at ¶ 621 (“While the BVAP levels in the majority-minority district have decreased, that was inevitable as a result of the fact that those districts where generally short of population. This is particularly true with regard to the three districts with which Defendants have taken issue.”). Georgia suggests that this alleged quandary insulates its plan from a section 5 challenge. See Pl.Post-Tr.Br. at 47-48. This argument is unavailing. First, even if the State is correct that some reductions in BVAP were inevitable — a proposition that it asserts without attempting to prove — it certainly does not follow that Georgia was compelled to move minorities out of Districts 2, 12, and 26 to the extent that it did. Indeed, the State actually removed some majority African American precincts from each of these districts, a decision that at least casts doubt on its cries of inevitability.40 While the precincts added to the proposed districts also contain significant African American communities, the overall effect of the reapportionment is to reduce the percentages of BVAP in these districts.

As the United States has illustrated, there were alternative plans available that would both have comported with the constitutional principle of one person-one vote and have allowed the retention of greater numbers of African American voters in the disputed districts. See U.S.Exs. 213-23;

SD Deviation Black reg Black Net % of Reg. reg., % of add. of Reg. voters removed votes added . reg. moved out_precincts added_precincts voters Black reg. voters, % of net add. reg. voter

24.37% 7,835/8,626 (approx. 42% of total added voters) 20.39% 21.83%* 23,225 24,465 * 14.97% 15.89% * 18,866 17,533 * 12.86% 11.8%*

12 - 17.77% 10,705 (approx. 91% of total added voters) 46.64% 17,228 30.28% 9,928 18.25%

26 - 28.65% (approx. 54% of 14,047 (approx. 54% of total added voters) 41.80% 41.80% 33,202 30,484 * 20.20% 23,295 19,678 * 11.0% 13.09%* *84313-18; 414-19. For example, in Senate District 2, the United States notes that Hunter Army Air Field was removed from the proposed district. The inclusion of the army base would have permitted the district, which was in need of population, to add population while avoiding a reduction in BVAP. See U.S.Ex. 733, 192: 1-14 (Meggers’ testimony); USPFF ¶¶ 214-15.

Moreover, the State, despite its suggestion to the contrary, was not forced to choose between complying with the Equal Protections Clause and the Voting Rights Act. It is true that the dictates of one person-one vote may at times force a redistricting state to reduce the minority population in a district that previously was strongly majority-minority. However, the mere fact that BVAP decreases in certain districts is not enough to deny preclearance to a plan under Section 5. Instead, retrogression concerns are implicated when it appears that the numerical changes may dimmish effective minority voting power. A State is free under Section 5 to reduce BVAP levels in a district in order to bring that district into compliance with the Fourteenth Amendment, so long as in so doing it does not limit the ability of the remaining minority voters to elect candidates of choice. Cf. Bush, 517 U.S. at 982-83, 116 S.Ct. at 1963 (suggesting exception when no feasible alternative plan exists). Read in this way, the Voting Rights Act continues to serve its historical purpose of protecting minorities in the exercise of their electoral franchise without running afoul of the Constitution’s overarching demand of equality.

c. Evidence of Racial Voting Patterns

Racially polarized voting “exists where there is a consistent relationship between the race of the voter and the way in which the voter votes.” Gingles, 478 U.S. at 54 n. 21, 106 S.Ct. 2752. Recognizing that “there is no simple doctrinal test for the existence of legally significant racial bloc voting,” Gingles noted that Section 2 inquiries are necessarily fact-sensitive. Id. at 58, 106 S.Ct. 2752. Justice Brennen suggested a number of factors that courts may consider under Section 2.

[A] white bloc vote that normally will defeat the combined strength of minority support plus white “cross-over” votes rises to the level of legally significant white bloc voting. The amount of white bloc voting that can generally “minimize or cancel” black voters’ ability to elect representatives of their choice, however, will vary from district to district according to a number of factors, including the nature of the allegedly dilutive electoral mechanism; the presence or absence of other potentially dilutive electoral device ...; the percentage of registered voters in the district who are members of the minority group; [and] the size of the distriet[.]

478 U.S. at 56, 106 S.Ct. 2752. As explained above, in the present case, racial polarization is critically important because its presence or absence in the Senate Districts challenged by the United States goes a long way to determining whether or not the decreases in BVAP and African American voter registration in those districts are likely to produce retrogressive effects.

Georgia’s own expert demonstrated that, given some varying amounts of racial polarization and crossover voting, minority voters’ ability to elect candidates of choice is likely to decrease as districts’ BVAP diminishes. Epstein’s probit analysis necessarily subsumes information about racial voting patterns and voter turnout.41 *85Whether an African American candidate of choice wins election is necessarily determined by the degree of racial bloc voting by African American voters and the level of white crossover votes received by the candidate. However, it is impossible to extrapolate these voting patterns from Epstein’s database. As Epstein admitted on cross-examination:

the whole point of my analysis is not to look at polarization per se. The question is not whether or not blacks and whites in general vote for different candidates.

Tr., 2/4/02, p.m. at 44-45.

Although not drawing any legal conclusions, Epstein testified that, in order to assess whether there has been retrogression in the voting rights of people who live in a reconfigured district, it would be necessary to consider whether the district was “significantly different” than the rest of the State. Tr., 2/5/02, a.m. at 70. Epstein also admitted that, in his opinion, a candidate running for office in a particular district would want to take into account whether white crossover was at 25% or at 52%. Id. at 72. It is inevitable that there will be “local variability” in elections. In fact, Epstein notes that his “final conclusion ... is not that there is no difference ... from one area to another.” Id.

Epstein did not perform any ecological regression analyses of voting patterns in Senate Districts 2 or 26. Epstein did, however, perform an ecological regression analysis for an election in Senate District 12. In the 1998 Senate election, he found that the winning white candidate was not the African American candidate of choice, and, in fact, received very little black support. Tr., 2/4/02, p.m. at 50. He testified that, to his recollection, the ecological regression analysis had shown “very high levels of polarization” in that election. Id.

Thus, with the exception of Epstein’s analysis of one election in Senate District 12, the only evidence of racially polarized voting before the court is Engstrom’s expert report. The report provides the results of two different methodologies of analyzing racial voting patterns in Senate Districts 2, 12, and 26. Engstrom considered local elections to assess the “extent of racial voting correlations” in the existing district. Tr., 2/5/02, p.m. at 12. He also performed a reaggregation analysis, which gathered precinct-level results from seven statewide elections between an African American candidate and a white candidate and predicted results for the proposed Senate Districts. The results of Eng-strom’s analysis are discussed at length in the court’s Findings of Fact.

The State would use Engstrom’s reaggregation analysis to prove its case. The State argues that the reaggregation analysis is the only record evidence that attempts to estimate voting patterns in the proposed districts. By Engstrom’s own admission, the reaggregation results suggest the presence of sufficient white crossover voting to permit some African American candidates of choice to prevail in the redrawn districts. Furthermore, Eng-strom can only guess at the reasons for the different voting patterns in statewide and local elections. However, we find that Engstrom has presented compelling evidence that racial voting patterns in State Senate races can be expected to differ *86from racial voting patterns in statewide races.

Before discussing Engstrom’s analysis of voting patterns in local elections, we must first address some concerns with the relevance of this analysis. Plaintiff underscores the absence of any information in Engstrom’s report that would identify the overlap of the jurisdictions in which the local elections analyzed by Engstrom took place and the proposed Senate Districts. Pl.PosL-Tr.Br. at 15; see also Tr., 2/5/02, p.m. at 20. Although the court is ultimately unable to determine the exact extent of overlap present, it is clear that a substantial portion of the jurisdictions considered by Engstrom will fall within the proposed Senate Districts.

Engstrom did not attempt to compare the proposed plan with the benchmark plan, Tr., 2/5/02, p.m. at 70, and, in fact, looked only at racial polarization in existing Senate Districts. In its Post-Trial Brief, the United States, without citing to any record evidence, states that “the county elections analyzed involve counties that make up the overwhelming majority of all three districts at issue.” U.S. Post-Trial Br. at 96. The court has searched the record and found evidence describing the overlap of the existing and proposed Senate Districts,42 as well as overlap of the existing and proposed Senate Districts.43 However, Engstrom looked at city elections for Savannah, Macon and Albany, and there is no evidence in the record that would assist the court in determining these cities’ limits. Engstrom also considered two State House races, but, again, no record evidence permits the court to compare the geographical reach of the House Districts to the proposed Senate Districts.

These gaps in the record do not require the court to disregard the local election data and analysis provided by Engstrom. It is clear that a substantial portion of the benchmark Senate Districts and Bibb County comprise the proposed districts. Furthermore, the State has not denied the United States’ assertion that the county and city elections analyzed involve a majority of the voters in the benchmark and proposed districts at issue. Ultimately, the burden of proof in this matter lies with the State. We look to the State to explain why retrogression is not present, and to prove the absence of racially polarized voting that might diminish African American voting strength in light of several districts’ decreased BVAPs. We find that Eng-strom’s report presents relevant information, and indicates that Senate elections in the redrawn districts will be marked by high levels of polarized voting. Plaintiff has presented no evidence to suggest otherwise.44

The State’s primary critique of Eng-strom’s report posits that his analysis of racial polarization is flawed because it only considers the level of white crossover voting.45 The State argues that a better indi-*87eator of racial polarization is the difference of votes received from African American and non-African American voters. See Pl.’s Resp. to USPFF ¶ 426 (arguing that, “in looking at the actual difference of issue, it is apparent that the patterns in the statewide elections was [sic] indistinguishable from that in the local elections”). For example, instead of focusing on Eng-strom’s estimate that Senator Thomas received 78.8% of African American votes in District 2, the State suggests that the difference between that number and the level of white crossover voting for Thomas (8.9%), is a more accurate reflection of race-based voting patterns.

Georgia relies on its analysis of “preference differentials” to suggest that the regression analyses done by Engstrom on the local election data demonstrate the same pattern as does his reaggregation analysis of statewide elections. Georgia identifies the range of the differences in support in Engstrom’s local election data, and posits that this range corresponds to the range seen, in the reaggregation results. Georgia asks the court to conclude that the similarity of “ranges” demonstrates that the level of racial polarization is the same in local elections as it is in elections for statewide office. The next step in this logic, according to the State, is to credit Table 4 over Tables 1-3, and to conclude that African American candidates have “good chance[s]” of winning election. Tr., 2/5/02, p.m. at 89 (Engstrom admits that reaggregation results show a “good chance” of African American candidate winning).

It is fanciful to think that the court will defer to counsel’s alternative “expert” theories. No expert testimony has been presented to suggest that comparing “ranges” of numbers will result in probative evidence. For example, the court does not see why a comparison of the ranges of the relative support levels is necessarily any more probative than a comparison of the average levels of the differences in support.46 The court is not persuaded that *88the method of comparing the ranges of the differentials suggested by plaintiffs counsel is at all probative. While open to the contention that the difference in levels of white and African American voter support may be probative of the degree of racial polarization, the court sees no way that it may competently interpret this information in the absence of expert evidence to this effect.

In light of the problems with the State’s own statistical evidence and its inability to cast significant doubt on that presented by the United States, we are compelled to conclude that the evidence of racial polarization suggests the likelihood of retrogression. Indeed, despite the importance of such information to the Section 5 inquiry, plaintiff has provided the court with no competent, comprehensive information regarding white crossover voting or levels of polarization in individual districts across the State. At the same time, the United States has produced credible evidence that suggests the existence of highly racially polarized voting in the proposed districts. As emphasized above, an African American candidate’s ability to succeed in a jurisdiction will depend on the levels of polarization and white crossover voting. Eng-strom’s testimony suggests that Senate races in the proposed districts will be marked by racially polarized voting. This evidence undermines the utility of Epstein’s probit analysis because that analysis fails to account for variations in levels of racial polarization. And plaintiff has presented no other evidence to persuade us that voting in future Senate races in the contested districts will not be racially polarized. Consequently, the court can not find by a preponderance of the evidence that the planned reductions in BVAP and in the number of African American registered voters in Senate Districts 2, 12 and 26 will not diminish African American voting strength in these .districts. Once again, we note that it may well be the case that any decrease in African American electoral power in Senate Districts 2, 12 and 26 will be offset by gains in other districts, but plaintiff has failed to present any such evidence.

d. Lay Testimony

The electoral history of the jurisdiction may shed considerable light on the effects that reapportionment will have on minority voting strength. One part of that electoral history is found in racial voting patterns and the willingness of voters to cast ballots for candidates of different races. Lay witness testimony regarding the use of race in politics and elections is also relevant.

The court is presented with two types of lay testimony regarding the electoral climate of the disputed districts and the State in general: (1) testimony by legislators and others involved in the drafting of the reapportionment plan; and (2) testimony by legislators and citizens providing opinions about minority voting strength. African American Legislators’ Support of Redistricting Plan

The State relies heavily on the unanimity of African American legislators’ support for the reapportionment plan to argue that *89the plan does not have a retrogressive effect. Plaintiff characterizes as “indisputable” the proposition that African American elected representatives are in the best position to judge “as a matter of fact” whether the reapportionment plans enhance or diminish minority voting strength. PPFF ¶ 618. The court, however, notes that the United States has presented extensive evidence of African American Senators’ misgivings about the Senate plan. Nevertheless, only two African American legislators voted against the plan. The court will not look behind those votes to question such inherently political decisions. A vote for legislation is almost always a compromise of some sort, motivated by a complex intersection of self-interest and external pressures. A court that tries to unpack these forces, and assign probative weight to them, treads a treacherous path. Accordingly, we are loathe to rely on testimony regarding the nature of legislative trade-offs, or on post-hoc expressions of doubt on the part of legislators who nevertheless voted for the contested plan. Certainly, as it relates to the plan’s possible retrogressive effect, this is dubious evidence indeed.

That said, it is undoubtedly true that support among minority legislators for a reapportionment plan can sometimes be relevant in assessing the legality of that plan under the Voting Rights Act:

The protection of existing relationships among incumbents and their constituents, and the benefits accruing to the state from the seniority its delegation may have achieved in Congress, are pragmatic considerations which often figure prominently in the drawing of congressional districts. These considerations are not talismanic, however, and may not serve to protect incumbents by imposing an electoral scheme which splinters a geographically concentrated black populace within a racially polarized parish, thus minimizing the black citizenry’s electoral participation.

Major v. Treen, 574 F.Supp. 325, 355 (E.D.La.1983). In this case, the State has presented no authority suggesting why the court should consider the support of African American legislators as evidence that the actual effect of the Senate redistricting plan will not be to decrease minority voters’ opportunities to elect candidates of choice. We believe that the legislators’ support is, in the end, far more probative of a lack of retrogressive purpose than of an absence of retrogressive effect.

Voting Patterns in Senate Districts 2, 12 and 26

The record is replete with lay testimony regarding individuals’ predictions as to whether the redrawn districts will permit minority voters to elect candidates of choice. Several incumbent Senators testified that they believed the plan gave them and others a “fair” or “reasonable” chance of victory in the redrawn districts. See, e.g., Brown dep. at 30, Walker dep. at 12. However, Senator Fort expressed concerns that the plans might be retrogressive. Int.Ex. 16 at 3.

The United States has offered the testimony of eleven witnesses from Chatham County with respect to the proposed Senate District 2, including two State Senators, four African American Commissioners, and three members of the Executive Committee of the Savannah Branch of the NAACP. They testified that they believe the boundaries of Proposed Senate District 2 will reduce the opportunity that African American voters have, under the benchmark Senate District 2, to elect candidates of choice. USPFF at ¶ 187.

*90Dr. Jackson, President of the Savannah Branch of the NAACP, sent a letter to the Governor of Georgia on August 6, 2001, expressing his opposition to the proposed changes, and suggesting that they “would serve only to dilute the Black votes and eventually delete the Black Legislators in this area.” U.S.Ex. 504, ¶23; U.S.Ex. 504, ¶¶ 10, 24 (NAACP executive member testifying that proposed districts would weaken minority voting strength); U.S.Ex. 509, ¶ 10 (same). The city aldermen expressed similar concerns that the opportunities for African American voters to elect candidates of choice will dimmish. U.S.Ex. 501, ¶ 5; U.S.Ex. 505, ¶5; U.S.Ex. 502, ¶ 16.

Two witnesses from Senate District 2 testified that African American voter registration rate is lower than the white voter registration rate in Chatham Count, leading to a majority of white registered voters, despite a majority African American population. U.S.Ex. 508 ¶ 6; U.S.Ex. 716, 11:25-13:11; 21:9-22:3; 25:19-26:5 (Jackson, NAACP); U.S.Ex. 719, 66:3-21 (Rivers, Chatham County Bd.Comm.). Furthermore, seven of the witnesses, including NAACP Executive members, County Commissioners, Aldermen and Senator Thomas, testified that white voter turnout is generally higher than African American voter turnout in Chatham County.47

Ms. Thomas notes that, in her experience, her white opponents have always won majority white precincts. U.S.Ex. 704, 104:11-19 (Thomas). Other witnesses from Chatham County testified to high levels of polarized voting, i.e. that voters cast their ballots for candidates of the same race as themselves. One NAACP member, who worked at a television station, testified that the staff routinely relied upon polarized voting patterns to predict election results from precincts that had not reported their returns. U.S.Ex. 504, ¶ 6 (Johnson). Another witness testified to the use of racial appeals to white voters in the 1995 Savannah mayoral election and the 1999 Savannah municipal election. USPFF ¶¶ 210, 211; U.S.Ex. 712 (Aider-man Jackson) (stating that white opponent labeled Alderman Jackson a “Farrakan supporter”).

With respect to Senate District 12, the United States offers the testimony of five African American declarants, one former and two current City Commissioners, a former Representative to the State House, and the former president of Doughtery County NAACP. Two of the witnesses, Mr. John White and Mr. Charles Sherrod, ran for the Senate District 12 seat.

City Commissioners Williams and White testified that they felt the flier distributed by White’s white opponent, Mark Taylor, in the course of the 1996 Democratic Party primary, was an attempt to encourage white voters to vote against African American candidates. U.S.Ex. 513 ¶ 12. The flier compared White to Williams, who was outspoken on issues of city contracts with black-owned businesses, affirmative action in city government and affordable housing, and urged people not to vote in the Republican Party primary, but to vote against White in the Democratic Party primary. Id. White characterized the flier as “a call to arms for white voters to rally together to defeat a black candidate.” U.S.Ex. 513 ¶ 12.

*91The United States’ lay witness evidence from Senate District 26 does not paint a picture of such a racially charged political environment, as is suggested by the lay testimony about Senate Districts 2 and 12. Senator Robert Brown is the incumbent and has been reelected without opposition three times, and once without opposition in the general election. Most witnesses believed that incumbent Senator Brown would be re-elected under the proposed plans; however, they were especially fearful that it will be difficult to elect a candidate of choice in the proposed Senate District 26 if Senator Brown did not decide to run for re-election. U.S.Ex. 516 ¶ 7 (Barnes, Bibb County Bd. of Ed.); U.S.Ex. 519 ¶ 8; U.S.Ex. 725, 36:15-37:3 (Hart, County Comm.). The witnesses from Senate District 26 expressed concerns about the reductions in the district’s minority population and testified that people tend to vote along racial lines. See, e.g., U.S.Ex. 515 ¶ 6; U.S.Ex. 723, 53:11-16 (Abrams, Bibb County Bd. of Ed.); U.S.Ex. 726, 28:11-29:10, 82:22-85:5 (Bivens, County Comm.). However, these witnesses did not describe the type of racial appeals or polarized voting evident in Senate District 2 and 12.

The State argues that admissions by the United States’ declarants that it would be possible to elect candidates of choice in the proposed districts are evidence that no retrogression will occur. For example, Alderman Clifton Jones states that, “although more difficult under the new plan for Senate District 2, blacks still have a fair opportunity to elect the candidate of their choice.” C. Jones dep. at 63. However, the State’s reliance on these statements is as misplaced as is its reliance on the “equal opportunity point.” The retrogression analysis does not ask if a proposed district is “fair” or whether a possibility exists that a minority preferred candidate would be elected. Indeed, Jones’ statement that it would be “more difficult” to elect a candidate of choice points to a retrogression in minority voting strength.

Legislative Influence

A final note is required due to an argument advanced by Georgia that rests on the expectation that the proposed districts will elect Democratic candidates. Georgia asserts that African American voters in Georgia tend to vote for Democratic Party candidates in excess of 90% of the time. While this assertion appears to be unsupported by empirical evidence, Epstein provides this number in a footnote, Pl.Ex. 25 at 17 n. 14, and several of plaintiffs lay witnesses support the proposition that African Americans overwhelmingly vote for Democratic candidates. However, it does not follow that anything that is good for the Democratic Party is good for African American voters — at least, within the context of this court’s Section 5 inquiry.

Georgia asks the court to equate African American voters’ electoral strength with the success of the Democratic Party at the polls. “The evidence in this case is absolutely uncontradicted that minority voting strength is enhanced by the Democratic strategy, joined in by both black and white Democrats, to maintain Democratic majority in the Senate.” PPFF at ¶ 618. The reapportionment plan has been crafted predominantly to ensure continued Democratic ' control of the State Senate, and plaintiff explains that the “political performance” statistics for African Americans influenced the redrawing of the districts.

However, Engstrom’s analysis of racial voting patterns demonstrates that white and African American voters regularly prefer different Democratic candidates at Democratic Party primaries. U.S.Ex. 601, Tables 1-3. Furthermore, the record demonstrates that many of the proposed *92senate districts were drawn to protect current white Democratic incumbents. See U.S.Ex. 703, 49:7-16, 80:4-81:20 (Senator Brown testified that he considered white Democrats to be the weakest link in the redistricting process, and that he excluded majority African American precincts from his district in order to assist neighboring white Democratic incumbents).

Plaintiff urges the court to consider “legislative influence” as a factor in assessing the effectiveness of minority voting strength. All of the African American legislators currently serving in the Georgia General Assembly are Democrats. Senator Brown testified that, were Republicans to be the majority in the Senate, Democratic African Americans would lose several positions as committee chairs. See, e.g., Pl.Ex. 20 at 23-24. Congressman John Lewis testified to similar conclusions:

I happen to believe that it is in the best interest of African American voters, and as far as I’m concerned, to all voters and all citizens, to have a continued Democratic-controlled legislature in Georgia. First of all, the great majority, and I think all of the black legislators presently elected, and most of the would-be black candidates, are Democrats or will be Democrats in the General Assembly.
So, if black elected officials, black state legislators, are going to have any sense of control, any sense of power, as chairs of full committees or chairs of subcommittees, it will come with being in the majority. And when in the minority, they lose all of that. They lose the possibility of maintaining a chairmanship of a full committee or subcommittee.
The great majority of the African American voters in the State of Georgia, 90 percent or more tend to vote the Democratic way. So, it’s in our best interest for us to maintain a Democratic-con-
trolled state legislature. I think that’s important. I think this is a must.

Pl.Ex. 21, at 18-21 (Congressman Lewis Test.).

The State has pointed to no authority that supports the proposition that a political party’s overall success, and accompanying positions of power for minority legislators, should be considered in assessing minority voters’ effective exercise of their franchise. Indeed, Whitcomb v. Chavis suggests otherwise. There, the Court specifically distinguished between “racial vote dilution” and “political defeat at the polls.” 403 U.S. 124, 153, 91 S.Ct. 1858, 29 L.Ed.2d 363 (1971). The Court concluded that the evidence demonstrated a lack of racial bias in the voting community. Id. Nothing in the record suggested that African American citizens were prevented from registering to vote, were excluded from political parties, or were overlooked in the candidate slating process. Id. at 149-50, 91 S.Ct. 1858. Had the Democrats won more elections, the Court concluded, the African American community would have had no complaints about its representation. Id. Because most of the elections under consideration were won by Republicans, however, “the failure of the [black community] 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’ ..., but this seems more a euphemism for political defeat at the polls.” Id. at 153, 91 S.Ct. 1858.

In League of United Latin American Citizens v. Clements, the Fifth Circuit considered a claim that a Section 2 violation was shown where plaintiffs argued that highly partisan voting was indicative of racially polarized voting because the majority of Republicans were white. 999 F.2d 831 (5th Cir.1993). While concluding *93that political parties did not serve as proxies for race, id. at 860, the court agreed that “courts should not summarily dismiss vote dilution claims in cases where racially divergent voting patterns correspond with partisan affiliation as ‘political defeats’ not cognizable under § 2.” Id. at 860-61.

We believe that the converse is also true. Where a change in voting procedures will favor a political party supported by African Americans, courts need not conclude that African American voters’ strength corresponds to that success. Rather, as did the Fifth Circuit, courts have an obligation to base their inquiry “upon a searching practical evaluation of the ‘past and present reality.’ ” Id. at 860. The inquiry in this Section 5 matter requires us to scrutinize the opportunities for minority voters in the redrawn districts to exercise their electoral franchise. The court emphatically rejects the notion that a plan that protects Democratic incumbents and a Democratic majority is necessarily a plan that does not retrogress with respect to African American voting strength. Whatever success the Georgia Democratic Party may enjoy as a result of the Senate redistricting plan does not and cannot immunize the plan’s racially retrogressive effects from Section 5 attack. The Voting Rights Act was not enacted to safeguard the electoral fortunes of any particular political party.

e. Conclusion

In asking this court to enter a declaratory judgment that its Senate reapportionment plan does not have the effect of worsening minority voters’ opportunities to effectively exercise their voting rights, the State must present competent evidence that will permit the court to engage in the analysis mandated by the Voting Rights Act.

We have engaged in a searching review of the record for evidence that would facilitate a competent comparison of the benchmark Senate plan and the proposed plan, and their consequences for the voting strength of Georgia’s African American population. We conclude that the State has not met its burden of proof. The expert testimony presented by the plaintiff was woefully inadequate. Epstein presented a report that was crafted to predict a “point of equal opportunity” that has little relevance to the retrogression inquiry mandated by Section 5. Plaintiffs retrogression analysis consisted of a simple comparison of the number of majority-minority districts, and the number of districts with “equal” chances of electing an African American preferred candidate. Defendants’ expert was only marginally more helpful. Engstrom provided a report that presented no conclusions with respect to the existence of retrogression in the Senate plan. Finally, intervenors’ expert offered no substantive evidence at all.

In short, the evidence presented by the parties, while voluminous, may be relied upon only for several limited conclusions. The number of Senate Districts with majorities of BVAP would, according to Georgia’s calculations, increase from twelve to thirteen; according to the Attorney General’s interpretation of the census data, the number would decrease from twelve to eleven. According to Georgia, the proposed Senate Districts 2, 12 and 26 would have BVAPs of 50.81%, 50.22% and 50.39% respectively; the Attorney General’s calculations of BVAPs are slightly less, with Senate District 2 falling below 50% to 49.81% BVAP. Under the proposed Senate plan, the number of Senate Districts with majorities of black registered voters will decrease from 11 to 8 districts.

Plaintiffs expert predicts that, as BVAP in a district decreases, the probability that *94an African American candidate of choice will be elected diminishes. The percentages of BVAP in 11 of 12 of the existing majority-minority districts will decrease by 3.42%-26.28% in the redrawn districts. In Senate Districts 2, 12 and 26, it will decrease by 10.27%, 4.77% and 11.65% respectively.

An analysis of local and regional elections demonstrates the presence of racially polarized voting in the benchmark Senate Districts and in several of the counties and cities included in the benchmark Senate Districts. Substantial portions of these jurisdictions fall within the proposed Senate Districts. Lay witness testimony also suggests the presence of racially polarized voting, especially in Senate Districts 2 and 12. There is, however, conflicting lay testimony regarding the probability that the redrawn Senate Districts will permit African American candidates of choice to win election.

Reaggregated results from statewide elections show that African American candidates may garner sufficient white crossover votes to win election in the proposed Senate Districts 2, 12 and 26. However, the record also supports a finding that African American candidates of choice running for State Senate seats are unlikely to receive the same levels of white crossover voting as may occur in statewide elections.

Finally, the decline in BVAPs in Senate Districts 2, 12 and 26 was not required by the State’s constitutional obligation to draw districts that comply with the principle of one person-one vote. Under the reapportionment plan, predominantly African American precincts are removed from the underpopulated Senate Districts 2, 12 and 26. Although the plan proposed to add African American population to these districts, the total result is a decrease in the districts’ BVAP. The State has offered no evidence to suggest that this decrease was inevitable or necessary.

In fight of these conclusions, the court finds that the State has failed to demonstrate by a preponderance of the evidence that the reapportionment plan for the State Senate will not have a retrogressive effect. The court simply is not persuaded, on the basis of the evidence before it, that minority voting strength will not be significantly diminished by the proposed redistricting. The plan proposes to decrease the BVAPs in existing majority-minority districts such that they would constitute only bare majorities, or slightly less than majorities. It was Georgia’s burden to produce some evidence to prove that these changes would not be retrogressive.

The State has produced no evidence to demonstrate that the demographics of the proposed Senate Districts counteract any reduction in BVAP. It has not attempted to show the number of white voters who cross over to vote for African American candidates of choice in the disputed districts and how that might affect the effective exercise of minority voters’ franchise. Nor has the State presented evidence regarding potential gains in minority voting strength in Senate Districts other than Districts 2, 12 and 26. There are, without doubt, numerous other ways, given the limited evidence of racially polarized voting in State Senate and local elections, that Georgia could have met its burden of proof in this case. Yet, the court is limited to reviewing the evidence presented by the parties, and is compelled to hold that the State has not met its burden. Accordingly, we are unable to conclude that the Senate reapportionment plan will not have a retrogressive effect on the voting strength of Georgia’s African American electorate.

*952. United States Congressional Plan

Although Georgia has three African American representatives in the United States Congress, there is currently only one majority-minority Congressional district, the Fifth Congressional District. The proposed redistricting plan would include two majority-minority Congressional districts. Plaintiff argues that the plan also creates additional districts where African Americans have the opportunity for election, by increasing BVAPs in several districts.

Intervenors assert that the Congressional redistricting plan violates Section 5. First, they argue that the Congressional plan does not create two majority-minority districts. Rather, when African American population is counted in accordance with the Attorney General’s Guidance, only the Fifth District remains a majority-minority district. However, the Fifth District has a majority BVAP. The court is not persuaded that the reduction in the African American population in the Fifth District necessarily constitutes retrogression.

Second, intervenors contend that the proposed Congressional plan retrogresses because of the reduction in the Fifth District’s BVAP, and due to purported racially polarized voting in one 1998 election for Fulton County Commission and one 2001 runoff for Atlanta City Council president. There is, however, no evidence in the record to support these assertions.

Finally, intervenors suggest that the presence of alternative plans proves the existence of retrogression. The fact that some of intervenors’ plans result in higher percentages of BVAP in certain districts (and lower percentages in others) does not establish that Georgia’s proposed Congressional plan retrogresses. Interve-nors’ evidence, at best, shows that the General Assembly could have adopted a more Republican plan. The evidence in the record establishes by a preponderance of the evidence that the Congressional redistricting plan does not result in retrogression in the position of African Americans with respect to the effective exercise of the electoral franchise.

3. State House Plan

According to the 2000 census data, the benchmark State House plan contains 40 districts with a total black population over 50%, 37 districts with total BVAP over 50%, and 38 districts with total black voting registration over 50%.

The proposed plan would have 38 or 39 seats in districts with majorities of BVAP. While some of the existing House districts would experience decreases in BVAP under the proposed plan, there is no evidence before the court of racially polarized voting in any House Districts that might suggest that these decreases will have a retroges-sive effect.

Intervenors argue that the House Plan is infirm because it contains multi-member districts. They suggest that courts have generally disfavored multi-member districts because such districts have the potential to “submerge” substantial minority populations, making it difficult for minorities to select candidates of choice. However, of the multi-member districts in the proposed plan, six are majority-minority districts with BVAPs ranging from 54.14% to 65.18%. The other multi-member districts have relatively low levels of BVAP, ranging from 0.58% to 33.52%. Plaintiff asserts that African American voters will maintain voting strength in the majority-minority multi-member districts, as they will have the opportunity to elect candidates to each of the seats. Intervenors present no evidence or arguments that would convince the court that the multi-*96member districts will diminish minority voting strength.

The proposed State House redistricting plan would create two additional seats in majority-minority districts. The court is persuaded that the House redistricting plan will not have a retrogressive effect on minority voting strength.

D. Purpose of the Reapportionment Plans

The State argues that all three redistricting plans were drawn with the intent to maintain and bolster the Democratic control of the State House and Senate. In order to reach these goals, Democratic legislators sought to “unpack” majority black districts, and to increase Democratic voting strength throughout the State. The testimony of State Democratic legislators reflects a belief that African American votes, which generally heavily favor Democrats, are “wasted” in districts with high percentages of African American voters. See, e.g., Pis.’ Resp. to USPFF ¶ 344 (“Increasing the BVAP in Senate District 12 would waste black voters and diminish their impact in other districts and the reapportionment plan of the State as a whole.”); Pl.Ex. 22 at 21 (Meggers). The plans also reflect an effort to prevent Democratic incumbents from being placed in the same districts.

In Bossier II, the Supreme Court held that “ § 5 does not prohibit preclearance of a redistricting plan enacted with a discriminatory but nonretrogressive purpose.” 528 U.S. 320, 341, 120 S.Ct. 866, 145 L.Ed.2d 845. The Court held that its “longstanding interpretation” of the effect prong as limited to a concern with retrogression applied to the “purpose” prong of Section 5, which must cover “only retrogressive dilution.” Id. at 327, 120 S.Ct. 866. In a Section 5 action, therefore, the government “need only refute the covered jurisdiction’s prima facie showing that a proposed voting change does not have a retrogressive purpose in order for pre-clearance to be denied.” Id. at 332, 120 S.Ct. 866.

Georgia has the burden of making a prima facie showing that the plans do not have a retrogressive purpose. Id. The blatantly partisan nature of the redistricting process and the goals of the authors of the redistricting plans do not violate Section 5. An intent to help the Democratic Party is not a retrogressive purpose, or an intent to decrease minority voting strength. Id.

Here, the State’s plans undoubtedly spread out Georgia’s African American voters. Plaintiff, however, contends, and no one has disputed, that this was done not to purposefully decrease minority voting strength, but instead to increase the electoral opportunities of Georgia’s Democratic Party. Those individuals involved in the redistricting process testified at length regarding their intent to bolster the Democratic Party’s majority in the Georgia General Assembly. See, e.g., Pl.Ex. 20 at 23-24. We see no legal infirmity in this effort. Just as partisan progression is no guarantor of Section 5 compliance, neither is it proscribed. Moreover, this court’s previous analysis makes clear that the dilution of minority voting blocs does not always constitute retrogression. Here, in the end, there is simply no evidence that plaintiff intended to diminish the opportunities of minority voters to elect candidates of their choice. As such, the court concludes that neither Georgia’s United States Congressional redistricting plan nor its State House plan was enacted with an impermissible retrogressive purpose. While the same record evidence would support a finding that the State Senate plan does not have a retrogressive purpose, we need not reach this issue as we *97hold that the Senate plan has a retrogressive effect.

IV. CONCLUSION

Accordingly, having considered the entire record herein and the relevant statutory and case law, it is hereby

ORDERED that the American Civil Liberties Union’s motion for leave to participate as amicus curiae [108-1] is DENIED; and it is

FURTHER ORDERED that plaintiffs motion to vacate the court’s grant of intervention [92-1] is DENIED; and it is

FURTHER ORDERED that defendants’ and intervenors’ motions to strike portions of Epstein’s testimony [149][150] are DENIED.

This court is divested of jurisdiction to consider Michael King’s motion for a stay of proceedings [145] and renewed motion to intervene [144].

For the foregoing reasons, the court holds that the State of Georgia has not demonstrated by a preponderance of the evidence that the State Senate redistricting plan would not have a retrogressive effect on African American voters’ opportunities to exercise their electoral power at the polls. Accordingly, the court DENIES plaintiffs request to enter a declaratory judgment that Act No. 1EX6 does not have the purpose or effect of denying or abridging the right to vote on account of race or color.

The State of Georgia has demonstrated by a preponderance of the evidence that the Congressional redistricting plan does not have the purpose or effect of denying or abridging the right to vote on account of race or color. Accordingly, a declaratory judgment is GRANTED that Act No. 2EX11 does not violate Section 5 of the Voting Rights Act.

The State of Georgia has demonstrated by a preponderance of the evidence that the State House redistricting plan does not have the purpose or effect of denying or abridging the right to vote on account of race or color. Accordingly, a declaratory judgment is GRANTED that Act No. 2EX23 does not violate Section 5 of the Voting Rights Act.

An appropriate Judgment accompanies this Opinion and Order.

HaRry T. Edwards, Circuit Judge, con-curving, with whom SullivaN, District Judge, concurs: I concur fully in Judge Sullivan’s comprehensive opinion for the court. I write separately only to respond to the arguments made by our dissenting colleague, whose opinion is, I believe, wrong on the law and misguided as to how to weigh the various forms of evidence that have been presented to this court.

First, the law: Our dissenting colleague argues that § 5 is satisfied whenever a covered jurisdiction adopts a plan that preserves an “equal or fair opportunity” for minorities to elect candidates of their choice. This is not an accurate statement of the law. What the dissent has effectively done is to import § 2’s focus on equality into the § 5 inquiry. This is not the way this court has been charged to adjudicate a § 5 case. The dissent cites no authority for this argument, which is not surprising considering that no such authority exists.

The Supreme Court has consistently made it clear that § 5 and § 2 are procedurally and substantively distinct provisions. See Reno v. Bossier Parish School Bd., 520 U.S. 471, 477, 117 S.Ct. 1491, 187 L.Ed.2d 730 (1997) (“Bossier I ”). Section 2 protects minority voters in every state against electoral plans that deny them equal or fair voting opportunities. See Voinovich v. Quilter, 507 U.S. 146, 155, 113 S.Ct. 1149, 122 L.Ed.2d 500 (1993). Section 5, in contrast, protects minority *98voters in certain southern states from changes in electoral plans that have the purpose or effect of diminishing their voting rights relative to the status quo that is proposed to be changed. See Reno v. Bossier Parish School Bd., 528 U.S. 320, 334, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000) (“Bossier II ”).

It is therefore clear that the standard against which a plan must be measured in a § 5 preclearance case is always the existing plan. The illusive, standards of equality or proportionality are not the guideposts in a § 5 case. The electoral opportunities available under the benchmark plan frame the inquiry in determining whether the corresponding opportunities afforded by the new plan satisfy § 5. A status quo protective of African American voting strength cannot be weakened merely because that status quo exceeds the law’s minimum requirements. That preclearance can be denied only when a plan retrenches on existing opportunities hardly suggests, as the dissent would have it, that substantive retrenchment can in certain cases simply be ignored. Instead, the one-way ratchet imposed by § 5 means that tangible gains made by African Americans voters need not be surrendered merely because the State has sought to undo those gains with a plan that is (perhaps) not independently unlawful under § 2.

To so hold would be to undermine the Supreme Court’s consistent efforts to construe § 2 and § 5 “to combat different evils and, accordingly, to impose very different duties upon the States.” Bossier I, 520 U.S. at 477, 117 S.Ct. 1491. When the idea of retrogression is taken seriously, as the dissent refuses to do, it is quite obvious that a proposed plan backslides from an existing plan if it merely affords the protected class an equal opportunity to a elect a fixed number of candidates and the existing plan affords the protected group a significantly better than equal chance of electing that same slate of candidates. Accordingly, all other things being equal, a state that converts a safe district into one where African Americans have only a “fair opportunity” would be hard-pressed to preclear its plan under the § 5 analysis described by the Supreme Court. For it simply could not be said of such a plan that it “is no more dilutive than [the plan] it replaces.” Bossier II, 528 U.S. at 335, 120 S.Ct. 866.

In sum, then, neither the Supreme Court nor any other court, has held — or even hinted — that preclearance under § 5 must be granted to a plan that protects equal electoral opportunities for minority voters, even though it materially reduces those opportunities available under the existing electoral scheme. This is hardly remarkable, as the entire concept of retrogression militates against such a result. Our dissenting colleague simply blinks legal reality in insisting otherwise.

* * # * * *

This legal error infects the whole of the dissent’s analysis. Discounting the need for a serious retrogression inquiry leads the dissent to overvalue the testimony of the African American legislators who spoke in favor of the proposed Senate plan. The dissenting opinion suggests that this evidence is more probative than what it describes as the “flawed expert testimony and conflicting lay witness testimony” presented by the plan’s opponents. This approach has several serious flaws.

In the first place, it represents an inappropriate attempt to reframe the case with the testimony of politicians at the center of the court’s inquiry. This is highly problematic, because all of the parties to this litigation agreed that, in light of the State’s deliberate (and apparently strategic) decision to significantly decrease the *99black voting populations in several Senate districts, the statistical evidence concerning polarization is the principal issue in the case. The parties generally agreed that, in order to assess polarization, we must consider the degree to which African American voters have had to rely on white crossover votes to elect preferred candidates. As the issue was framed by both Georgia and the United States, these data were highlighted as the determinative factor in assessing whether the Senate plan should be precleared. The parties certainly did not suggest that the inconsistent testimony of African American politicians could somehow overcome the statistical data on polarization.

Indeed, the State’s evidentiary case is built on the assertion that racially polarized voting is uniform across the state, and uniformly low. The utility of Dr. Epstein’s analysis, which Georgia put forward as the centerpiece of its evidence, rises and falls on the validity of this proposition. The dissent ignores this and offers its own view on how the plaintiffs case should have been argued. In other words, the dissent has attempted to do for the State what it did not seek to do for itself: to elevate the testimony of three political leaders (only one of whom had first-hand knowledge of the districts most hotly in dispute) above the empirical evidence regarding the effect of declining BVAP on African American electoral opportunities. Even if it were permissible for a judge to change the theory of a plaintiffs case (with no notice to the other side) and then make evidentiary findings to which no party subscribes, the dissent is still wrong in the conclusion that it reaches. Why? Because the testimony on which the dissent focuses can in no way carry the State’s burden of proving that its Senate plan will more likely than not generate no retrogressive effect.

^ ^

Before turning to the politicians’ testimony and how it should be weighed, I first reject the dissent’s assessment of the importance and the clarity of the statistical evidence in the record before us. The only such evidence presented by Georgia was Dr. Epstein’s probit analysis. For this analysis to be at all relevant, it was crucial for the State to demonstrate that it was based on a realistic assessment of racial polarization. Otherwise, Dr. Epstein’s testimony is of virtually no utility in determining whether the numerical decreases in the disputed districts will actually affect African American voting strength. Yet, the entire basis for the State’s assumption of low polarization in these areas are data from statewide races, which the defendant’s expert, Dr. Engstrom, rejected as having little bearing on the likely degree of white crossover in local Senate elections.

Dr. Engstrom testified that polarized voting patterns persist in local elections in the Senate districts challenged by the United States. His data indicated that in nonstatewide races in the proposed Districts 2, 12, and 26, racial polarization was considerable and white crossover persistently low. He offered his expert opinion that this polarization was statistically significant and that it fatally undermined Dr. Epstein’s probit analysis. The State put on no expert testimony to refute this damning claim; nor did it submit its own statistical analysis of the data on which the Dr. Engstrom relied. Instead, using only the decidedly non-expert analyses put together by its lawyers, Georgia asked the court to accept Dr. Engstrom’s data, but to ignore the conclusions that the expert had drawn from that data. This offer, which the dissent has now credulously accepted, requires the court to cast itself adrift from the expert’s data and analyses and instead to rely on its own inexact and untested impressions on the significance of cross*100over voting patterns. Such amateur theorizing is entirely inappropriate. The simple and controlling fact in this case is that Dr. Engstrom’s expert data and analyses have not been refuted by expert testimony from the State.

Moreover, while it is true that the polarization trends appear less problematic in statewide races, that fact tells us little about the elections that are at issue in the Senate plan. It was the plaintiffs burden to demonstrate that the statewide data rehabilitate Dr. Epstein. It did not do so. The dissent now suggests several benign explanations for why white crossover may be greater in the regional races. This is all well and good, but such post hoc efforts by a judge cannot disguise the fact that the State presented not a shred of evidence — statistical or otherwise — -to support such suggestions. With Georgia bearing the ultimate burden of proof, I am simply unwilling to gamble that the serious racial bloc voting that has undeniably occurred in numerous local elections in the disputed districts will not continue under the new Senate plan.

Ultimately, we must decide this ease on the basis of the record made by the parties; it is not our role to engage in idle speculation about what might be, or what could have been, in lieu of examining and evaluating what actually has been proven — and, more importantly, what has not been proven. The State has not met its burden of proof on the crucial issues of crossover patterns and polarization. Judicial speculation cannot overcome this reality.

In an apparent effort to rescue the State from the weakness of its own statistical evidence, the dissent suggests that the testimony of the three African American legislators, only one of whom is actually qualified to speak about the districts most important to this case, accomplishes what the State’s statistical data could not. The dissent’s view appears to be that if a few well-respected African American leaders are satisfied with the challenged plan, it should pass muster under § 5. Respectfully, this is not the law.

I do not mean to suggest that support among minority representatives can never be relevant to whether a proposed reapportionment plan comports with the Voting Rights Act. In this case, however, such evidence does not save Georgia’s otherwise infirm plan. For not only is the testimony of Congressman Lewis, and State Senators Brown and Walker largely irrelevant to' the legal issue before the court, but, insofar as it is relevant, it does not contradict the statistical data that the dissent seeks to supplant.

First, nowhere do any of these esteemed politicians purport to compare the proposed Senate plan with the existing apportionment scheme. Accordingly, while some of what they have to say bear upon (albeit only in the most general terms) the opportunities available to minority candidates under the new plan, their testimony simply does not address retrogression, which, as we have explained, in the only relevant legal inquiry under § 5. Nor do these legislators address the polarization problem that is at the heart of the Court’s decision to deny preclearance. Their statements therefore cannot refute the detailed evidence presented by the United States regarding the effects of this crucial phenomena in the places where it matters most. If the lack of positive racial polarization data was the gap at the center of the State’s case, the evidence presented by these estimable men does not come close to filling that void. Whatever the dissent has taken from their testimony can therefore be but marginally related to the task *101assigned to this court by the Voting Rights Act.

Indeed, of these three legislators, neither Congressman Lewis nor Senator Walker had any direct knowledge of the demographics and voting patterns in the contested districts. Accordingly, their testimony about the general BVAP levels at which African American preferred candidates have a fair opportunity to compete, in addition to being tangential to the question of retrogression, has little bearing as to what levels might be required in Districts 2,12, and 26. The dissent’s puzzling genuflection to this vague evidence therefore does not support the result it claims.

Moreover, the testimony of Senator Brown, who represents District 26, leaves little doubt that he was speaking primarily as a loyal Democrat, interested in advancing the political fortunes of his own party. As he was actively involved in drawing the proposed lines for his home district, his support for that new shape is unremarkable. What is more telling, however, is Senator Brown’s ready recognition that his motives were primarily partisan. “[0]f course, since I am currently serving the district I looked at where I think the votes would be that would, you know, enhance the districts in terms of the likelihood the Democrats would be able to retain it as a Senate plan seat.” Defendant’s Exhibit 729, at 32. Brown acknowledged that “I was not looking at race as a predominant concern,” id. at 34, but instead that “I am looking at the overall objective that I have always wanted to maximize Democrat performance,” id. at 35.

Given that Senator Brown was the only African American state Senator from a contested district who testified on the State’s behalf, his candor is especially significant. That he and a number of his peers support the proposed Senate plan may well bespeak sound politics by partisans of the Democratic party. But that Georgia’s African American politicians sought to make their state safer for Democratic candidates does not establish (or even imply) that in so doing they did not make it worse for African American voters. Indeed, Senator Brown’s enthusiasm for the plan seems to reflect a general agreement among Georgia Democrats that the present reapportionment will preserve their partisan interests more effectively than any alternative. While such considerations are not impermissible under the Voting Rights Act, they are certainly not sufficient to satisfy the demands of § 5.

The dissent also tries to rely on the testimony of Senator Regina Thomas that she expects to win reelection. It is hardly surprising that a politician seeking reelection expresses the optimistic view that she will win her race. It would do her no good to announce that she will lose. What is more noteworthy, however, is that Senator Thomas voted against the new Senate plan, arguing that it is retrogressive.

Furthermore, Senator Thomas’ assessments regarding her chances of reelection are irrelevant. The Voting Rights Act does not protect minority incumbents; it protects minority voters. It is thus a dangerous business to conflate a politician’s assessment of her own continued electoral prospects with the genuine protection of African American voting strength.

Finally, insofar as the dissenting opinion turns to this legislative evidence on the assumption that it is less confused and contradictory than the statistical evidence, it chases a false prophet of its own invention. The three politicians on whom the dissent relies represent but a small slice of the testimony presented regarding the attitudes of Georgia’s African American political leadership to the proposed Senate plan. Indeed, it is simply inaccurate to suggest that those leaders have spoken *102with a single voice. The United States offered the testimony of a number of prominent African Americans, from each of the three Senate Districts that it has challenged, in which those witnesses expressed considerable concern about the effect of the proposed changes on minority voting strength. The Senate plan, for whatever support it has received, cannot fairly be said to represent the unanimous preferences or desires of the African American leaders in the State of Georgia. The reality is simply more complicated than our dissenting colleague suggests.

In the face of such heterogeneity of views, it seems entirely inappropriate to say, as the dissent does, that the opinions of some African Americans pale in importance compared to those of others. In determining the likely impact of the proposed changes, I can see no principled basis upon which to conclude, for instance, that Congressman Lewis’ conviction that the plan will inure to the benefit of African Americans should, in the context of this § 5 litigation, be given more probative weight than Senator Thomas’ opinion that it will not. Nor does our dissenting colleague supply any such basis. Instead, he simply declares that those witnesses whose testimony accords with his own conclusion are credible and probative, while simultaneously discounting (or ignoring altogether) the testimony of others whose views diverge from his own.

In sum, then, the political opinions on which the dissent relies simply cannot bear the weight that has been placed upon them. Why? Because these individual expressions of support do not even purport to seriously address the issues of crossover voting patterns, polarization, and, most significantly, retrogression, which are the subjects of this § 5 litigation. And, in addition, the cited political opinions do not conclusively reflect the sentiment of African American leaders in Georgia. Our dissenting colleague’s effort to convert these opinions into a legally sufficient basis for approving the disputed State Senate plan is therefore entirely unconvincing. Indeed, it is quite strange to find the dissenting opinion hinged to a political judgment whose very political nature makes it an inaccurate and unreliable indicator of the very thing that the dissent would use it to predict. The dissent not only relies on this dubious evidence, but seeks to absolve this court of its responsibility of judging by blindly deferring to the judgment of the very politicians whose actions we have been charged with scrutinizing.

Judge Sullivan’s opinion for the court is, in my view, correct in its statement and application of the law, correct in its finding of facts, and correct in its weighing of the various forms of evidence that have been presented to this court.

. Covered jurisdictions are determined by Section 4 of the Act, 42 U.S.C. § 1973b(b) (2001), and include states where voter participation was below 50% in 1964, and where a test or device was used to determine eligibility to vote in that year. Id. Georgia is a covered jurisdiction. 28 C.F.R. § 51, App. (Mar. 5, 2002).

. Section 2 of the Voting Rights Act provides that "[n]o 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.” 42 U.S.C. § 1973(a).

. Section 5 provides:

Whenever a State or political subdivision with respect to which the prohibitions set forth in section 4(a) are in effect shall enact or seek to administer any voting qualification or prerequisite to voting, or standard, practice, or procedure with respect to voting different from that in force or effect on November 1, 1964, such State or subdivision may institute an action in the United States District Court for the District of Columbia for a declaratory judgment that such qualification, prerequisite, standard, practice, or procedure does not have the purpose and will not have the effect of denying or abridging the right to vote on account of race or color, and unless and until the court enters such judgment no person shall be denied the right to vote for failure to comply with such qualification, prerequisite, standard, practice, or procedure: Provided, that such qualification, prerequisite, standard, practice, or procedure may be enforced without such proceeding if the qualification, prerequisite, standard, practice, or procedure has been submitted by the chief legal officer or other appropriate official of such State or subdivision to the Attorney General and the Attorney General has not interposed an objection within sixty days after such submission, except that neither the Attorney General’s failure to object nor a declaratory judgment entered under this section shall bar a subsequent action to enjoin enforcement of such qualification, prerequisite, standard, practice, or procedure. Any action under this section shall be heard and determined by a court of three judges in accordance with the provisions of section 2284 of title 28 of the United States Code and any appeal shall lie to the Supreme Court.

42 U.S.C. § 1973c (1994).

. The bulk of testimony in this case was submitted on paper, in accordance with the court’s scheduling order and with the consent of the parties. While the court’s order directed only that direct testimony be submitted in writing, the parties were permitted to designate deposition testimony in place of cross-examining the witnesses at trial. The live testimony at trial, therefore, consisted of cross-examination and redirect questioning of the parties’ expert witnesses.

. The court is acutely aware of its responsibilities to pro se litigants. See, e.g., Fox v. Strickland, 837 F.2d 507 (D.C.Cir.1988) (requiring courts to inform pro se plaintiffs of standard of review for summary judgment motions). Despite the fact that Mr. King is an attorney, this court has taken care to ensure that King was aware of the progress of this case.

*34The court contacted King upon receipt of his motion to intervene and directed him to file his motion in accordance with the local rule governing three-judge courts, which requires that parties file all pleadings in quadruplicate. See Local Civil Rule 9.1.

. To date, King has not filed or attempted to file an amicus curiae brief.

. Chambers faxed King a copy of the order because at no point did King subscribe to the court’s automatic faxing program, which allows parties to automatically receive facsimile copies of court orders at the time they are docketed. Counsel for plaintiff, federal defendants and intervenors all participate in the Court’s automatic faxing program and receive. facsimile copies of all orders docketed in this case.

.Footnote 1 of Plaintiff's Memorandum of Points and Authorities in Opposition to King's Renewed Motion to Intervene, filed January 17, 2002, reflects these events:

Although Mr. King’s Certificate of Service indicates he mailed his "Renewed Motion to Intervene” by mail to only two of the attorneys (only D.C. counsel and no Atlanta counsel) in this case on January 14, 2002, the undersigned counsel represents that as of the close of business on January 16, 2002, none of the counsel in this case received Mr. King’s papers. The undersigned obtained a copy of Mr. King’s Renewed Motion via fax from the Court’s law clerk on January 16, 2002, and immediately faxed copies to other counsel in this case.

. The benchmark plan is the existing district-ing plan in effect, or the last legally enforceable plan.

. It appears from the 1990 and 2000 census reports provided by the parties that the census asked respondents to identify themselves as "black.” Thus, while we refer to "African American” individuals and voters throughout the opinion, this section will use "black” to identify individuals’ responses to the census.

. The court will refer to Georgia's method of counting BVAP as "BVAP (Ga.)”, and to the United States’ as "BVAP (U.S.)”.

. Plaintiff admitted this finding of fact, PPFF ¶ 33; however, in plaintiffs own findings of fact, it states in a footnote that it believes the proper number of census responses for non-Hispanic VAP in Georgia is 1,591,421, or 26.44%. PPFF ¶ 32, n. 6; but see id. n. 7 (suggesting that proper percentages is 27.83%).

. The official legislative website for the 2001 redistricting process by the Georgia General Assembly is http://www.legis.state.ga.us/Legis /2001 — 02/reapp/index.htm. PPFF ¶ 72.

. Intervenors object that these were not "hearings” but otherwise do not contest the dates and locations. Intervenors’ Pre-trial Proposed Findings of Fact & Conclusions of Law ¶ 116.

.Regular sessions of the Georgia General Assembly commence on the second Monday in January of each year and may continue in session for a period of no longer than 40 legislative days. Ga. Const., Art. Ill, § IV, ¶ I; PPFF ¶ 10. The Governor of Georgia may convene the General Assembly in special session by proclamation and the only laws that may be enacted during such a special session are those that relate to the purposes stated in the proclamation or in any other *41amendment thereto. Ga. Const., Art. V, § II, KVII(a); PPFF ¶ 11. Special sessions of the General Assembly are limited to a period of 40 legislative days unless extended in accordance with the Georgia Constitution. Ga. Const., Art. V, § II, ¶ VII(c); PPFF ¶ 12.

. These percentages reflect BVAP, as calculated by the State of Georgia.

. For purposes of this table, we use BVAP as calculated by the State of Georgia.

. These are benchmark House Districts 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 64, 65, 66, 68, 69, 70, 71, 72, 73, 75, 93, 96, 116, 117, 118, 120, 121, 124, 127, 131, 133, 134, 136, 140, 148, 149, 151, 161 and 162. Pl.Ex. 11D; PPFF ¶ 143.

. These are benchmark House Districts 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 64, 65, 66. 68. 69. 70. 71. 72. 73. 93. 116. 117. 118, 120, 121, 124, 127, 133, 134, 136, 140, 148, 149, 151, 161, and 162. Pl.Ex. 11D; PPFF ¶ 144.

.These are benchmark House Districts 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 64, 65, 66, 68, 69, 70, 71, 72, 73, 75, 93, 96, 116, 117, 118, 120, 124, 127, 133, 134, 136, 140, 148, 149, 158, 161, and 162. Pl.Ex. HE.

. These are House Districts 43 (two-member district), 44, 45, 47, 48 (four-member district), 49, 50, 51, 55, 57, 58, 59 (three-member district), 60 (three-member district), 61 (three-member district), 62, 81, 83, 97, 98, 100, 103, 105, 107, 111, 113, 114, 124 (two-member district), 125, 128, 135, and 136. Pl.Ex. 12C.

. These are House Districts 43 (two-member district), 44, 45, 47, 48 (four-member district), 49, 50, 51, 55, 57, 58, 59 (three-member district), 60 (three-member district), 61 (three-member district), 62, 81, 83, 97, 98, 100, 105, 107, 111, 113, 114, 124 (two-member district), 135 and 136. Pl.Ex. 12C.

.These are House Districts 43 (two-member district), 44, 45, 47, 48 (four-member district), 49, 50, 51, 55, 57, 58, 59 (three-member district), 60 (three-member district), 61 (three-member district), 62, 81, 83, 97, 98, 100, 105, 107, 111, 113, 114, 124 (two-member district), 135 and 136. Pl.Ex. 12D.

. The Democratic performance numbers indicate the degree to which the voters in the given precincts or district have supported Democratic candidates.

. "TPOP” refers to total population.

. "TVAP” refers to total voting age population.

. BVAP numbers reflect Georgia’s method of calculating BVAP.

. This BVAP is calculated according to Georgia's method; according to the Attorney General’s Guidance, the BVAP of District 44 is 48.52%.

. The margin of victory was between 73 and 76 votes.

. Epstein estimates the level of African American support for the white candidates by using a methodology called Dr. King’s regression, or ecological inference, analysis. This methodology is discussed in more detail below.

. The court again notes that, although this information appears to be crucial to the methodology employed by Epstein, it was not provided to defendants or intervenor-defendants until shortly before the beginning of the trial. Nevertheless, defendants thoroughly cross-examined Epstein about his calculations and his conclusion that the differences were not statistically significant.

The court also emphasizes that Epstein did not attempt to rely on the table’s calculations to demonstrate voting patterns in the districts, and calculated crossover in the existing, and not the proposed, Senate districts. Tr., Day 1, p.m. at 71.

. Specifically, one African American candidate of choice was elected at 53% BVAP, and four candidates of choice were elected at 54% BVAP. Candidates of choice were not elected at 49% BVAP, 50.0% BVAP, 53% BVAP, 53.5% BVAP and an indeterminate result was found for a district with 52.28% BVAP. Pl.Ex. 25, app. II.

There are no districts in the database between 36% and 49% BVAP, and no districts in the database between 54 and 56% BVAP.

. Epstein stated that “with the McKinney election, ... there’s a coding rule at issue, and so I think it’s appropriate to do some analysis on whether or not that’s a crucial coding issue.” Tr., 2/5/02, a.m. at 43. It does not appear, however, that he performed this analysis, and there is no testimony in the record as to what constitutes a "crucial coding issue.” Id.

. Ecological regression analysis provides an estimate of the support for various candidates among both African American and non-African American voters based on statistically significant correlations between voters' race and voting patterns. One court described ecological regression as a standard statistical technique that “comparéis] the votes a candidate received in an election with the racial composition of the electorate, and producéis] estimates of the voting behavior of [minority voters] and white voters.” Old Person v. Cooney, 230 F.3d 1113, 1123 (9th Cir.2000).

The “King ecological inference” approach is a technique developed in 1997, which uses all available data to generate a more accurate estimate of voting behavior. Epstein testified that a scholarly article recently predicted that the King method of ecological regression improves upon traditional methods by approximately 16%, thus getting "closer to the truth,” but not solving the inherent aggregation bias problem. Tr., 2/5/02, a.m. at 106. Aggregation bias is a problem that arises from attempting to infer individual voting behavior from aggregate-level behavior, or, in other words, in predicting how voters of different races voted in an election by looking at a precinct's demographic characteristics and data such as voter turn-out.

. A homogeneous precinct analysis considers the election results from precincts that are closest to racially homogeneous in character. For example, the analysis generally reports the percentage of the votes a candidate or set of candidates receive within the precincts in which over 90% of the registered voters or people receiving ballots was not African American and within those in which over 90% was African American.

. Engstrom reports predicted voting patterns as calculated by regression analysis, Ring’s ecological inference, and homogeneous precinct analysis. Engstrom, however, testified that King’s ecological inference is generally considered to be the most accurate method of calculation, and the court will therefore refer to these estimates.

. Effectively, then, the State would have us adopt the converse of the argument rejected by the Supreme Court in the Bossier Parish cases. There, the Court rebuffed the claim that preclearance must be denied where a proposed plan violates Section 2 (or the Constitution). See Bossier II, 120 S.Ct. at 877; Bossier I, 520 U.S. at 476-85, 117 S.Ct. 1491.

. The court does not consider Senate District 34 in this range, as the district is a newly created "majority-minority” district.

. On redirect, Epstein relied on his S-curve to testify that, in District 43, with 65.18% BVAP, he estimated the probability at 90% to 95%; in district 48, with 61.13%, he estimated probability at approximately 95%. Tr., 2/5/02, a.m. at 84.

. The following table shows the percentages of black registered voters residing in the precincts that were moved out of the proposed districts, and those that were added to the districts.

. Plaintiff's responses present different calculations. These are marked with an asterisk. USPFF at ¶¶ 181, 326, 380; PL’s Resp. to ¶¶ 181, 326, 380.

. Epstein testified: "the great advantage of using probit analysis here, is, I’m not concerned with trying to figure out whether or *85not black candidates are getting elected via greater turnout, crossover, registration, ... campaign appeals. Those all might be there and they’re important, but if all you want to do is ... estimate something on the level of district characteristics, then you’re using estimated district characteristics.... I'm looking at the different kinds of districts, see who elected a black candidate of choice and who didn’t.” Tr., 2/5/02, a.m. at 108-09.

. 65.39% of the proposed Senate District 2 is comprised of the population from the existing District 2. U.S.Ex. 112. 72.68% of the proposed Senate District 12 is comprised of the population from the existing District 12. Id.

. The United States introduced a listing of counties included in the proposed Senate Districts. This listing indicates that Dougherty County is split between the proposed Senate Districts 12, 13, and 14. U.S.Ex. 111. In contrast, Bibb County appears to be wholly located within the proposed Senate District 26. Id.

. To the extent that African American legislators testified that they believe candidates of choice could be elected under the proposed plan, they did not suggest that racially polarized voting did not exist in Georgia.

. The State advances two additional arguments, which we address only briefly. Plaintiff argues that Engstrom's analysis is flawed *87because he failed to consider the outcomes of the elections included in his report. The State belabors the point that several of the African American candidates in races, in which Engstrom finds racially polarized voting, were elected. However, plaintiff offered no testimony that would persuade this court that these candidates’ victories disprove the existence of racially polarized voting. Rather, Engstrom testified that the fact that an African American candidate won in the general election would not change any of his opinions with respect to the levels of racially polarized voting. Tr., 2/5/02, p.m. at 86-87. He further explained that the victories were typically in majority-minority jurisdictions. Id. Contrary to the suggestion of our dissenting colleague, the court in no way considers electoral victories by African American candidates to be aberrations; rather, the court can conclude only that, given specific demographics and voting patterns, African American candidates of choice may win election despite racially polarized voting.

The State of Georgia also argues that Eng-strom's report does not take into account the level of African American voter turn-out at Democratic primaries. Plaintiff's contention appears to rest on the theory that, if African American voters generally participate in Democratic Party primaries at a higher rate than other voters, they are able to influence the outcome of the primaries, and that the Democratic candidate was likely to win in the general election. However, there is no reliable record evidence concerning voter turn-out, at primaries or at general elections.

. By way of illustration only, the court provides the following calculations of average differences in support for African American preferred candidates in those elections listed by Engstrom in Senate Districts 2 and 26, and in Bibb County, which appears to be primarily, if not totally, within District 12.

Average difference in percentages of African American and Non-African American Support for African American Preferred Candidates District 2_District 12 District 26

*88Senate 2 Statewide Senate 12 elections elections elections _(primaries) Statewide Bibb County Statewide elections elections elections

63.58%45.99%63.05% 28.43%55.72%40.2%

The differentials are based on the King Ecological Inference numbers contained in Engstrom's Tables 1-3.

The differentials are based on the King Ecological Inference numbers contained in Eng-*89strom’s Tables 1-3.

. See, e.g., U.S.Ex. 704, 104:11-19 (Senator Thomas); U.S.Ex. 716, 22:4-24 (Dr. Jackson, NAACP); U.S.Ex. 708, 65:7-14, 68:1-11 (Johnson, NAACP); U.S.Ex. 505 ¶ 9; U.S.Ex. 709, 35:5-20; 36:2-7; 63:17-65:4 (Alderman Jones); U.S.Ex. 712, 53:20-54:5 (Alderman Jackson); U.S.Ex. 507 ¶ 12; U.S.Ex. 717, 22:20-25:8, 27:23-28:15 (County Bd.Comm. Odell); U.S.Ex. 719, 66:24-67:3 (County Bd. Comm. Rivers).