Full opinion text
OPINION and ORDER RIPPLE, Circuit Judge. OPINION TABLE OF CONTENTS 1. Background.. .843 A. Reapportionment in Wisconsin. . .844 B. Drafting of Act 43... 846 C. Prior Court Challenges to Act 43...853 II. Procedural History.. .854 A. Allegations of the Complaint.. .854 B. Motion to Dismiss... 855 C. Motion for Summary Judgment. . .856 D. Witnesses Testifying at Trial.. .857 E. Post-Trial Briefing.. .862 III. The Legal Landscape... 863 A. The Foundational Case Law.. .864 B. Present Supreme Court Precedent. . ,867 IV.Elements of the Cause of Action. . .883 A. Discriminatory Intent or Purpose. . .884 B. Discriminatory Effect of Act 43...898 V. Justification.. .910 VI. Standing.. .927 VII. Order... 930 A. Remedy.. .930 B. Evidentiary Matters... 930 Appendices The plaintiffs have brought this action alleging that Act 43, the redistricting plan enacted by the Wisconsin Legislature in 2011, constitutes an unconstitutional partisan gerrymander. Specifically, they maintain that the Republican-controlled legislature drafted and enacted a redistricting plan that systematically dilutes the voting strength of Democratic voters statewide. We find that Act 43 was intended to burden the representational rights of Democratic voters throughout the decennial period by impeding their ability to translate their votes into legislative seats. Moreover, as demonstrated by the results of the 2012 and 2014 elections, among other evidence, we conclude that Act 43 has had its intended effect. Finally, we find that the discriminatory effect is not explained by the political geography of Wisconsin nor is it justified by a legitimate state interest. Consequently, Act 43 constitutes an unconstitutional political gerrymander. This opinion constitutes our findings of fact and conclusions of law pursuant to Federal Rule of Civil Procedure 52(a)(1). I BACKGROUND We begin our consideration of the plaintiffs’ claims by examining Wisconsin’s statutory requirements for redistricting as well as its recent redistricting history. A. Reapportionment in Wisconsin 1. The State’s constitutional and statutory framework Reapportionment of state legislative districts is a responsibility constitutionally vested in the state government. See, e.g., Growe v. Emison, 507 U.S. 25, 34, 113 S.Ct. 1075, 122 L.Ed.2d 388 (1993) (citing U.S. Const. art I., § 2); Chapman v. Meier, 420 U.S. 1, 27, 95 S.Ct. 751, 42 L.Ed.2d 766 (1975). Although some states have chosen to avoid the problem of partisan gerrymandering by vesting this power in a neutral body designed specifically to perform that delicate function, see Arizona State Legislature v. Arizona Indep. Redistricting Comm’n, — U.S.-, 135 S.Ct. 2652, 2661-62 & n.6, 192 L.Ed.2d 704 (2015), the people of Wisconsin have so far chosen to rely on its legislature to reapportion its districts after the decennial census. They have vested responsibility in the bicameral legislature composed of the Wisconsin State Senate and the Wisconsin State Assembly. Wis. Const. art. IV, §§ 1, 3. According to Wisconsin law, “[t]he state is divided into 33 senate districts, each composed of 3 assembly districts. Each senate district shall be entitled to elect one member of the senate. Each assembly district shall be entitled to elect one representative to the assembly.” Wis. Stat. § 4.001. The Wisconsin Constitution directs the Wisconsin legislature, “[a]t its first session after each enumeration made by the authority of the United States,” to “apportion and district anew the members of the senate and assembly, according to the number of inhabitants.” Wis. Const, art. IV, § 3. The Wisconsin Constitution also imposes specific requirements for reapportionment plans. Assembly districts are “to be bounded by county, precinct, town or ward lines, to consist of contiguous territory and be in as compact form as practicable.” Id. § 4. With respect to political subdivisions, a prior federal district court observed that, “[although avoiding the division of counties is no longer an inviolable principle, respect for the prerogatives of the Wisconsin Constitution dictate that wards and municipalities be kept whole where possible.” Baumgart v. Wendelberger, Nos. 01-C-0121 & 02-C-0366, 2002 WL 34127471, at *3 (E.D. Wis. May 30, 2002), amended by 2002 WL 34127473 (E.D. Wis. July 11, 2002). The Wisconsin Constitution further requires that “no assembly district shall be divided in the formation of a senate district.” Wis. Const, art. IV, § 5. In addition to the state constitutional requirements, the Wisconsin legislature must comply with federal law when redistricting. In particular, state legislatures must ensure that districts are approximately equal in population, so that they do not violate the “one-person, one-vote” principle embedded in the Equal Protection Clause of the Fourteenth Amendment. See Reynolds v. Sims, 377 U.S. 533, 568, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964) (“[T]he Equal Protection Clause requires that the seats in both houses of a bicameral state legislature must be apportioned on a population basis.”); see also Brown v. Thomson, 462 U.S. 835, 842-43, 103 S.Ct. 2690, 77 L.Ed.2d 214 (1983) (holding “that an apportionment plan with a maximum population deviation under 10%” is presumptively constitutional, while a population deviation larger than 10% must be justified by the state); Harris v. Arizona Indep. Redistricting Comm’n, — U.S. -, 136 S.Ct. 1301, 1306-07, 194 L.Ed.2d 497 (2016) (same). Further, states also must comply with § 2 of the Voting Rights Act of 1965, which focuses on preserving the voting power of minority groups. 52 U.S.C. § 10301; see also Thornburg v. Gingles, 478 U.S. 30, 47, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). Redistricting laws in Wisconsin are enacted, in large measure, in the same manner as other legislation, specifically, by way of bills originating in either house of the legislature, see Wis. Const, art. IV, § 19. Tad Ottman, aide to the Senate Majority Leader, explained in some detail this legislative process: [Legislators will work either on their own or with drafters or with a small group of people to develop legislation. Usually it’s developed among members of your own party, if not just the individual legislator. They create a proposal with the assistance of the Legislative Reference Bureau. At that point, the bill is often, but not always, circulated among other legislators to see if anybody else would want to sign on .... The bill is then circulated. At some point it is introduced.... And then once they are introduced, they are assigned to a committee. The committee chairman or chairwoman can choose to hold a public hearing on that piece of legislation. Most of the time a public hearing is held.... And then that legislation is forwarded to the full body, either the Senate or the Assembly, for debate and then it is passed over to the other House where a similar process occurs. A bill must then “be presented to the governor,” who can sign or veto the bill. Wis. Const, art. V, § 10. The caucus system plays a significant role in the legislative process. Caucus meetings are held in the morning prior to the legislative session to vet legislation internally before a vote on the floor. Professor William Whitford, a named plaintiff and retired professor of law from the University of Wisconsin, testified that important “debate and discussion,” as well as the “vote[] that matters,” occur within the caucus meetings. “Once the party caucuses come to a majority result, the other members of the party are expected to follow the party line ....” Thus, it is “extremely difficult” to pass legislation through a bipartisan coalition. 2. The modern history of reapportionment in Wisconsin In the wake of the 1980 census, the plan that had been enacted in 1972 could no longer satisfy the constitutional requirement of “one-person one-vote.” See Wis. State AFL-CIO v. Elections Bd., 543 F.Supp. 630, 631 (E.D. Wis. 1982). In response to these changes in population, a redistricting plan was drafted and enacted by the Wisconsin legislature, which had a Democratic majority, but it was vetoed by the Republican governor. Consequently, a federal district court was asked to devise a remedy. See id. at 632-33. Upon reviewing several plans submitted by legislators and interest groups, the court “reluctantly concluded” that it could “be more faithful to the goals of reapportionment” by drafting its own plan. Id. at 637. In doing so, the court focused on ensuring population equality, avoiding the dilution of racial minority voting strength, and keeping communities of interest together. Id. at 637-39. This “AFL-CIO Plan” remained in effect for one election in 1982. As a result of that election, the Democratic Party held control of both houses of the Wisconsin legislature and also gained the governor’s office. The legislature passed, and the governor signed, a new apportionment plan that lasted for the rest of the decennial period. See 1983 Wis. Sess. Laws 633. Following the 1990 election, the Wisconsin government again was divided between two political parties. See Prosser v. Elections Bd., 793 F.Supp. 859, 862 (W.D. Wis. 1992). The Democratic Party controlled both houses of the Wisconsin legislature while the governor was a Republican. Id. “For that or other reasons, no bill to reapportion the legislature had been enacted into law” by January 1992, leading several Republican legislators to challenge the existing apportionment plan “as unconstitutional and violative of the Voting Rights Act.” Id. As a result, the federal court was asked to draft a new plan. In an attempt to play a more limited role in the redistricting process, the court “asked the parties at the outset whether they had any objection ... to [the court’s] selecting the best of the submitted plans rather than trying to create [its] own plan.” Id. at 865 (emphasis removed). Upon receiving these submissions, however, the court determined that the plans bore “the marks of their partisan origins.” Id. at 865. It therefore used parts of one Republican plan and one Democratic plan. The court plan preserved the strengths of the partisan plans, “primarily population equality and contiguity and compactness,” while “avoiding] their weaknesses.” Id. at 870. The plan remained in effect through the 2000 election. Following the 2000 census, a divided Wisconsin legislature again was unable to agree upon a redistricting plan. Arrington v. Elections Bd., 173 F.Supp.2d 856, 862 (E.D. Wis. 2001). In an ensuing law suit, the federal district court determined that “the existing Wisconsin Assembly and Senate districts,” which had not been redrawn since 1992, were “violative of the ‘one person, one vote’ standard.” Baumgart, 2002 WL 34127471, at *1. A new plan was therefore necessary. The court considered sixteen plans that had been submitted by legislators and other interest groups, but “found various unredeemable flaws” in all of them. Id. at *6. The court therefore drew a plan “in the most neutral way it could conceive—by taking the 1992 reapportionment plan as a template and adjusting it for population deviations.” Id. at *7. In making these changes, the court attempted to “maintain[] municipal boundaries and unit[e] communities of interest.” Id. The “Baumgart Plan” was in effect from 2002 until 2010. B. Drafting of Act 43 In 2010, for the first time in over forty years, the voters of Wisconsin elected a Republican majority in the Assembly, a Republican majority in the Senate, and a Republican Governor. This uniformity in control led the Republican leadership to conclude that a legislatively enacted redistricting plan was possible. In January 2011, Scott Fitzgerald, Wisconsin Senate Majority Leader, and Jeff Fitzgerald, Speaker of the Wisconsin Assembly, retained attorney Eric McLeod and the law firm of Michael Best & Fried-rich, LLP, to assist with the reapportionment of the state legislative districts. The firm supervised the work of Tad Ottman, staff member to Senate Majority Leader Fitzgerald; Adam Foltz, staff member to Speaker Fitzgerald; and Joseph Handriek, a consultant with the law firm Reinhart Boerner Van Deuren s.c., in planning, drafting, and negotiating the new district-ing plan. Ottman, Foltz, and later Han-drick, worked in a room located in the offices of Michael Best & Friedrich, which they referred to as the “map room.” Ottman, Foltz, and Handriek also received assistance from Professor Ronald Keith Gaddie, a professor of political science at the University of Oklahoma. Michael Best & Friedrich had retained Professor Gaddie “as an independent advisor on the appropriate racial and/or political make-up of legislative and congressional districts in Wisconsin.” Professor Gaddie described his job as “devisfing] measures and consult[ing] ... about measures” of partisanship, compactness, “the integrity of counties, the integrity of city boundaries, the so-called good government principles of redistricting.” “Where [he] ... spent most of [his] time was trying to disentangle the performance of the majority/minority districts in Milwaukee County.” A “significant part” of his work was “building a regression model to be able to test the partisan makeup and performance of districts as they might be configured in different ways.” As explained by one of the plaintiffs’ experts, Professor. Kenneth Mayer, “[regression is a technique where we can seek to explain a dependent variable, the variable that we’re trying to account for.... [W]e attempt to explain the values that a dependent variable take[s] with what are called independent variables or underlying causal variables.” In this instance, Professor Gaddie’s dependent variable was the baseline partisanship of a unit of geography, which then could be aggregated into different configurations of Assembly districts. In this way, Professor Gaddie was able to assess the partisanship of the Assembly maps that the drafters passed on to him for analysis. Professor Mayer testified that “the political science literature is essentially unanimous” that the approach taken by Professor Gaddie is “the appropriate method,” and Professor Mayer used the same methodology to construct his Demonstration Plan. Ottman, Foltz, and Handriek began drafting the map that would become Act 43 in April 2011, after they received census data from the Legislative Technology Services Bureau (“LTSB”). The LTSB also had provided them with computers loaded with the redistricting software, auto-Bound. Ottman described in detail how the software was used: [Y]ou would open up a plan that you’d been working on or label a new plan and assign it the Assembly district that you wanted to work with and then you could also pick a color that you wanted that Assembly district to be. It’s sort of like a color-by-number exercise. ... You also determine what other layers that you want to look at on the screen. There were a number of different overlays that you have, anywhere from existing Senate and Assembly districts, ... count[y] boundaries, municipal boundaries, ward boundaries all the way down to census block boundaries. As a practical matter what you tried to do is you would zoom in the region of your screen to the area that you’re looking at to the smallest amount that you could see and then have kind of the fewest layers displayed that you would need because the more information that you were requiring it to display slows down the computer speed a lot and makes it really slow to render. And then what you would do is there were a couple different ways that you could add population to the district. Ottman further explained that, in more populated areas, the drafters worked more at the ward level: “So you would have the wards displayed and you would literally draw a circle, click on it, and it would assign it to the map and fill it in.” “In other parts of the state ... you might do that at the county level because it’s so sparsely populated so you’d grab three or four counties at [a] time.” When the drafters would increase the area size of the districts that they were drawing, autoBound provided demographic information for the area that the drafter had included, such as the number of people in the district, the deviation from the ideal population, voting-age population, and different minority group populations. It also allowed the user to include “customized ... demographic data.” One piece of “customized demographic data” employed by the drafters was a composite partisan score. From the time that Ottman, Foltz, and Handrick received the census data from the LTSB, they worked to develop a composite partisan score that accurately reflected the political make-up of the population units. Having this measure was necessary so that, when they aggregated those units into new districts, they could assess the partisan make-up of the new district they had drawn. On April 19, 2011, they developed a composite of “all statewide races from [20]04 to 2010” that “seem[ed] to work well.” They sent this composite measure to Professor Gad-die, who tested it against his regression model. Professor Gaddie confirmed to Handrick that “the partisanship proxy you are using (all races) is an almost perfect proxy for the open seat vote, and the best proxy you’ll come up with.” Once Professor Gaddie confirmed the usefulness of their composite measure, Ottman, Foltz, and Handrick could “assess the partisan impact of the map[s] that [they] drew.” Although Ottman, Foltz, and Handrick worked in the same room at Michael Best & Friedrich, they worked independently on their own maps. They drew several statewide maps, and even more regional maps from which the legislative leadership eventually would choose. As they drew the maps, they would ensure that the districts were “close-to-ideal population.” They did an “eyeball test” for “compactness and contiguousness.” They “looked at ... what the core of the existing district was compared to the new district,” “looked at municipalities that were split,” whether the new district had changed Senate districts, and “where incumbents lived.” The drafters were attentive to traditional districting criteria like population equality, compactness, and municipal splits throughout the drafting process. When the drafters had created a statewide map with which they were satisfied, they would export the district-by-district partisanship scores from autoBound into a spreadsheet for that “finalized” “statewide” plan. The drafters used their composite score to evaluate the statewide maps that they had drawn based on the level of partisan advantage that they provided to Republicans. In many instances, the names of the maps reflected the level of partisan advantage achieved by the districting plan; for instance, there are maps labeled “Assertive” and “Aggressive.” Foltz testified that “aggressive” in this context meant “probably that [the map] was a more aggressive map with regard to GOP leaning.” The drafters created spreadsheets which collected the partisan scores, by district, for each of the statewide map alternatives. Each spreadsheet included a corresponding table comparing the partisan performance of the draft plan to the prior map drawn by the Baumgart court, which they called the “Current Map.” These performance comparisons were made on the following criteria: “Safe” Republican seats, “Lean” Republican seats, “Swing” seats, “Safe” Democratic seats, and “Lean” Democratic seats. The process of drafting and evaluating these alternative district maps spanned several months. In early April 2011, the drafters produced a document comparing the partisan performance of the Current Map to two early draft maps: Joe’s Base-map Basic and Joe’s Basemap Assertive. Under the Current Map, the drafters anticipated that the Republicans would win 49 Assembly seats. This number increased to 52 under the Joe’s Basemap Basic map and to 56 under the Joe’s Base-map Assertive map. The number of safe and leaning Republican seats increased from 40 under the Current Map to 45 under the Joe’s Basemap Basic map and 49 under the Joe’s Basemap Assertive map; the number of swing seats decreased from 19 to 14 to 12, The number of safe and leaning Democratic seats, however, remained roughly the same under all three maps, hovering between 38 and 40. The drafters prepared and evaluated the partisan performance of at least another six statewide alternative maps. Each of these maps improved upon the anticipated pro-Republican advantage generated in the initial two draft plans. The total number of safe and leaning Republican seats now ranged between 51 and 54, and the number of swing seats was decreased to between 6 and ll. The number of safe and leaning Democratic seats again remained about the same under each draft map, ranging between 37 and 39. The drafters sent their completed draft maps to Professor Gaddie for further analysis. For each map, Professor Gaddie created an “S” curve—a “visual aide[ ] to demonstrate the partisan structure of Wisconsin politics.” These “S” curves show how each map would operate within an array of electoral outcomes. The “S” curves give a visual depiction of how each party’s vote share (on the x axis), ranging from 40% to 60%, relates to the number of Assembly seats that party likely will secure (on the y axis). Democratic seats are depicted by shades of blue, and Republican seats by shades of red. To produce the “S” curves, Professor Gaddie first used his regression analysis to calculate the expected partisan vote shares for each new district. He then shifted the vote share of each district ten points in either direction, from 40% to 60%, and assigned a color to districts that “tend[ed]” towards, or were “safe” seats, for that party. The “S” curves—at least some of which were printed in large format and kept in the map room—allowed a non-statistician, by mere visual inspection, to assess the partisan performance of a particular map under all likely electoral scenarios. On one occasion, Senator Fitzgerald came to the map room, and Professor Gaddie showed him one of the large printouts of the “S” curves and “basically explain[ed] how to interpret” them. Not long after Professor Gaddie had performed his analyses, the Republican legislative leadership contacted the drafters and indicated that they wanted to be prepared to act on a redistricting plan. Over several days in early June, the drafters presented a selection of regional maps drawn from their statewide drafts, approximately three to four per region, to the Republican leadership. - Along with these regional alternatives, the leadership “saw the partisan scores for the maps that [the drafters] presented to them in those alternatives.” Foltz testified during his deposition that, although he could not recall a particular example, he was sure that he was asked- by the leadership about the partisan performance of the various regional options. Following this meeting, the drafters amalgamated the regional'alternatives chosen by the leadership. Foltz testified that “the draft map called team map emerged as a result of the ... leadership’s choices at those meetings.” Under the Team Map, which was also referred to as the “Final Map,” the Republicans could expect to win 59 Assembly seats, with 38 safe Republican seats, 14 leaning Republican, 10 swing, 4 leaning Democratic, and 33 safe Democratic seats. In a document bearing the heading “Tale of the Tape,” the drafters, among other things, compared the partisan performance of the Team Map directly to the Current Map on each of these criteria. They highlighted specifically that under the Current Map, 49 seats are “50% or better” for Republicans, but under the Team Map, “59 Assembly seats are 50% or better.” The Team Map underwent even more intense partisan scrutiny in a document identified as “summary.xlsx.” The drafters divided the new Team Map districts into six categories of partisan performance, listing beside each district its “new incumbent” and its Republican vote share under the Current Map and the Team Map. The drafters considered five districts to be “Statistical Pick Up[s],” meaning they were currently held by a Democratic incumbent but likely to become Republican; they grouped fourteen districts under the heading “GOP seats strengthened a lot”; they designated eleven districts “GOP seats strengthened a little”; they labeled three districts as “GOP seats weakened a little”; they considered another three GOP districts “likely lost”; and, finally, they identified four districts where the Democrats were “weakened.” The drafters also listed the twenty Republican Assembly members who, under the Team Map, could be considered “GOP Donors to the Team”: “Incumbents with numbers above 55% that donate[d] to the team.” These representatives stood in contrast to “GOP non-donors,” who were Republican incumbents with “over 55% who d[id] not donate points.” The Team Map was then sent to Professor Gaddie, who conducted an “S” curve analysis. The. Team Map demonstrated that Republicans would maintain a majority under any likely voting scenario; indeed, they would maintain a 54 seat majority while garnering only 48% of the statewide vote. The Democrats, by contrast, would need 54% of the statewide vote to capture a majority. Once the map had been finalized, Foltz presented each Republican member of the Assembly with information on his or her new district. The memos prepared for the Assembly members informed them whether the district number had changed, whether adjustment to the district population was necessary based on the census numbers, and provided a “[cjomparison of [k]ey [rjaces” in the new district compared to the old. Specifically, the memorandum detailed what percentage of the population in the old and new districts voted for Republican candidates in representative statewide and national elections held since 2004. This information also was provided in terms of raw votes. The memoranda did not provide the individual legislators with any information about contiguity, compactness, or core population. Ottman engaged in a similar process with Republican members of the State Senate. For each meeting, he created a talking-points memo that included information about population, where changes in the district’s population had occurred, and the geography of the new district. These also contained information on how the reconfigured district had voted in national and statewide elections. Ottman also made a presentation to the Republican caucus. His notes for that meeting state: “The maps we pass will determine who’s here 10 years from now,” and “[w]e have an opportunity and an obligation to draw these maps that Republicans haven’t had in decades.” On July 11, 2011, the redistricting plan was introduced by the Committee on Senate Organization. On July 13, 2011, a public hearing was held, during which Ott-man and Foltz presented the plan and fielded questions. The Senate and Assembly passed the bill on July 19, 2011, and July 20, 2011, respectively. The Governor signed the bill, and it was published as Wisconsin Act 43 on August 23, 2011. C. Prior Court Challenges to Act 43 Even before Act 43 was passed, two actions were brought challenging the plan on constitutional and statutory grounds, including under Section 2 of the Voting Rights Act. See Baldus v. Members of the Wis. Gov’t Accountability Bd., 849 F.Supp.2d 840, 846-47 (E.D. Wis. 2012). The court consolidated the actions for decision and concluded that the plan did not violate the “one-person, one-vote” principle, nor did it violate the Equal Protection Clause by “disenfranchise[ing]” voters who were moved to a new Senate district and were unable to vote for their state senator for another two years. Id. at 849-51, 852-53. However, the court did find that the plaintiffs were entitled to relief on their claim that Act 43 violated the Voting Rights Act by diluting the voting power of Latino voters in Milwaukee County, and it ordered the State to redraw these districts. Id. at 859. The remainder of Act 43, however, remained intact and governed the 2012 and 2014 Assembly elections. In 2012, the Republican Party received 48.6% of the two-party statewide vote share for Assembly candidates and won 60 of the 99 seats in the Wisconsin Assembly. In 2014, the Republican Party received 52% of the two-party statewide vote share and won 63 assembly seats. II PROCEDURAL HISTORY A. Allegations of the Complaint We now turn to the dispute before this court. Plaintiffs William Whitford, Roger Anclam, Emily Bunting, Mary Lynne Do-nohue, Helen Harris, Wayne Jensen, Wendy Sue Johnson, Janet Mitchell, James Seaton, Allison Seaton, Jerome Wallace, and Don Winter are United States citizens registered to vote in Wisconsin. They reside in various counties and legislative districts throughout Wisconsin. All of them are “supporters of the Democratic party and of Democratic candidates and they almost always vote for Democratic candidates in Wisconsin elections.” Defendants are Beverly R. Gill, Julie M. Glan-cey, Ann S. Jacobs, Steve King, Don Millis, and Mark L. Thomsen, each in his or her official capacity as a member of the Wisconsin Elections Commission. According to the plaintiffs, in drafting Act 43, the Republicans employed two gerrymandering techniques: “cracking”— “dividing a party’s supporters among multiple districts so that they fall short of a majority in each one”—and “packing”— “concentrating one party’s backers in a few districts that they win by overwhelming margins,” in order to dilute the votes of Democrats statewide. This “cracking and packing resulted] in “wasted’ votes: votes cast either for a losing candidate (in the case of cracking) or for a winning candidate but in excess of what he or she needs to prevail (in the case of packing).” They therefore urge the court to adopt a new measure for assessing the discriminatory effect of political gerrymanders—the efficiency gap (or “EG”). “The efficiency gap is the difference between the parties’ respective wasted votes in an election, divided by the total number of votes cast.” When two parties waste votes at an identical rate, a plan’s EG is equal to zero. An EG in favor of one party, however, means that the party wasted votes at a lower rate than the opposing party. It is in this sense that the EG arguably is a measure of efficiency: Because the party with a favorable EG wasted fewer votes than its opponent, it was able to translate, with greater ease, its share of the total votes cast in the election into legislative seats. In short, the complaint alleges that Act 43 purposely distributed the predicted Republican vote share with greater efficiency so that it translated into a greater number of seats, while purposely distributing the Democratic vote share with less efficiency so that it would translate into fewer seats. The plaintiffs’ complaint incorporated the EG into a proposed three-part test for partisan gerrymandering. First, plaintiffs would have to establish that a State had an intent to gerrymander for partisan advantage. Second, the plaintiffs would need to prove a partisan effect, by proving that the EG for a plan exceeds a certain numerical threshold (which the plaintiffs proposed, based on historical analysis, to be 7%). If a plan exceeds that threshold, the plaintiffs asserted that it should be presumptively unconstitutional. Third, and finally, the plaintiffs placed the burden on the defendants to rebut the presumption by showing that the plan “is the necessary result of a legitimate state policy, or inevitable given the state’s underlying political geography.” If the state is unable to rebut the presumption, then the plan is unconstitutional. The plaintiffs alleged that they had satisfied all of these elements. According to the complaint, Act 43 “was drafted and enacted with the specific intent to maximize the electoral advantage of Republicans and harm Democrats to the greatest possible extent.” Additionally, Act 43 “produced a pro-Republican efficiency gap of 13% in 2012 and 10% in 2014.” They further claimed that this EG is unjustified because one of their experts, Professor Mayer, had crafted a “Demonstration Plan” with “an efficiency gap of just 2% in 2012,” which “perform[ed] at least as well as [Act 43] on every other relevant metric.” For these reasons, plaintiffs claimed that Act 43 “treats voters unequally, diluting their voting power based on their political beliefs, in violation of the Fourteenth Amendment’s guarantee of equal protection,” and “unreasonably burdens their First Amendment rights of association and free speech,” They requested a declaration that Act 43 is unconstitutional, an injunction prohibiting further elections under the map, and the drawing of a new redistricting map. B. Motion to Dismiss The defendants filed a motion to dismiss on August 18, 2015, which contended that the court could not grant relief for three primary reasons. First, the defendants argued that the EG was directly analogous to the proportional-representation standard rejected by the Supreme Court in Vieth v. Jubelirer, 541 U.S. 267, 287-88, 124 S.Ct. 1769, 158 L.Ed.2d 546 (2004). Second, the defendants argued that the EG failed to account for the impact of traditional districting criteria like, contiguity and compactness. Finally, the defendants argued that the plaintiffs lacked the standing to challenge Act 43 on a statewide basis;- and instead could only challenge their individual districts. In an order dated December 17, 2015, we denied defendants’ motion to dismiss. We first noted that the claim was justicia-ble, and that, “[u]ntil a majority of the Supreme Court rules otherwise, lower courts must continue to search for a judicially manageable standard.” We acknowledged the defendants’ argument that the EG was analogous to a proportionality standard, but noted that the plaintiffs’ experts disagreed with the defendants’ contention and that factfinding therefore was needed. We concluded that “[a] determination whether plaintiffs’ proposed standard is judicially manageable relies at least in part on the validity of plaintiffs’ expert opinions” and that a more developed record would be necessary to resolve that question. Finally, we concluded that the plaintiffs had standing, explaining that “[bjecause plaintiffs’ alleged injury in this case relates to their statewide representation, it follows that they should be permitted to bring a statewide claim.” We noted, however, that the defendants were “free to raise this issue again on a more developed record.” C. Motion for Summary Judgment Defendants subsequently filed a motion for summary judgment, raising new challenges to the plaintiffs’ claims. In the motion, the defendants argued that the EG metric was overinclusive and captured several plans—including court-drawn plans in Wisconsin—that were not drawn with any partisan intent. Furthermore Democratic voters tended to live in cities, which created a “natural packing” effect and distorted the EG. The defendants acknowledged the plaintiffs’ argument that a requirement of partisan intent could remedy this over-inclu-sivity problem, but noted that the intent element was not sufficiently demanding. The defendants contended that “[a]s long as redistricting is done by a legislature, it should not be very difficult to prove that the likely political consequences of the reapportionment were intended.” The intent element proposed by the plaintiffs was, therefore, “meaningless,” and the Supreme Court’s decision in Vieth already had ruled out the more demanding standard of “predominant intent.” See 541 U.S. at 284-86, 124 S.Ct. 1769 (plurality opinion); id. at 308, 124 S.Ct. 1769 (Kennedy, J., concurring). The defendants levied two additional criticisms of the plaintiffs’ test. First, they noted that the plaintiffs’ “Demonstration Plan” was based on a counterfactual scenario and therefore failed to address concerns raised by some Justices about a standard which dealt with a “hypothetical state of affairs.” Second, they alleged that the EG is highly sensitive to “vote-switchers” in swing districts. Had voters in close (or competitive) elections voted for the other party, and had a few candidates of the other party won those seats, then the EG might be dramatically different. In their view, a plan that included such competitive districts could be found unconstitutional under the plaintiffs’ proposed standard. We denied the motion for summary judgment. We explained that judgment “as a matter of law would be premature because there [we]re factual disputes regarding the validity of plaintiffs’ proposed measurement.” We also noted that there was conflicting evidence on the “natural packing” of Democrats in Wisconsin. We further observed that the defendants’ arguments might serve as “a suggestion to alter the threshold of the plaintiffs’ test and, perhaps, shift the burdens of production or proof.” In particular, we left open the question of the requisite level of intent and directed the plaintiffs to “be prepared to present the strongest evidence that they have on this issue ... in order to meet even the most demanding intent requirement.” We therefore set the case for trial. D. Witnesses Testifying at Trial During the four-day trial, from May 24, 2016, through May 28, 2016, the parties presented their cases through eight witnesses. Some of the testimony of the witnesses involved in the passage of Act 43 has been set forth above, so it is not necessary to summarize it again here. An overview of the remaining testimony is set forth below. 1. William Whitford First to testify was William Whitford, one of the plaintiffs in this litigation and a resident of the 76th Assembly District. Professor Whitford testified to his longtime affiliation with the Democratic Party. He related that he consistently has voted for Democratic candidates, has made donations to Democratic Assembly candidates outside of his own district, has raised money on their behalf, and has donated to the Assembly Democratic Campaign Committee. According to Professor Whitford, given Wisconsin’s caucus system, “[t]he only practical way to accomplish [his] policy objectives is to get a majority of the Democrats in the Assembly and the Senate,” which is “virtually impossible under this apportionment [plan].” 2. Ronald Keith Gaddie Professor Gaddie was deposed by the plaintiffs on March 9, 2016, and a video of that deposition was admitted into evidence and played at trial. As explained in some detail above, Professor Gaddie testified that he was retained by Michael Best & Friedrich on April 11, 2011, to “serv[e] as an independent advisor on the appropriate racial and/or political make-up of legislative and congressional districts in Wisconsin.” In particular, Professor Gaddie took “the electoral data ... and constructed a regression analysis ... in order to create an estimate of the vote performance of every district.” He explained that this analysis “could be used to create a set of visual aids to demonstrate the partisan structure of Wisconsin politics.” As noted above, Professor Gaddie’s regression analysis was employed to confirm the validity of the composite measure developed by Foltz, Ottman, and Handrick. Professor Gaddie also used his regression analysis to assess each of the drafters’ proposed maps and to create “S” curves to illustrate how the Republican seat share would change based on changes in the party’s statewide vote share. In Professor Gaddie’s words, the “S” curves were “designed to tease out a potential estimated vote for the legislator in the district and then allow you to also look at that and say, okay, what if the Democrats have a good year? What if the Republicans have a good year? How does it shift?” At least some of the “S” curves were printed and kept in the map room at Michael Best & Fried-rich; in print form, the “S” curves were large enough to “cover half th[e] table.” 3. Adam Foltz Foltz worked as a legislative aide for Speaker Fitzgerald and served as one of the primary drafters of Act 43. One additional aspect of Foltz’s testimony at trial, however, is worthy of note. His testimony revealed a shortcoming in the drafters’ composite partisan measure. Specifically, the composite score likely was skewed to show a greater Republican advantage because of an error in the data for the 2006 Governor’s race (one of the components of the composite score). As a result of this error, the partisan estimates in the drafters’ spreadsheets were distorted and differed from the estimates reached by Professor Gaddie in his “S” curves. Foltz testified that he had not noticed this discrepancy at the time of drafting. He explained that, at the time, he “didn’t spend a whole lot of time with” Professor Gaddie so he “[did]n’t really understand the nuts and bolts” of the “S” curves. 4. Tad Ottman Ottman testified to his involvement in the drafting and passage of Act 43. 5. Kenneth Mayer Kenneth Mayer, a professor of political science at the University of Wisconsin, served as an expert witness for the plaintiffs. His ultimate goal was to design an alternative districting plan to Act 43 “that had an efficiency gap as low to zero as I could get it” while also complying with traditional districting criteria to the same extent as Act 43. He first created a regression model that estimated partisanship for each geographic area, so that he could compare his plan to Act 43. To ensure the model was accurate, Professor Mayer compared the predictions made by his regression model to the actual results in 2012. Once he was confident in his model, Professor Mayer “used a GIS redistricting program called Maptitude ... to ... complete the task of actually drawing the Assembly district map.” Professor Mayer’s alternative “Demonstration Plan” yields a 2.2% EG in favor of the Republicans, compared to an 11.69% EG yielded by Act 43. According to Professor Mayer, “[o]n all constitutional requirements, the Demonstration Plan is comparable to Act 43.” On cross-examination, however, the defendants pointed out that Professor Mayer did not take account of incumbents when drawing the plan. As a result, his plan paired a greater number of incumbents than Act 43, including one pairing in a majority-minority district. Further, Professor Mayer had not drawn any Senate districts, and therefore had not taken account of disenfranchisement. In addition to discussing the Demonstration Plan, Professor Mayer responded to points made by the defendants’ experts in their reports. Specifically, Professor Mayer testified that he had conducted a sensitivity analysis to address concerns about the effect of “wave” elections—elections that dramatically favor one party—on the EG calculations for both the Demonstration Plan and Act 43. He first looked over the last twenty years of elections in Wisconsin and found the greatest and smallest statewide vote shares for each party. Using these vote shares as the likely electoral spectrum, he performed a swing analysis where the Democrats received an additional 3% of the statewide vote (compared to their 2012 share) and the Republicans received an additional 5% of the statewide vote (again compared to their 2012 share) “to see what effect that would have on [his] efficiency gap calculations for the Demonstration Plan.” Professor Mayer’s analysis revealed that the Demonstration Plan’s EG remained below 4%, regardless of the swing. Act 43’s EG, however, increased during a Democratic swing but significantly decreased during a Republican swing. Professor Mayer noted that this is because “we’ve swung the Republican vote percentage up to 54 percent” but “[t]he number of [Republican] seats doesn’t change.” In Professor Mayer’s view, the result “is a confirmation that the bias in Act 43 is about the maximum that you can get.” 6. Simon Jackman Simon Jackman, a professor of political science and statistics at Stanford University, also served as an expert witness for the plaintiffs. Professor Jackman primarily testified about the reliability and practicability of the EG. He conducted a survey of 786 state legislative elections (under 206 different districting plans) in the United States between 1972 and the present day, in order to ascertain whether there was a baseline EG which should “trigger scrutiny” and also to compare Act 43 to other redistricting plans. Professor Jackman sought to determine how much the EG varied from election year to election year, and whether a dis-tricting plan had any impact on that EG. Professor Jackman presented a “scatter-plot,” which graphed the relationship between the EG in the first election year of a redistricting plan (set forth on the x axis) and the average EG over the lifetime of the plan (set forth on the y axis). He found a “relatively strong predictive relationship,” meaning that a high EG in the first year of a redistrieting plan likely means that the EG will remain high for the lifetime of the plan. Based on his research, Professor Jack-man proposed that an EG of 7% or higher should be legally significant: I arrived at 7 percent because that seemed to be a reasonable threshold for saying yes, if the first election under a plan produces an efficiency gap score at least that big, then you can be confident now that you’ve seen not just a one-off, but something that’s going to persist over the life of the plan as a signal of—a reliable signal as to the set of efficiency gap scores and the average efficiency gap score you might see if the plan were allowed to run. In other words, an EG of 7% in favor of one party in the first election year of a plan almost certainly means that the EG will favor that same party in each subsequent election year under that plan. Professor Jackman noted that the EGs for the 2012 and 2014 races in Wisconsin— 13% and 10%, respectively—were particularly high by historical levels. The EG in 2012 was, according to Professor Jackman, “among the largest scores we’ve seen anywhere” and “in the top 3 percent in terms of magnitude.” Act 43’s average EG ranked fifth out of the 206 plans that Professor Jackman surveyed. He testified that he was “virtually certain” that “Act 43 will exhibit a large and durable advantage in favor of Republicans over the rest of the decade.” 7. Sean Trende Sean Trende, Senior Elections Analyst for the website RealClearPolitics, served as an expert witness for the defendants. Mr. Trende primarily testified on the political geography of Wisconsin and its potential effect on the EG. Mr. Trende explained that, as a general matter, political geography of the United States currently favors Republicans. In his view, the Democratic coalition has contracted geographically and is now concentrated heavily in urban areas. This concentration, in turn, has hurt the Democratic Party in congressional elections, which tend to favor parties with wider geographic reach. Mr. Trende also testified to the political geography of Wisconsin itself, which he analyzed using a measure called the “partisan index” (“PI”). The purpose of the PI is “to determine the partisan lean of political units,” in order to “compare results across elections.” Mr. Trende explained that the county and ward PI values within Wisconsin have shifted such that the Democratic Party’s influence was strengthening in areas “that already leaned Democratic,” but was contracting geographically. Mr. Trende then applied his PI analysis to Wisconsin’s wards in what he referred to as a “nearest neighbor” analysis, which assessed the median distance between heavily Democratic wards compared to the median distance between heavily Republican wards. From this analysis, Mr. Trende concluded that it has “become[] progressively harder to draw ... Democratic districts elsewhere in the state,” which in his view explained at least some of the EG. However, he did not determine exactly how much of the EG was attributable to geography. 8. Nicholas Goedert Nicholas Goedert, a visiting professor of political science at Lafayette College, was retained by the defendants to offer opinions on using the EG to measure partisan gerrymandering. Professor Goedert’s main objection to the EG was its perceived volatility. In Professor Goedert’s view, “wave elections are the norm,” meaning that “much more often than not one party wins by 5 percent or more” of the vote. Therefore, relying on an EG from one election year, which might have taken place during a close election, might not be reliable. Professor Goedert opined that, “at a very minimum, ... you need to have some sort of robust sensitivity testing that would be codified if you were going to use the efficiency gap in any way.” Professor Goedert also raised a series of policy concerns. First, he pointed out that the EG measure arguably rests on a “2-to-1” vote-to-seats ratio and therefore a certain standard of proportionality. He also noted that there are “normatively good reasons why a state might cho[ose] to draw a map in a certain way and even under these normatively good reasons we could and actually do observe very high efficien[cy] gaps.” For example, Professor Goedert noted that some states may wish to create a more proportional system or encourage competitive elections. In his view, states might be discouraged from pursuing these policy goals if the court adopted the EG as the standard for partisan gerrymandering. E. Post-Trial Briefing Both parties filed post-trial briefs, which summarized their views of the case in light of the evidence presented at trial. The plaintiffs contended that they satisfied them proposed three-part test by proving discriminatory intent, discriminatory effect, and an absence of a justification for that effect. On intent, the plaintiffs focused in particular on the alternative maps that the drafters rejected, the “S” curves drawn by Professor Gad-die, and memos written by Foltz and Ottman. On effect, the plaintiffs stressed that the EG was not only likely to favor Republicans for the lifetime of the plan, but that it also was likely to stay relatively high. The plaintiffs also highlighted the sensitivity testing that had been conducted by Professors Jackman and Mayer. On justification, the plaintiffs pointed out that the previous Assembly maps in Wisconsin, the alternative plans drafted by the defendants, and Professor Mayer’s Demonstration Plan all exhibited lower EGs while arguably complying as well with traditional districting criteria. In response, the defendants contended that “a plan that complies with all neutral districting criteria, and whose efficiency gap is consistent with prior court-drawn plans” cannot be unconstitutional. The defendants noted that Act 43’s districts were congruent, compact, and fairly equal in population. Further, much of the secrecy surrounding Act 43’s enactment was consistent with how bills typically are enacted in Wisconsin. The defendants also pointed to evidence that the political geography in Wisconsin favors Republicans, which they contend explains the trend in EGs towards that party over the past two decades. In the defendants’ view, this evidence also illustrates the unreliability of the EG. The defendants concluded that the plaintiffs had not presented enough of a reason for a court to intervene in the redistrieting process. We express our appreciation to both parties for their thorough and informative presentation, and now turn to the legal principles that must guide our analysis of the case. Ill THE LEGAL LANDSCAPE The plaintiffs’ claim is that Act 43 violates their First and Fourteenth Amendment rights because it discriminates against Democratic voters by diminishing the strength of their votes in comparison to their Republican counterparts. We note, as a prefatory matter, that we have acknowledged, throughout this litigation, that the plaintiffs’ standing to maintain a cause of action is a threshold issue. See, e.g., Tierney v. Advocate Health and Hosps. Corp., 797 F.3d 449, 450 (7th Cir. 2015). Indeed, in our disposition of the defendants’ motion to dismiss, we addressed extensively standing and “conclude[d] that plaintiffs’ alleged injury [wa]s sufficiently concrete and particularized under current law-to satisfy Lujan [v. Defenders of Wildlife, 504 U.S. 555, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992),] with respect to a statewide challenge to the districting plan.” “We reaeh[ed] the same conclusion with respect to [Lujan’s] second and third elements of standing, which are causation and redressability.” We noted, though, that the “defendants [we]re free to raise this issue again on a more developed record.” Lujan explains that, because the elements of standing “are not mere pleading requirements but rather an indispensable part of the plaintiffs case, each element must be supported in the same way as any other matter on which the plaintiff bears the burden of proof, i.e., with the manner and degree of evidence required at the successive stages of the litigation.” 504 U.S. at 561, 112 S.Ct. 2130. Our assessment of the evidence, as well as our elucidation of the political gerrymandering cause of action, therefore will inform our standing analysis. Consequently, we postpone a plenary discussion of standing until we fully have set forth the evidence as well as the constitutional standard. As a precursor, however, wé conclude that the plaintiffs have established a concrete and particularized injury: “[a]s a result of the statewide partisan gerrymandering, Democrats do not have the same opportunity provided to Republicans to elect representatives of their choice to the Assembly. As a result, the electoral influence of plaintiffs and other Democratic voters statewide has been unfairly [and] disproportionately ... reduced” for the life of Act 43. Additionally, the plaintiffs have shown causation: Act 43 was designed with the purpose of solidifying Republican control of the legislature for the decennial period and, indeed, has had that effect. Finally, the plaintiffs have established that their injury is re-dressable: adopting a different statewide districting map would redress the constitutional violation by removing the state-imposed impediment on Democratic voters. In resolving the plaintiffs’ claim, we face a significant analytical problem. Although the Supreme Court’s political gerrymandering cases establish that “an excessive injection of politics is unlawful,” Vieth, 541 U.S. at 293, 124 S.Ct. 1769 (plurality opinion) (emphasis removed), the Court has not come to rest on a single, judicially manageable or discernible test for determining when the line between “acceptable” and “excessive” has been crossed. Indeed, a signature feature of these cases is that no single opinion has garnered a majority of the Court. But the absence of a well-trodden path does not relieve us of the obligation to render a decision. True, we cannot anticipate that the Court will alter course from the decisional law, however sparse, that currently exists. See State Oil Co. v. Khan, 522 U.S. 3, 20, 118 S.Ct. 275, 139 L.Ed.2d 199 (1997) (noting that lower courts should apply outstanding precedent until explicitly overruled by the Supreme Court). Nor can we cobble together the opinions of the various Justices who have written on the matter and call the resulting amalgam binding precedent. See Rodriguez de Quijas v. Shearson/American Exp., Inc., 490 U.S. 477, 484, 109 S.Ct. 1917, 104 L.Ed.2d 526 (1989) (noting that lower courts should follow precedent despite expressed dissatisfaction by various members of the Court until the precedent is overruled explicitly). Nevertheless, understanding that we are in an area where the navigational signs are not yet well-placed, we must decide the case before us and satisfy our “duty ... to say what the law is,” Marbury v. Madison, 5 U.S. (1 Cranch) 137, 177, 2 L.Ed. 60 (1803), or at least what we believe it to be. We begin by examining the cases that set forth the constitutional principles which later informed the Court’s political gerrymandering decisions. A. The Foundational Case Law 1. Over half a century ago, the Supreme Court recognized that the constitutionality of legislative apportionments is governed by the Equal Protection Clause of the Fourteenth Amendment. See Reynolds v. Sims, 377 U.S. 533, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964). Reynolds was not a political gerrymandering case, but addressed allegations that an outdated apportionment scheme resulted in “serious discrimination with respect to the allocation of legislative representation” in violation of the Equal Protection Clause. Id. at 540, 84 S.Ct. 1362. Nevertheless, the Supreme Court spoke to the importance and nature of the right to vote in terms that also inform our consideration of the plaintiffs’ claims. The Court first observed that the right to vote “is a fundamental matter in a free and democratic society. Especially since the right to exercise the franchise in a free and unimpaired manner is preservative of other basic civil and political rights, any alleged infringement of the right of citizens to vote must be carefully and meticulously scrutinized.” Id. at 561-62, 84 S.Ct. 1362. The Court explained that “[m]ost citizens” exercise their “inalienable right to full and effective participation in the political process” by voting for their elected representatives. Id. at 565, 84 S.Ct. 1362. “Full and effective participation by all citizens in state government requires, therefore, that each citizen have an equally effective voice in the election of members of his state legislature.” Id. Moreover, the concept of equal protection has been traditionally •viewed as requiring the uniform treatment of persons standing in the same relation to the governmental action questioned or challenged. With respect to the allocation of legislative representation, all voters, as citizens of a State, stand in the same relation regardless of where they live. Id. The Court explained, however, that the requirement of equal treatment was not limited to where a voter resided. Instead, “[a]ny suggested criteria for the differentiation of citizens are insufficient to justify any discrimination, as to the weight of their votes, unless relevant to the permissible purposes of legislative apportionment.” Id. (emphasis added). The Court therefore concluded that, [s]ince the achieving of fair and effective representation for all citizens is con-cededly the basic aim of legislative apportionment, ... the Equal Protection Clause guarantees the opportunity for equal participation by all voters in the election of state legislators. Diluting the weight of votes because of place of residence impairs basic constitutional rights under the Fourteenth Amendment just as much as invidious discriminations based upon factors such as race or economic status. Id. at 565-66, 84 S.Ct. 1362 (citations omitted). Reynolds therefore establishes that, in electing state representatives, the votes of citizens must be weighted equally. If an apportionment scheme violates the principle of one-person, one-vote, it must be justified on the basis of other, permissible, legislative considerations. 2. The Court soon had the opportunity to apply the principles set forth in Reynolds to allegations of vote-dilution brought by racial minorities. In Fortson v. Dorsey, 379 U.S. 433, 85 S.Ct. 498, 13 L.Ed.2d 401 (1965), the Court considered the constitutionality of an apportionment scheme which included traditional single-member districts and multimember districts, where citizens reside in a comparatively larger district and vote for multiple representatives. Voters alleged that these multimem-ber districts were “defective because county-wide voting in multi-district counties could, as a matter of mathematics, result in the nullification of the unanimous choice of the voters of a district.” Id. at 437, 85 S.Ct. 498. The district court granted summary judgment to the plaintiffs