OPINION
PER CURIAM:Plaintiffs, individual voters registered in the State of Arizona, challenge the map drawn for state legislative districts by the Arizona Independent Redistricting Commission for use starting in 2012, based on the 2010 census. They argue that the Commission underpopulated Democrat-leaning districts and overpopulated Republican-leaning districts for partisan reasons, in violation of the Fourteenth Amendment’s one-person, one-vote principle. The Commission denies that it was driven by partisanship, explaining that the population deviations were driven by its efforts to comply with Section 5 of the Voting Rights Act. We conclude that the population deviations were primarily a result of good-faith efforts to comply with the Voting Rights Act, and that even though partisanship played some role in the design of the map, the Fourteenth Amendment challenge fails.1
*1047The one-person, one-vote requirement of the Equal Protection Clause of the Fourteenth Amendment does not require that legislative districts have precisely equal population, but provides that divergences must be “based on legitimate considerations incident to the effectuation of a rational state policy.” Reynolds v. Sims, 377 U.S. 533, 579, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964). The majority of the overpopulated districts in the map drawn by the Commission were Republican-leaning, while the majority of the underpopulated districts leaned Democratic. Plaintiffs’ complaint alleged that this correlation was no accident, that partisanship drove it, and that partisanship is not a permissible reason to deviate from population equality in redistricting.
The Commission does not argue that the population deviations came about by accident, but it disputes that the motivation was partisanship. Most of the underpopulated districts have significant minority populations, and the Commission presented them to the Department of Justice as districts in which minority groups would have the opportunity to elect candidates of their choice. Section 5 of the Voting Rights Act required that the Commission obtain preclearance from the Department before its plan went into effect. To obtain preclearance, the Commission had to show that any proposed changes would not diminish the ability of minority groups to elect the candidates of their choice. The Commission argues that its effort to comply with the Voting Rights Act drove the population deviations.
For the purpose of this opinion, we assume without deciding that partisanship is not a legitimate reason to deviate from population equality. We find that the primary factor driving the population deviation was the Commission’s good-faith effort to comply with the Voting Rights Act and, in particular, to obtain preclearance from the Department of Justice on the first try. The commissioners were aware of the political consequences of redistricting, however, and we find that some of the commissioners were motivated in part in some of the linedrawing decisions by a desire to improve Democratic prospects in the affected districts. Nonetheless, the Fourteenth Amendment gives states some degree of leeway in drawing their own legislative districts and, because compliance with federal voting rights law was the predominant reason for the deviations, we conclude that no federal constitutional violation occurred.
We do not decide whether any violations of state law occurred. Though plaintiffs have alleged violations of state law and the Arizona Constitution, we decided early in the proceedings and announced in a prior order that Arizona’s courts are the proper forum for such claims. We discuss that subject further below, at 1065-66. We express no opinion on whether the redistricting plan violated the equal population clause of the Arizona Constitution, whether the Commission violated state law in *1048adopting the grid map with population variations rather than strict population equality, or whether state law prohibits adjusting legislative districts for partisan reasons. All that we consider is whether a federal constitutional violation occurred.
At trial, plaintiffs focused on three districts that they argued were not true Voting Rights Districts and therefore could not justify population deviations: Districts 8, 24, and 26. Accordingly, this opinion largely focuses on the population shifts associated with the creation of these three districts.
1. Course of Proceedings
Plaintiffs filed this action on April 27, 2012, and subsequently filed a First Amended Complaint. This three judge district court was convened pursuant to 28 U.S.C. § 2284(a). Plaintiffs sought a declaration that the final legislative map violated both the Equal Protection Clause of the Fourteenth Amendment of the U.S. Constitution and the equal population requirement of the Arizona Constitution, an injunction against enforcing the map, and a mandate that the Commission draw a new map for legislative elections following the 2012 elections. Originally, not only was the Commission a defendant in this action, but so too were each of the five commissioners in their official capacities.2
Defendants moved to dismiss the complaint for failure to state a claim. In a reasoned order, we denied the motion. Plaintiffs then filed a Second- Amended Complaint.
Prior to trial, the parties filed several motions that the court summarily disposed of on February 22, 2013. First, defendants moved to stay the case pending the resolution of state-law claims in state court, which we denied. Defendants also moved for a protective order on the basis of legislative privilege, which we denied. Finally, defendants moved for judgment on the pleadings, asking for dismissal of the individual commissioners as defendants and for dismissal of plaintiffs’ claim for relief under the equal population requirement of the Arizona Constitution. We granted this motion, dismissing the individual commissioners from the suit and dismissing plaintiffs’ second claim for relief. We explain the bases for our rulings on these motions later in this opinion, at 1063-71.
Starting March 25, 2013, we presided over a five-day bench trial. Among other witnesses, all five commissioners testified.
II. Findings of Fact
Most of the factual findings below, based in large part on transcripts of public hearings and other documents in the public record, were not disputed at trial. Rather, what was most controverted was what inferences about the Commission’s motivation we should draw from the largely undisputed facts. We discuss that issue, whether and to what extent partisanship motivated the Commission, at the end of this section, at 1063-71.
To the extent any finding of fact should more properly be designated a conclusion of law, it should be treated as a conclusion of law. Similarly, to the extent any conclusion of law should more properly be designated a finding of fact, it should be treated as a finding of fact.
A. The Approved Legislative Redistricting Plan
The first election cycle using the legislative map drawn by the Commission took *1049place in 2012. Arizona has thirty legisla-tire districts, each of which elects two representatives and one senator. Ariz. Const, art. IV, pt. 2, § 1. The following chart summarizes pertinent electoral re-suits and population statistics for the Corn-mission’s 2012 legislative map, which we explain in greater detail below.
[[Image here]]
Figure 1. 2012 Legislative Map Statistics.
In the 2012 elections, Republicans won a total of 36 out of the 60 house seats, winning both seats in 17 districts and 1 seat in 2 districts. Democrats won the remaining 24 house seats, winning 2 seats in 11 districts and 1 seat in 2 districts. Republicans won 17 out of 30 senate seats, and Democrats won the remaining 13. The Democratic senate candidate narrowly won in District 8, but the Republican candidate might have won if not for the presence of a Libertarian candidate in the *1050race.3 In all, 16 districts elected only Republicans to the state legislative houses, 11 districts elected only Democrats, and 3 districts elected a combination of Republicans and Democrats.
Ideal population is the average per-district population, or the population each district would have if population was evenly distributed across all districts. Of the 16 districts that elected only Republicans to the state legislature, 15 were above the ideal population and 1 was below. Of the 11 districts that elected only Democrats to the state legislature, 2 were above the ideal population and 11 were below. District 8 was below ideal population, and the other 2 districts that elected legislators from both parties were above ideal population.
Of the 10 districts the Commission presented to the Department of Justice as districts in which minority candidates could elect candidates of their choice, or “ability-to-eleet districts,” all 10 only elected Democrats to the state legislature in 2012. Nine out of ten of these ability-to-elect districts were below the ideal population, and one was above.
Of the 9 districts presented to the Department of Justice as districts in which Hispanics could elect a candidate of their choice, all but District 24 elected at least one Hispanic candidate to the state legislature in the 2012 elections. In District 26, only one of the three legislators elected in 2012 was of Hispanic descent. Of the 27 state legislators elected in the purported ability-to-elect districts, 16 were of Hispanic descent.
District 7 was presented to the Department of Justice as a district in which Native Americans could elect candidates of their choice, and it elected Native American candidates in all three of its state legislative races.
Maximum population deviation refers to the difference, in terms of percentage deviation from the ideal population, between the most populated district and the least populated district in the map. In the approved legislative map, maximum population deviation was 8.8 percent; District 12 had the largest population, at 4.1 percent over the ideal population, and District 7 had the smallest population, at 4.7 percent under the ideal.
B. Formation of the Commission
In 2000, Arizona voters amended the state constitution by passing Proposition 106, an initiative removing responsibility for congressional and legislative redistricting from the state legislature and placing it in the newly established Independent Redistricting Commission. See Ariz. Const, art. IV, pt. 2, § 1(3). Five citizens serve on the Commission, consisting of two Republicans, two Democrats, and one unaffiliated with either major party. See id. § l(3)-(5). Selection of the commissioners begins with the Arizona Commission on Appellate Court Appointments, which interviews applicants and creates a slate of ten Republican candidates, ten Democratic candidates, and five independent or unaffiliated candidates. See id. § l(4)-(5). Four commissioners are appointed from the party slates, one by each of the party leaders from the two chambers of the legislature. See id. § 1(6). Once appointed, those four commissioners select the fifth commissioner from the slate of unaffiliated candidates, and the fifth commissioner also serves as the commission chair. Id. § 1(8).
*1051Pursuant to these requirements, Republican commissioners Scott Freeman and Richard Stertz were appointed by the Speaker of the House and the President of the Senate, respectively, and Democratic commissioners Jose Herrera and Linda McNulty were appointed by the House Minority Leader and Senate Minority Leader, respectively. Commissioners Freeman, Stertz, Herrera, and McNulty then interviewed all five candidates on the unaffiliated slate.
In his interview notes, Commissioner Stertz noted his concerns with the liberal leanings of most of the candidates on the unaffiliated list. For example, he wrote that Kimber Lanning’s fundraising efforts were almost all for Democrats, and that her Facebook page indicated a fondness for Van Jones.4 Paul Bender, another candidate, served on the board of the ACLU. Margaret Silva identified Cesar Chavez as her hero, and her Facebook profile picture featured her alongside Nancy Pelosi, the Democratic leader in the U.S. House of Representatives. Ray Bla-dine was his first choice for the position, whom Stertz described as balanced despite Bladine’s former tenure as chief of staff for a Democratic mayor.
In a public meeting, the four commissioners unanimously selected Colleen Mathis as the fifth commissioner and chairwoman. In his interview notes Commissioner Stertz described her as balanced, though noting that she and her husband had supported Democratic candidates. Mathis and her husband had also made contributions to Republican candidates.
C. Selection of Counsel and Mapping Consultant
The Commission has authority to hire legal counsel to “represent the people of Arizona in the legal defense of a redistricting plan,” as well as staff and consultants to assist with the mapping process. Ariz. Const, art. IV, pt. 2, §§ 1(19), (20). The selection of the Commission’s counsel and mapping consultant sparked public controversy, and plaintiffs argue that the process reflected a partisan bias on the part of Chairwoman Mathis.
The previous Commission, after the 2000 census, had retained a Democratic attorney and a Republican attorney. Chairwoman Mathis expressed interest in hiring one attorney instead of two, as the counsel hired would represent the entire Commission. The other four commissioners preferred to hire two attorneys with different party affiliations, however. That is what the Commission decided to do.
The Commission used the State Procurement Office to help retain counsel and interviewed attorneys from six law firms. Among the interviewees were the two attorneys who had worked for the previous Commission: Lisa Hauser, an attorney with the firm of Gammage & Burnham and a Republican, and Michael Mandell, an attorney with the Mandell Law Firm and a Democrat. Other attorneys interviewed by the Commission included Mary O’Grady, a Democrat with Osborn Maledon, and Joe Kanefield, a Republican with Ballard Spahr. Osborn Maledon and Ballard Spahr received the highest scores from the Commission based on forms provided by the State Procurement Office for use in the selection process. Nonetheless, Commissioner Herrera expressed a preference for retaining Mandell as Democratic counsel, and Commissioners Stertz and Freeman preferred Hauser and Gammage & Burnham as Republican counsel.
*1052In a public meeting, Commissioner Herrera moved to retain Osborn Maledon and Ballard Spahr at Chairwoman Mathis’s suggestion. Commissioner Herrera later explained that while Mandell was his first choice, Osborn Maledon and Ballard Spahr received the highest evaluation scores. Commissioner Freeman expressed his preference for Gammage & Burnham, and said he would give deference to the Democratic commissioners’ preference for Democratic counsel if they would do the same for the Republican commissioners. Commissioner Stertz then made a motion to amend, to instead retain the Mandell Law Firm and Gammage & Burnham. The amendment was defeated on a 2-3 vote, with Commissioners Stertz and Freeman voting for it and Commissioners Mathis, Herrera, and McNulty voting against. The motion to retain Osborn Maledon and Ballard Spahr carried with a 3-2 vote, with Commissioners Mathis, Herrera, and McNulty voting for the motion and Commissioners Stertz and Freeman voting against. The Commission thus selected a Republican attorney for whom neither of the Republican commissioners voted.
In selecting a mapping consultant, the Commission initially worked with the State Procurement Office. An applicant for the position had to submit, among other things, an explanation of its capabilities to perform the work, any previous redistricting experience, any partisan connections, and a cost sheet. In the initial round of scoring, each applicant was scored on a 1000-point scale. Each commissioner independently filled out a scoring sheet, which considered capability to do the work but not cost, rating each applicant on a 700-point scale. The State Procurement Office rated each applicant on a 300-point scale, 200 points of which evaluated the relative cost of the bid.
The Commission considered the first round of scoring, and then announced a short list of four firms that it would interview for the mapping consultant position. Those firms were Strategic Telemetry, National Demographics, Research Advisory Services, and Terra Systems Southwest. National Demographics, which had served as mapping consultant for the previous Commission, had received the highest score in the first round of evaluations.
The Commission interviewed the four selected firms in a public meeting. During the interview of the head of National Demographics, Commissioner Herrera expressed concern that there was a perception that the firm was affiliated with Republican interests. National Demographics had worked for both Democratic and Republican clients, though more Republicans than Democrats. In interviewing Strategic Telemetry, Commissioners Freeman and Stertz asked whether, because Strategic Telemetry had worked for a number of Democratic clients but no Republican clients, the firm would be perceived as biased.
After these interviews, the commissioners conducted a second round of scoring before selecting a firm. In this round of scoring, Commissioners Mathis, Herrera, and McNulty all gave Strategic Telemetry a perfect score. Strategic Telemetry came out of this round with the highest overall score. Prior to the public meeting in which the Commission voted to retain a mapping consultant, Chairwoman Mathis made a phone call to Commissioner Stertz and asked him to support the choice of Strategic Telemetry.
The Commission selected Strategic Telemetry as the mapping consultant on a 3-2 vote, with Commissioners McNulty, Herrera, and Mathis voting in favor, and Commissioners Freeman and Stertz voting against. Before the vote, Commissioners *1053Freeman and Stertz had expressed a preference for National Demographics.
At subsequent meetings, the Commission heard extensive criticism from members of the public about the selection of Strategic Telemetry. Much of the criticism related to the Democratic affiliations of the firm and to the fact that it was based out of Washington, D.C., rather than Arizona. Strategic Telemetry was founded primarily as a microtargeting firm, which uses statistical analyses of voter opinions to assist political campaigns. Ken Strasma, president and founder of Strategic Telemetry, considered himself a Democrat, as did most of the other employees of the firm. The firm had worked for Democratic, independent, and nonpartisan campaigns, but no Republican campaigns. While Strasma had redistricting experience in more than thirty states before he founded the firm in 2003, the firm itself had no statewide redistricting experience at the time of its bid, nor any redistricting experience in Arizona. Also making Strategic Telemetry a controversial choice was that it had submitted the most expensive bid to the Commission. All of this was known to the Commission when Strategic Telemetry was selected as the mapping consultant for the Commission and when Commissioners Mathis, Herrera, and McNulty each gave Strategic Telemetry a perfect score of 700 points during the second round of scoring.
D. The Grid Map
The Commission was required to begin the mapping process by creating “districts of equal population in a grid-like pattern across the state.” Ariz. Const, art. IV, pt. 2, § 1(14). The Commission directed its mapping consultant to prepare two alternative grid maps. Believing that the Arizona Constitution intended the Commission to begin with a clean slate, several commissioners expressed interest in having an element of randomness in the generation of the grid map. The Commission decided, after a series of coin flips, that the consultant would generate two alternative grid maps, one beginning in the center of the state and moving out counterclockwise, and the other with districts starting in the southeast corner of the state, moving inwards clockwise.
After the two maps were presented, the Commission voted to adopt the second alternative. The grid map selected had a maximum population deviation' — the difference between the most populated and least populated district — of 4.07 percent of the average district population.
E. Voting Rights Act Preclearance Requirement
During the redistricting cycle at issue, Arizona was subject to the requirements of Section 5 of the Voting Rights Act.5 Before a state covered by Section 5 can implement a redistricting plan, the state must prove that its proposed plan “neither has the purpose nor will have the effect of denying or abridging the right to vote on account of race or color.” 42 U.S.C. § 1973c(a).6 The state must either *1054institute an action with the U.S. District Court for the District of Columbia for a declaratory judgment that the plan has no such purpose or effect, or, as the Commission did here, submit the plan to the U.S. Department of Justice. If the Justice Department does not object within sixty days, the plan has been precleared and the state may implement it. See id.
A plan has an impermissible effect under Section 5 if it “would lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.” Reno v. Bossier Parish Sch. Bd., 520 U.S. 471, 478, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997). A redistricting plan leads to retrogression when, compared to the plan currently in effect, the new plan diminishes the ability of minority groups to “elect their preferred candidates of choice.” See id.; 42 U.S.C. § 1973c(b). There is no retrogression so long as the number of ability-to-elect districts does not decrease from the benchmark to the proposed plan. Texas v. United States, 887 F.Supp.2d 133, 157 (D.D.C.2012) (citing Abrams v. Johnson, 521 U.S. 74, 97-98, 117 S.Ct. 1925, 138 L.Ed.2d 285 (1997)), vacated and remanded, — U.S. -, 133 S.Ct. 2885, 186 L.Ed.2d 930 (2013) (remanding for further consideration in light of Shelby County v. Holder, — U.S. -, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013)).
A district gives a minority group the opportunity to elect the candidate of its choice not only when the minority group makes up a majority of the district’s population (a majority-minority district), but also when it can elect its preferred candidate with the help of another minority group (a coalition district) or white voters (crossover districts). Texas, 887 F.Supp.2d at 147-49. A minority group’s preferred candidate need not be a member of the racial minority. Cf. Ruiz v. City of Santa Maria, 160 F.3d 543, 552 (9th Cir.1998) (discussing minority candidates of choice for the purposes of Section 2 of the Voting Rights Act). “Ability to elect” properly refers to the ability to elect the preferred candidate of Hispanic voters from the given district, which is not necessarily the same thing as the ability to elect a Hispanic candidate from that district, though there is obvious overlap between those two concepts.
In determining the ability to elect in districts in the proposed and benchmark plan, the Department of Justice begins its review of a plan submitted for preclearance by analyzing the districts with current census data. 76 Fed.Reg. 7470, 7472 (Feb. 9, 2011). The analysis is a complex one relying on more than just census numbers, however, and does not turn on reaching a fixed percentage of minority population. Rather, the Department looks at additional demographic data such as group voting patterns, electoral participation, election history, and voter turnout. Id. at 7471; see also Texas, 887 F.Supp.2d at 150 (“There is no single, clearly defined metric to determine when a minority group has an ability to elect, so we use a multi-factored approach to determine when a coalition or crossover district achieves that ability.”).
Several aspects of the preclearance process encourage states to do more than the bare minimum to avoid retrogression. First, state officials do not know exactly what is required to achieve preclearance. As explained above, the Department of Justice relies on a variety of data in assessing retrogression, rather than assessing a fixed goal that states can easily ascertain. Bruce Cain, an expert in Vot*1055ing Rights Act compliance in redistricting who served as a consultant to the Commission following the 2000 census and was retained for this lawsuit by the current Commission, testified at trial that the lack of clear rules creates “regulatory uncertainty” that forces states “to be cautious and to take extra steps.”
Moreover, the preclearance process with respect to any particular plan is generally an opaque one. When the Department of Justice objects to a plan, the state receives an explanation of the basis for the objection. When the Department does not object, by contrast, the state receives no such information. In other words, the state does not know how many benchmark districts the Department believed there were nor how many ability-to-elect districts the Department concluded were in the proposed plan. Nor does it know whether the new plan barely precleared or could have done with fewer ability-to-elect districts.
Consultants and attorneys hired by a state to assist with the preclearance process may also tend to encourage taking additional steps to achieve preclearance. The professional reputation of a consultant gives him a strong incentive to ensure that the jurisdictions he advises obtain pre-clearance. The Commission, for example, asked applicants to serve as its mapping consultant whether they had previously worked with states in redistricting and whether those jurisdictions had succeeded in gaining preclearance on the first try.
These factors may work together to tilt the board somewhat because they encourage a state that wants to obtain preclearance to overshoot the mark, particularly if it wants its first submission to be approved. Because it is not clear where the Justice Department will draw the line, there is a natural incentive to provide a margin of error or to aim higher than might actually be necessary. Attorneys and consultants, aware that their professional reputations may be affected, can be motivated to push in that direction.
The Arizona Commission early in the process identified obtaining preclearance on its first attempt as a priority. All of the commissioners, Democrats and Republicans alike, shared this goal. In prior decades, Arizona had never obtained pre-clearance from the Department of Justice for its legislative redistricting plan based upon its first submission. The Commission was aware that, among other consequences, failure to preclear would make Arizona ineligible to bail out as a Section 5 jurisdiction for another ten years. See 42 U.S.C. § 1973b(a)(l). Although the Commission considered and often adjusted lines to meet other goals, it put a priority on compliance with the Voting Rights Act and, in particular, on obtaining preclearance on the first attempt.
F. The Draft Map
After adopting a grid map, the Commission was directed by the Arizona Constitution to adjust the map to comply with the United States Constitution and the federal Voting Rights Act. Ariz. Const, art. IV, pt. 2, § 1(14). It was also instructed to adjust the map, “to the extent practicable,” to comply with five other enumerated criteria: (1) equality of population between districts; (2) geographic compactness and contiguity; (3) respect for communities of interest; (4) respect for visible geographic features, city, town and county boundaries, and undivided census tracts; and (5) competitiveness, if it would “create no significant detriment to the other goals.” Id. The map approved by the Commission after the first round of these adjustments was only a draft map, which was required to undergo public comment and a further round of revisions before final approval. Id. § 1(16).
*1056Before beginning to adjust the grid map, the Commission received presentations on the Voting Rights Act from its attorneys, its mapping consultant, and its Voting Rights Act consultant Bruce Adelson. Adelson previously worked for the Department of Justice, where he led the team that had reviewed and objected to the first legislative map submitted by Arizona for preclearance in 2002. Adelson gave the Commission an overview of the preclearance process. He explained that determining whether a minority population had the ability to elect was a complex analysis that turned on more than just the percentage of minorities in a district. He explained, for example, that in reviewing Arizona’s submission from the prior decade, the Department had found a district where it concluded that minorities had an ability to elect even though they made up only between 30 and 40 percent of the population. Adelson informed the Commission at that time that he believed the 2002 map that was ultimately approved had nine districts in which minorities had an ability to elect their preferred candidates. Because the preclearance process focused on making sure there was no retrogression, that number was the benchmark, meaning that the new plan had to achieve at least the same number of ability-to-elect districts.
One of the most important factors the Department of Justice considers in determining the ability to elect in a district is its level of racial polarization, which is a measure of the voting tendencies of whites and minorities in elections pitting a white candidate against a minority candidate. A racial polarization study is a statistical analysis of past election results to determine the level of racial polarization in a district. When it first started considering potential benchmark districts, the Commission did not have any formal racial polarization analysis at its disposal and relied primarily on demographic data from the 2010 census. The Commission eventually retained Professor Gary King, a social scientist at Harvard University recommended by the Commission’s counsel, to conduct a racial polarization analysis.
Until the Commission had a formal racial polarization analysis, it often used what it called the “Cruz Index” to assess whether voters in an area might support a Hispanic candidate. Devised by Commissioners McNulty and Stertz, the Cruz Index used data from the 2010 election for Mine Inspector, a statewide race pitting Joe Hart, a Republican, non-Hispanie white (or Anglo) candidate, against Manuel Cruz, a Democrat, Hispanic candidate. The Cruz Index, sometimes described by commissioners and staff as a “down and dirty” measure, was not intended to be the Commission’s only analysis of cohesion in minority voting in proposed districts, but rather a rough proxy until the Commission had formal racial polarization analysis. In the end, however, the voting pattern estimates derived from the Cruz Index wound up corresponding closely to the voting pattern estimates King derived from his formal statistical analysis.
To explore possible adjustments to the grid map, the commissioners could either direct the mapping consultant to create a map with a certain change or use mapping software to make changes themselves. They referred to these maps as “what if’ maps because the maps simply showed possible line changes that the Commission might choose to incorporate into the draft map. Willie Desmond was the Strategic Telemetry employee with primary responsibility for assisting commissioners with the mapping software or creating “what if’ maps at their direction.
The Commission originally operated on the assumption that it had to create nine ability-to-elect districts, based on Adel-*1057son’s report that there were nine benchmark districts. As a result, the earliest “what if’ maps focused on creating nine minority ability-to-elect districts. Commissioner Freeman, for example, directed Desmond to create several maps that would create nine ability-to-elect districts.
Soon, however, the Commission began considering the possibility that there might be ten benchmark districts. Counsel advised that there were some districts without a majority-minority population that had a history of electing minority candidates, such as District 23 from the 2002 legislative map. Counsel further explained that, even though there were seven majority-minority benchmark districts and two to three other districts where minorities did not make up the majority, they nonetheless might be viewed as having the ability to elect. Because it was uncertain how many benchmark and ability-to-elect districts the Department of Justice would determine existed, counsel advised that creating ten districts would increase the odds of getting precleared on the first attempt.
The Commission worked to make Districts 24 and 26 ones in which, despite lacking a majority of the population, His-pamos could elect candidates of their choice. At this point, the Commission was still relying on the Cruz Index to predict minority voting patterns in proposed districts. As the Commission explored shifting boundaries to create ability-to-elect districts, their mapping consultant apprised the Commission of the effects of the shifts on various statistics, such as minority voting population, the Cruz Index, and the deviation from average district population. Counsel advised the Commission that some population disparity was permissible if it was a result of compliance with the Voting Rights Act.
On October 10, 2011, the Commission approved a draft legislative map on a 4-1 vote, with all but Commissioner Stertz voting in favor of the map. That map had ten districts identified by the Commission as minority ability-to-elect districts.
G. The Effort to Remove Chairwoman Mathis
The Arizona Constitution prescribes at least a thirty-day comment period after the adoption of the draft map. Ariz. Const, art. IV, pt. 2, § 1(16). The Commission did not begin working on the final map until late November, however, because of a delay resulting from an effort to remove Chairwoman Mathis from the Commission.
On October 26, Governor Janice Brewer sent a letter to the Commission alleging it had committed “substantial neglect of duty and gross misconduct in office” for, among other things, the manner in which it selected the mapping consultant. On November 1, the Governor’s office informed Chairwoman Mathis that it would remove her from the Commission for committing gross misconduct in office, conditioned upon the concurrence of two-thirds of the Arizona Senate. The Arizona Constitution permits the governor to remove a member of the Commission, with concurrence of two-thirds of the Senate, for “substantial neglect of duty” or “gross misconduct in office.” Ariz. Const, art. IV, pt. 2, § 1(10). After the Senate concurred in the removal of Chairwoman Mathis in a special session, the Commission petitioned the Arizona Supreme Court for the reinstatement of Chairwoman Mathis on the basis that the Governor had exceeded her authority under the Arizona Constitution. Ariz. Indep. Redistricting Comm’n v. Brewer, 229 Ariz. 347, 275 P.3d 1267, 1270 (2012). On November 17, that court ordered the reinstatement of Chairwoman Mathis, conclud*1058ing that the Governor did not have legal cause to remove her. Id. at 1268, 1276-78.
H. The Final Map
On November 29, the Commission began working to modify the draft map to create the final map it would submit to the Department of Justice. Because of the delay caused by the effort to remove Chairwoman Mathis, the Commission felt under pressure to finalize its work in time to permit election officials and prospective candidates to prepare for the 2012 elections, knowing that the preclearance process would also take time.
The Commission received a draft racial polarization voting analysis prepared by King and Strasma. According to the draft analysis, minorities would be able to elect candidates of their choice in all ten proposed ability-to-elect districts in the draft map.
The Commission received advice from its attorneys and consultants as to the importance of presenting the Department of Justice with at least ten ability-to-elect districts. Adelson said that, based on the information he had received since his earlier assessment, he believed the Department would conclude that there were ten benchmark districts. He also emphasized that, due to the uncertainty in determining what constitutes a benchmark district, the Department might determine there were more benchmark districts than what the Commission had concluded. Counsel advised the Commission that it would be “prudent to stay the course in terms of the ten districts that are in the draft map and look to ... strengthen them if there is a way to strengthen them.”
The Commission also received advice that it could use population shifts, within certain limits, to strengthen these districts. Adelson advised the Commission that un-derpopulating minority districts was an acceptable tool for complying with the Voting Rights Act, so long as the maximum deviation remained within ten percent. According to Adelson, underpopulating districts to increase the proportion of minorities was an “accepted redistricting tool” and something that the Department of Justice looked at favorably when assessing compliance with Section 5. According to Strasma, underpopulation could strengthen the districts in several ways. First, it could increase the percentage of minority voters in a district. Second, it could account for expected growth in the Hispanic districts, which might otherwise become overpopulated in the decade following the implementation of a new map.
The Commission directed Strasma and Adelson to look for ways to strengthen the ability-to-elect districts and report back. At a subsequent meeting, Strasma, Adel-son, and Desmond presented a number of options for improving the districts along with the trade-offs associated with those changes. Strasma identified Districts 24 and 26 in particular as districts that might warrant further efforts to strengthen the minority ability to elect. Doing so would increase the likelihood that the Department of Justice would recognize those districts as ability-to-elect districts and thus the likelihood that the plan would obtain preclearance.
The Commission adopted a number of changes to Districts 24 and 26, including many purportedly aimed at strengthening the minority population’s ability to elect. Between the draft map and final map, the Hispanic population in District 24 increased from 38.6 percent to 41.3 percent, and the Hispanic voting-age population increased from 31.8 percent to 34.1 percent. In District 26, the Hispanic population increased from 36.8 percent to 38.5 percent, and the Hispanic voting-age population increased from 30.4 percent to 32 percent.
*1059A consequence of these changes was an increase in population inequality. District 24’s population decreased from 0.2 percent above the ideal population to 3 percent below. District 26’s population increased from 0.1 percent above the ideal population to 0.3 percent above.
Commissioner McNulty asked Desmond to explore possibilities for making either District 8 or 11 more competitive. Desmond presented an option to the Commission that would have made District 8 more competitive. The Republican commissioners expressed opposition to the proposed change. Commissioner Stertz argued that the change favored Democrats in District 8 while “hyperpacking” Republicans into District 11. Commissioner Freeman argued that competitiveness should be applied “fairly and evenhandedly” across the state rather than just advantaging one party in a particular district. The Republican commissioners were correct that the change would necessarily favor Democratic electoral prospects given that the voter registration in the existing versions of both Districts 8 and 11 favored Republicans and that Commissioner McNulty did not propose any corresponding effort to make any Democratic-leaning districts more competitive. Commissioner McNulty was absent from the meetings in which these initial discussions occurred, but Commissioner Herrera noted that competitiveness was one of the criteria the Commission was required to consider and expressed support for the change.
Commissioner McNulty asked Desmond to try a few other ways of shifting the lines between Districts 8 and 11, one of which would have kept several communities with high minority populations together in District 8. Commissioner McNulty, noting that the area had a history of having an opportunity to elect, raised the possibility that the change might also preserve that opportunity. Adelson opined that, if the minority population of District 8 were increased slightly, the Commission might be able to present it to the Department of Justice as an eleventh opportunity-to-elect district, which would “unquestionably enhance the submission and enhance chances for preclearance.” Counsel suggested that having another possible ability-to-elect district could be helpful because District 26 was not as strong of an ability-to-elect district as the other districts.
District 8 contained many of the same concentrations of minority populations as the district identified as District 23 in the previous decade’s plan. The comparable district in that region of the state had a history of electing minority candidates pri- or to the 2002 redistricting cycle. In 2002, the Department of Justice identified that district as one of the reasons why the Commission did not obtain preclearance of its first proposed plan in that cycle. Although the Commission later argued to the Department of Justice in its 2012 submission that the minorities could not consistently elect their candidate of choice in that district between 2002 and 2012, several minority candidates had been elected to the state legislature from the district in that time period.
The Commission voted 3-2 to implement Commissioner McNulty’s proposed change into the working map and send it to Dr. King for further analysis, with the Republican commissioners voting against. This was the only change order that resulted in a divided vote.
This change order also affected the population count of Districts 11, 12, and 16. The order changed the deviation from ideal population from 1.5 percent to -2.3 percent in District 8, from 1.9 percent to 0.3 percent in District 11, from 1.7 percent to 4.3 percent in District 12, and from 1.9 percent to 4.8 percent in District 16. Be*1060cause of subsequent changes, the population deviations in these districts in the final map was -2.2 percent for District 8, 0.1 percent for District 11, 4.1 percent for District 12, and 3.3 percent for District 16. Therefore, the change in population deviation for each district that is both attributable to Commissioner McNulty’s change order and that actually remained in the final map was an increase in deviation of 0.7 percent for District 8, a decrease in deviation of 1.6 percent for District 11, an increase of 2.4 percent for District 12, and an increase in deviation of 1.4 percent for District 16.
These changes increased the percentage of Hispanic population in District 8 from 25.9 percent in the draft map to 34.8 percent in the final map, with Hispanic voting-age population from 22.8 percent to 31.3 percent. The Commission ultimately concluded, however, that while District 8 came closer to constituting a minority ability-to-elect district than the previous District 23, it did not ensure minority voters the ability to elect candidates of their choice. The changes were nonetheless retained in the final map.
The Commission approved the final legislative map on January 17, 2012, on a 3-2 vote, with the Republican commissioners voting against.
On February 28, 2012, the Commission submitted its plan to the Department of Justice for preclearance purposes. In its written submission, the Commission argued that the benchmark plan contained seven ability-to-elect districts, comprised of one Native American district and six Hispanic districts. The Commission argued that the new map was an improvement over the benchmark plan, as the new map contained ten districts (one Native American district and nine Hispanic districts) in which a minority group had the opportunity to elect the candidate of its choice. The Commission also noted that while District 8 was not an ability-to-elect district, its performance by that measure was improved over its predecessor, Benchmark District 23.
On April 26, the Department of Justice approved the Commission’s map.
I. The Motivation for the Deviations
As noted previously and explained in more detail below, at 1071-74, we conclude as a matter of law that the burden of proof is on plaintiffs. To prevail, plaintiffs must prove that the population deviations were not motivated by legitimate considerations or, possibly, if motivated in part by legitimate considerations, that illegitimate considerations predominated over legitimate considerations.7 We assume that seeking partisan advantage is not a legitimate consideration, and we conclude, as discussed at 1073-77, that compliance with the Voting Rights Act is a legitimate consideration.
We find that plaintiffs have not satisfied their burden of proof. In particular, we find that the deviations in the ten districts submitted to the Department of Justice as minority ability-to-elect districts were predominantly a result of the Commission’s good-faith efforts to achieve preclearance under the Voting Rights Act. Partisanship may have played some role, but the primary motivation was legitimate.
*1061With respect to the deviations resulting from Commissioner McNulty’s change to District 8 between the draft map and the final map, we find that partisanship clearly played some role. We also find, however, that legitimate motivations to achieve pre-clearance also played a role in the Commission’s decision to enact the change to District 8.
We acknowledge that it is difficult to separate out different motivations in this context. That is particularly true in this instance because the cited motivations pulled in exactly the same direction. As a practical matter, changes that strengthened minority ability-to-elect districts were also changes that improved the prospects for electing Democratic candidates. Those motivations were not at cross purposes. They were entirely parallel.
The Cruz Index, used by the commissioners in considering changes to the map aimed at strengthening minority districts, illustrates the overlap of these two motivations. It applied results from an election contest between a Hispanic Democrat and a white, non-Hispanic (Anglo) Republican. The commissioners used votes for candidate Cruz to reflect a willingness to vote for a Hispanic candidate — which was itself a proxy for the ability of the Hispanic population to elect its preferred candidate, regardless of that candidate’s ethnicity— but the voters could have been motivated, as much or even more, to vote for a Democrat. Similarly, voters who voted for Cruz’s opponent may have been willing to vote for a Hispanic candidate but were actually motivated to vote for a Republican. In using the Cruz Index to adjust district boundaries in order to strengthen the minority population’s ability to elect its preferred candidate, the commissioners used a measure that equally reflected the ability to elect a Democratic candidate.
The practical correlation between these two motivations was confirmed by the results of the 2012 election, conducted under the map that is the subject of this lawsuit. The legislators elected from districts identified by the Commission as minority ability-to-elect districts were all Democrats. As noted above, 19 of the 30 legislators elected from those districts were Hispanic or Native American.
It is highly likely that the members of the Commission were aware of this correlation. Individuals sufficiently interested in government and politics to volunteer to serve on the Commission and to contribute hundreds of hours of time to the assignment would be aware of historic voting patterns. If they weren’t aware before, then they would necessarily have become aware of the strong correlation between minority ability-to-elect districts and Democratic-leaning districts in the course of their work.
That knowledge could open the door to partisan motivations in both directions. If an individual member of the Commission were motivated to favor Democrats, that could have been accomplished under the guise of trying to strengthen minority ability-to-elect districts. Similarly, a member motivated to favor Republicans could have taken advantage of the process to concentrate minority population into certain districts in such a way as to leave a larger proportion of Republicans in the remaining districts.
Recognizing the difficulty of separating these two motivations, we find that the Commission was predominantly motivated by a legitimate consideration, in compliance with the Voting Rights Act.
All five of the commissioners, including the Republicans, put a priority on achieving preclearance from the Department of Justice on the first try. To maximize the chances of achieving that goal, the Com*1062mission’s counsel and consultants recommended creating ten minority ability-to-elect districts. There was not a partisan divide on the question of whether ten districts was an appropriate target.
After working to create ten such districts in the draft map, including Districts 24 and 26, all but Commissioner Stertz voted for the draft map. Commissioner Stertz’s reason for voting against the draft map, however, was not that he objected to the population deviations resulting from the creation of the ability-to-elect districts. Rather, he felt that the Commission had not paid sufficient attention to the other criteria that the Arizona Constitution requires the Commission to consider, such as keeping communities of interest together.
In short, the bipartisan support for the changes leading to the population deviations in the draft map undermines the notion that partisanship, rather than compliance with the Voting Rights Act, was what motivated those deviations.
We also find that the additional population deviation in these ten districts resulting from changes occurring between the passage of the draft map and the final map were primarily the result of efforts to obtain preclearance, some reservations by the Republican commissioners notwithstanding. After the draft map was completed, both Republican commissioners expressed concern about further depopulating minority ability-to-elect districts. At the hearing in which the Commission began work on the final map, Commissioner Stertz said that it was his “understanding that the maps as they are currently drawn do meet [the Voting Rights Act] criteria,” and that he didn’t want to “overpack Republicans into Republican districts ... all being done on the shoulders of strengthening [Voting Rights Districts].” Commissioner Freeman shared Commissioner Stertz’s concerns.
But the Commission’s counsel and consultants responded that there was uncertainty as to whether the map would pre-clear without strengthening those districts. And despite their initial reservations, the Republican commissioners did not vote against any of the change orders further strengthening the minority ability to elect in those districts. Commissioner Stertz even expressed support for these changes. In a public hearing that took place after the Commission made additional changes to the Voting Rights Act districts, Commissioner Stertz said that apart from a change order affecting Districts 8 and 11— which were not ability-to-elect districts and which we discuss next — he was “liking where the map has gone” and thought there was “a higher level of positive adjustments that have been made than the preponderance of the negative design of Districts 8 and 11.” At trial, Commissioner Stertz testified that he relied on counsel’s advice that ten benchmark districts were necessary, and that he thought those ten districts were “better today than when they were first developed in draft maps.” The bipartisan support for the goal of pre-clearance, and the bipartisan support for the change orders strengthening these ten districts to meet that goal, support the finding that preclearance motivated the deviations.
We make this finding despite plaintiffs’ contention that the selection of counsel and mapping consultant prove that Chairwoman Mathis was biased towards Democratic interests. We agree that giving Strategic Telemetry a perfect score is difficult to justify and reflects Mathis taking an ends-oriented approach to the process to select her preferred firm, Strategic Telemetry.
But even if Chairwoman Mathis preferred Strategic Telemetry for partisan reasons rather than the neutral reasons she expressed at the time, it would not *1063prove that partisanship was the reason she supported the creation of ability-to-elect districts. As we have discussed, strong evidence shows that preclearing on the first attempt was a goal shared by all commissioners, not just Chairwoman Mathis.
With respect to the changes to District 8 occurring between the draft map and final map, the evidence shows that partisanship played some role. Though Commissioner McNulty first presented the possible changes to Districts 8 and 11 as an opportunity to make District 8 into a more competitive district, that simply meant making District 8 into a more Democratic district. Because Districts 8 and 11 both favored Republicans before the proposed change, any shift in population between the two districts to make one of them more “competitive” necessarily increased the chances that a Democrat would win in one of those districts. In fact, in a close senate race in the newly drawn District 8, the Democrat did win. We might view the issue differently had Commissioner McNulty proposed to create a series of competitive districts out of both Democrat- and Republican-leaning districts, or applied some defined standards evenhandedly across the state. Instead, she sought to make one Republican-leaning district more amenable to Democratic interests. Moreover, the Commission was well aware of the partisan implications of the proposed change before adopting it. Both Republican commissioners made their opposition to the change, on the basis that it packed Republican voters into District 11 to aid Democratic prospects in District 8, known early on.
Nonetheless, while partisanship played a role in the increased population deviation associated with changing District 8, so too did the preclearance goal play a part in motivating the change. While Commissioner McNulty originally suggested altering Districts 8 and 11 for the sake of competitiveness, she subsequently suggested that District 8 could become an ability-to-elect district. Consultants and counsel endorsed this idea, in part because they had some doubts that District 26 would offer the ability to elect. It was not until after the consultants and counsel suggested pursuing these changes for the sake of preclearance that Chairwoman Mathis endorsed the idea. While the Commission ultimately concluded that it could not make a true ability-to-elect district out of District 8, the submission to the Department of Justice did cite the changes made to that district’s boundaries in arguing that the plan deserved preclearance. Compliance with the Voting Rights Act was a substantial part of the motivation for the treatment of District 8.
III. Resolution of Pretrial Motions
The parties filed several motions prior to trial that this court disposed of summarily in its order dated February 22, 2013, with an opinion explaining the bases of the rulings to follow. Before we turn to our conclusions of law on the merits of the case, we explain our rulings on those motions.
A. First Motion for Judgment on the Pleadings
Defendants’ first motion for judgment on the pleadings sought two forms of relief. First, defendants requested dismissal of the commissioners based on legislative immunity. Second, defendants requested dismissal of plaintiffs’ state-law claim as barred by the Eleventh Amendment. We now explain why both forms of relief were granted.
1. Standard of Judgment on the Pleadings
Judgment on the pleadings is appropriate when there is “no issue of mate*1064rial fact in dispute, and the moving party is entitled to judgment as a matter of law.” Fleming v. Pickard, 581 F.3d 922, 925 (9th Cir.2009). In assessing defendants’ motion, we “accepted] all factual allegations in the complaint as true and construefd] them in the light most favorable to the non-moving party.” Id.
2. The Commissioners Were Immune from Suit
It was not entirely clear from the complaint but plaintiffs’ claims against the commissioners appeared to be based solely on the commissioners’ official acts. That is, plaintiffs’ claims rested on the commissioners’ actions in connection with the adoption of a particular final legislative map. Plaintiffs’ federal claim sought relief pursuant to 42 U.S.C. § 1983 based on their belief that the adoption of that map constituted a violation of the Equal Protection Clause of the Fourteenth Amendment. The Commission argued legislative immunity forbade plaintiffs from pursuing this claim against the commissioners.
“The Supreme Court has long held that state and regional legislators are absolutely immune from liability under § 1983 for their legislative acts.” Kaahumanu v. Cnty. of Maui, 315 F.3d 1215, 1219 (9th Cir.2003). This immunity applies to suits for money damages as well as requests for injunctive relief. See Supreme Court of Va. v. Consumers Union of the United States, Inc., 446 U.S. 719, 734, 100 S.Ct. 1967, 64 L.Ed.2d 641 (1980). Litigants often disagree over whether legislative immunity applies to a particular individual or to particular acts performed by an individual occupying a legislative office. Kaahumanu, 315 F.3d at 1219 (legislative immunity applies only to “legislative rather than administrative or executive” actions). But plaintiffs effectively conceded the commissioners qualified as legislators performing legislative acts. So instead of the normal lines of attack, plaintiffs argued that Ex parte Young, 209 U.S. 123, 28 S.Ct. 441, 52 L.Ed. 714 (1908), prevented legislative immunity from requiring dismissal of the commissioners. Plaintiffs also claimed their request for attorneys’ fees permitted them to maintain suit against the commissioners. Neither argument was convincing.
Ex parte Young creates a legal fiction to avoid suits against state officials from being barred by the Eleventh Amendment. See, e.g., Miranda B. v. Kitzhaber, 328 F.3d 1181, 1189 (9th Cir.2003) (per curiam) (“[T]he doctrine of Ex parte Young avoids an Eleventh Amendment bar to suit....”). That fiction permits only “actions for prospective declaratory or injunctive relief against state officers in their official capacities for their alleged violations of federal law.” Coal, to Defend Affirmative Action v. Brown, 674 F.3d 1128, 1134 (9th Cir.2012). Plaintiffs did not cite any case where a court employed the fiction of Ex parte Young to avoid the otherwise applicable bar of legislative immunity. And existing case law reaches the opposite conclusion. See, e.g., Scott v. Taylor, 405 F.3d 1251, 1257 (11th Cir.2005) (finding legislative immunity barred claim for prospective injunctive relief). Thus, Ex parte Young was not sufficient to overcome the bar of legislative immunity.
Even if the court had agreed Ex parte Young might permit the naming of the commissioners in certain circumstances, it was particularly inapt here. Pursuant to Ex parte Young, the “state official sued ‘must have some connection with the enforcement of the act.’ ” Coal, to Defend Affirmative Action, 674 F.3d at 1134 (quoting Ex parte Young, 209 U.S. at 157, 28 S.Ct. 441). That connection must be “fairly direct” and a “generalized duty to enforce state law or general supervisory *1065power over the persons responsible for enforcing the challenged provision” is not sufficient. L.A Cnty. Bar Ass’n v. Eu, 979 F.2d 697, 704 (9th Cir.1992). Accordingly, Ex parte Young does not allow a plaintiff to sue a state official who cannot provide the relief the plaintiff actually seeks. See id.
Under Arizona’s redistricting process, the commissioners have no direct connection to implementing the final legislative map nor do they have any supervisory power over those state officials implementing the final legislative map. Rather, it is the Secretary of State who enforces the map. See Ariz. Minority Coal. for Fair Redistricting v. Ariz. Indep. Redistricting Comm’n, 211 Ariz. 337, 121 P.3d 843, 857 (Ariz.Ct.App.2005) (per curiam) (“Once the Commission certifies the maps, the secretary of state must use them in conducting the next election.”). Plaintiffs named the Secretary of State as a defendant and the Secretary of State conceded he is responsible for enforcing the map. In light of this, assuming Ex parte Young allows suit against the commissioners in some circumstances, the present suit did not qualify.
Finally, plaintiffs argued the commissioners’ “presence [was] essential to maintaining section 1983 relief, which includes an award of attorneys’ fees under 42 U.S.C. § 1988.” In other words, plaintiffs wanted to keep the commissioners as defendants to ensure the possibility of plaintiffs recovering their attorneys’ fees. Plaintiffs did not cite, and the court could not find, any authority permitting the issue of fees to determine the propriety of keeping certain defendants in a suit. Moreover, plaintiffs’ issue regarding fees was a problem of their own creation in that the Secretary of State undoubtedly was an appropriate defendant and plaintiffs could have sought fees from him. At oral argument, however, plaintiffs’ counsel conceded the complaint did not seek an award of fees from the Secretary of State.8 The fact that plaintiffs made a choice not to seek fees against one party from whom they could clearly obtain fees was not a sufficient basis to allow plaintiffs to continue this suit against inappropriate parties.
Neither Ex parte Young nor the impossibility of plaintiffs collecting fees from the remaining defendants justified keeping the commissioners as defendants. Therefore, the commissioners were entitled to judgment on the pleadings.
3. Plaintiffs State-Law Claim Was Barred by the Eleventh Amendment
In addition to their § 1983 claim, plaintiffs also asserted a state-law claim that the final legislative map “violates the equal population requirement of Ariz. Const. art. 4, pt. 2, § 1(14)(B).” Defendants moved to dismiss this state-law claim as barred by the Eleventh Amendment pursuant to Pennhurst State School & Hospital v. Halderman, 465 U.S. 89, 104 S.Ct. 900, 79 L.Ed.2d 67 (1984). Plaintiffs did not dispute that a straightforward application of Pennhurst established their state-law claim was barred by the Eleventh Amendment. Instead, plaintiffs argued defendants waived their Eleventh Amendment immunity. Plaintiffs were incorrect.
“For over a century now, [the Supreme Court] has consistently made clear that ‘federal jurisdiction over suits against unconsenting States was not contemplated by the Constitution when establishing the *1066judicial power of the United States.’ ” Sossamon v. Texas, — U.S. -, 131 S.Ct. 1651, 1657-58, 179 L.Ed.2d 700 (2011) (quoting Seminole Tribe of Fla. v. Florida, 517 U.S. 44, 54, 116 S.Ct. 1114, 134 L.Ed.2d 252 (1996)). A state may choose to waive its immunity, but the “test for determining whether a State has waived its immunity from federal-court jurisdiction is a stringent one.” Id. at 1658 (quoting Coll. Sav. Bank v. Fla. Prepaid Postsecondary Ed. Expense Bd., 527 U.S. 666, 675, 119 S.Ct. 2219, 144 L.Ed.2d 605 (1999)). That test consists of determining whether “the state’s conduct during the litigation clearly manifest[ed] acceptance of the federal court’s jurisdiction or [was] otherwise incompatible with an assertion of Eleventh Amendment immunity.” Hill v. Blind Indus. & Servs. of Md., 179 F.3d 754, 759 (9th Cir.1999). For example, the Ninth Circuit concluded waiver occurred when a state appeared, actively litigated a case, and waited until the first day of trial to claim immunity. Id. at 763. The situation in the present case was significantly different.
Plaintiffs filed their original complaint on April 27, 2012. The parties then engaged in protracted pre-answer maneuvers that ended on November 16, 2012, when the court denied defendants’ motion to dismiss. Approximately three weeks later, defendants filed their answer asserting Eleventh Amendment immunity as well as a formal motion seeking judgment on the pleadings based on that immunity. Thus, while the case had been pending for over nine months at the time immunity was first asserted, the vast majority of that time was consumed by briefing and deciding a motion to dismiss. There was no meaningful delay between issuance of the order on the motion to dismiss and defendants’ assertion of the Eleventh Amendment. And while defendants might have raised immunity earlier, the actual sequence of events falls short of meeting the “stringent” test for establishing waiver. Sossamon, 131 S.Ct. at 1658. Therefore, defendants were entitled to judgment on the pleadings regarding plaintiffs’ state-law claim.
B. Motion for Abstention
Citing Railroad Commission of Texas v. Pullman Co., 312 U.S. 496, 61 S.Ct. 643, 85 L.Ed. 971 (1941), defendants moved to stay this case and defer hearing plaintiffs’ federal claim until plaintiffs obtained resolution of state-law issues in state court or, in the alternative, to certify any state-law questions to the Arizona Supreme Court. A majority of the court summarily denied the motion, with Judge Silver dissenting.
Because “Congress imposed the duty upon all levels of the federal judiciary to give due respect to a suitor’s choice of a federal forum for the hearing and decision of his federal constitutional claims,” Pullman abstention is available only in narrowly limited, special circumstances. Zwickler v. Koota, 389 U.S. 241, 248, 88 S.Ct. 391, 19 L.Ed.2d 444 (1967). At its core, it “reflect[s] a doctrine of abstention appropriate to our federal system whereby the federal courts, ‘exercising a wise discretion,’ restrain their authority because of ‘scrupulous regard for the rightful independence of the state governments’ and for the smooth working of the federal judiciary.” Pullman, 312 U.S. at 501, 61 S.Ct. 643. “It is better practice, in a case raising a federal constitutional or statutory claim, to retain jurisdiction, rather than to dismiss.” Zwickler, 389 U.S. at 244 n. 4, 88 S.Ct. 391. Pullman abstention generally is appropriate only if three conditions are met: (1) the complaint “requires resolution of a sensitive question of federal constitutional law; (2) the constitutional question could be mooted or narrowed by a defini*1067tive ruling on the state law issues; and (3) the possibly determinative issue of state law is unclear.” Potrero Hills Landfill, Inc. v. Cnty. of Solano, 657 F.3d 876, 888-89 (9th Cir.2011) (quoting Spoklie v. Mont., 411 F.3d 1051, 1055 (9th Cir.2005)). Proper application of these conditions is meant to ensure federal courts defer “to state court interpretations of state law” while avoiding “ ‘premature constitutional adjudication’ that would arise from ‘interpreting state law without the benefit of an authoritative construction by state courts’.” Id. (quoting Gilbertson v. Albright, 381 F.3d 965, 971 n. 6 (9th Cir.2004) (en banc)) (internal quotation marks omitted).
When deciding whether to exercise its discretionary equity powers to abstain, a court also must consider that “abstention operates to require piecemeal adjudication in many courts,” possibly “delaying ultimate adjudication on the merits for an undue length of time.” Baggett v. Bullitt, 377 U.S. 360, 378-79, 84 S.Ct. 1316, 12 L.Ed.2d 377 (1964). That delay can work substantial injustice because forcing “the plaintiff who has commenced a federal action to suffer the delay of state court proceedings might itself effect the impermissible chilling of the very constitutional right he seeks to protect.” Zwickler, 389 U.S. at 252, 88 S.Ct. 391.
Delay caused by abstention is especially problematic in voting rights cases. Harman v. Forssenius, 380 U.S. 528, 537, 85 S.Ct. 1177, 14 L.Ed.2d 50 (1965). The Ninth Circuit noted in a redistricting case that due to the “special dangers of delay, courts have been reluctant to rely solely on traditional abstention principles in voting cases.” Badham v. U.S. Dist. Court for the N. Dist. of Cal., 721 F.2d 1170, 1173 (9th Cir.1983). Expressing specific concern about the possibility of a potentially defective redistricting plan being left in place for an additional election cycle, it held that “before abstaining in voting cases, a district court must independently consider the effect that delay resulting from the abstention order will have on the plaintiffs right to vote.” Id.
Given the importance of prompt adjudication of voting rights disputes, we exercised our discretion and decided not to abstain. The three conditions precedent to applying Pullman abstention identified above might have been present here, but we concluded that we should deny the motion without having to make that determination because of the likely delay that would have resulted.
If we abstained as defendants requested, it was not likely that a resolution could be reached in time to put a new plan in place, if necessary, for the 2014 election cycle. Not only are voting rights disputes particularly important, they are also particularly complex. The last round of litigation over redistricting in Arizona, concerning Arizona’s legislative redistricting maps following the 2000 census, commenced in March 2002. See Ariz. Minority Coal. for Fair Redistricting v. Ariz. Indep. Redistricting Comm’n, 220 Ariz. 587, 208 P.3d 676, 682 (2009) (en banc). The state trial court did not issue its decision until January 2004, twenty-two months later. See id. The appellate process did not conclude until the Arizona Supreme Court’s final decision in May 2009. Id. at 676. The Commission’s motion for abstention came before us in December 2012. At the time of our decision on the motion, in February 2013, no state court action was pending. Thus, deferring ruling on the federal claim would have delayed adjudication on the merits until a state court action was initiated and concluded, which likely would have pre-*1068eluded relief in time for the 2014 election cycle.9
Furthermore, we could not resolve the state-law issues as this case no longer included the state-law claim because the State of Arizona’s Eleventh Amendment immunity under Pennhurst precluded relief on that claim in federal court. And, it was also unclear whether any state law issues were implicated in plaintiffs’ remaining federal claim. In sum, this case is unlike the typical case warranting Pullman abstention, where the federal court will necessarily construe a state statute that the state courts themselves have not yet construed in order to decide the sensitive question of whether the state statute violates the federal Constitution. See, e.g., Potrero Hills Landfill, 657 F.3d at 889. Here, by contrast, we did not need to resolve any question of state law as a predicate to deciding the merits of the federal claim. Therefore, we concluded that the special circumstances necessary for exercising discretion to defer ruling on plaintiffs’ federal claim did not exist.
As an alternative to their request for abstention, defendants requested the court certify any state-law questions to the Arizona Supreme Court. A basic prerequisite for a court to certify a question to the Arizona Supreme Court is the existence of a pending issue of Arizona law not addressed by relevant Arizona authorities. See, e.g., Seltzer v. Paul Revere Life Ins. Co., 688 F.3d 966, 968 (9th Cir.2012). In addition, Arizona’s certification statute requires the presence of a state-law question that “may be determinative” of the case. A.R.S. § 12-1861. With the dismissal of plaintiffs’ state-law claim, there was no pending issue of Arizona law in this case. Therefore, the request in the alternative for certification also was denied.
C. Motion for Protective Order
Prior to discovery, the Commission moved for a protective order on the basis of legislative privilege. The Commission requested that the panel prohibit the depositions of the commissioners, their staff, and their consultants, as well as limit the scope of documents and interrogatories during discovery. We ordered the commissioners, at the time defendants in this case, to inform the court through counsel whether they would exercise legislative privilege if asked questions covered by the privilege. Commissioners Mathis, Herrera, and McNulty informed the court that they would invoke legislative privilege, *1069while Commissioners Freeman and Stertz indicated they would waive it. We later denied the motion for a protective order, and we now explain the basis for doing so.
Whether members of an independent redistricting commission can withhold relevant evidence or refuse to be deposed on the basis of legislative privilege is an issue of first impression. Neither the Ninth Circuit nor, as far as we can tell, any other court has decided whether members of an independent redistricting commission can assert legislative privilege in a challenge to the redistricting plan they produced. In the present litigation, we conclude that members of the Arizona Independent Redistricting Commission cannot assert a legislative evidentiary privilege.
State legislators do not have an absolute right to refuse deposition or discovery requests in connection with their legislative acts. In United States v. Gillock, 445 U.S. 360, 100 S.Ct. 1185, 63 L.Ed.2d 454 (1980), the Supreme Court held that a state senator could not bar the introduction of evidence of his legislative acts in a federal criminal prosecution. Although Gillock could have claimed protection under the federal Speech or Debate Clause had he been a Member of Congress, the Court refused “to recognize an evidentiary privilege similar in scope to the Federal Speech or Debate Clause” for state legislators. Id. at 366, 100 S.Ct. 1185. The Court reasoned that “although principles of comity command careful consideration, ... where important federal interests are at stake, as in the enforcement of federal criminal statutes, comity yields.” Id. at 373, 100 S.Ct. 1185. The Court in Gillock held that no legislative privilege exists in federal criminal prosecutions. It did not opine on the existence or extent of legislative privilege for state legislators in the civil context.
The Ninth Circuit has recognized that state legislators and their aides may be protected by a legislative privilege. See Jeff D. v. Otter, 643 F.3d 278, 289-90 (9th Cir.2011). That case did not consider legislative privilege in the redistricting context, however, let alone whether citizen commissioners could assert the privilege. Moreover, its discussion of legislative privilege was limited. The decision did not indicate whether state legislators might assert an absolute legislative privilege in all civil litigation, or whether any privilege state legislators held must yield when significant competing interests exist.
Whether or not state legislators might be able to assert in federal court an absolute legislative privilege in some circumstances, we do not think that the citizen commissioners here hold an absolute privilege. The Fourth Circuit has recognized, albeit not specifically in any redistricting cases, a seemingly absolute privilege against compulsory evidentiary process for state legislators and other officials acting in a legislative capacity. See EEOC v. Wash. Suburban Sanitary Comm’n, 631 F.3d 174, 180-81 (4th Cir.2011). The purposes underlying an absolute privilege for state legislators are that it “allows them to focus on their public duties by removing the costs and distractions attending lawsuits [and] shields them from political wars of attrition in which their opponents try to defeat them through litigation rather than at the ballot box.” Id. at 181. However, these are not persuasive reasons for extending the privilege to appointed citizen commissioners. Unlike legislators, the commissioners have no other public duties from which to be distracted. See Ariz. Const. art. IV, pt. 2, §§ 1(3), (13) (providing that commissioners cannot hold elected office during or for the three years following their service on the Commission). They cannot be defeated at the ballot box *1070because they don’t stand for election. Indeed, the process is not supposed to be governed by what happens at the ballot box. The reason why Arizona transferred redistricting responsibilities from the legislature to the Commission was to separate the redistricting process from politics. See Ariz. Proposition 106 (2000), available at http://www. azsos.gov/election/2000/info/pubpamphlet/prop2-C-2000.-htm (on the ballot title of the initiative creating the Commission, stating one purpose behind the law as “ending the practice of gerrymandering”).
In addition, to the extent comity is a rationale underlying legislative privilege, the Supreme Court has held that comity can be trumped by “important federal interests.” Gillock, 445 U.S. at 373, 100 S.Ct. 1185. The federal government has a strong interest in securing the equal protection of voting rights guaranteed by the Constitution, an interest that can require the comity interests underlying legislative privilege to yield. Cf. Badham, 721 F.2d at 1173 (observing that federal courts are more reluctant to abstain in voting rights cases and noting that the “right to vote is fundamental because it is preservative of all rights” (internal quotations marks and alterations omitted)).
For similar reasons, we also refuse to extend a qualified legislative privilege to the commissioners in this case. Some courts have recognized a qualified privilege for state legislators in redistricting cases, in which a balancing test determines whether particular evidence is barred by the privilége. See, e.g., Rodriguez v. Pataki, 280 F.Supp.2d 89, 101 (S.D.N.Y.2003), aff'd, 293 F.Supp.2d 302 (S.D.N.Y.2003). These cases did not involve an independent redistricting commission, however, and several of these cases even suggested that a legislative privilege would not apply to citizen commissioners. See Favors v. Cuomo, 285 F.R.D. 187, 220 (E.D.N.Y.2012) (concluding that permitting discovery would have minimal chilling effect on future legislative redistricting deliberations because New York had recently passed a law creating an independent redistricting commission composed of non-legislators); Rodriguez, 280 F.Supp.2d at 101 (distinguishing between discovery requests aimed at the legislature itself and those aimed at an advisory redistricting commission composed of legislators and non-legislators, because the latter was “more akin to a conversation between legislators and knowledgeable outsiders”); Marylanders for Fair Representation, Inc. v. Schaefer, 144 F.R.D. 292, 301 n. 19, 304-05 (D.Md. 1992) (holding that legislators were protected by the privilege, but not citizens serving on a redistricting advisory committee).
In determining whether a qualified privilege applies to state legislators, the courts that recognize a qualified privilege often balance the following factors: “(i) the relevance of the evidence sought to be protected; (ii) the availability of other evidence; (iii) the ‘seriousness’ of the litigation and the issues involved; (iv) the role of the government in the litigation; and (v) the possibility of future timidity by government employees who will be forced to recognize that their secrets are violable.” Rodriguez, 280 F.Supp.2d at 101. These factors weigh heavily against recognizing a privilege for members of an independent redistricting commission. Because what motivated the Commission to deviate from equal district populations is at the heart of this litigation, evidence bearing on what justifies these deviations is highly relevant. In the event that plaintiffs’ claims have merit, and that the commissioners were motivated by an impermissible purpose, the commissioners would likely have kept out of the public *1071record evidence making that purpose apparent. See Cano v. Davis, 193 F.Supp.2d 1177, 1181-82 (C.D.Cal.2002) (Reinhardt, J., concurring in part and dissenting in part) (“Motive is often most easily discovered by examining the unguarded acts and statements of those who would otherwise attempt to conceal evidence of discriminatory intent.”). The federal interest in protecting voting rights is a serious one, as discussed earlier, and can require comity concerns to yield.
Perhaps most importantly, the nature and purpose of the Commission undermines the claim that allowing discovery will chill future deliberations by the Commission or deter future commissioners from serving. See Favors, 285 F.R.D. at 220. The commissioners will not be distracted from other duties because they have no other duties, and their future actions will not be inhibited because they have no future responsibility. See Ariz. Const. art. IV, pt. 2, §§ 1(3), (13). And, as the majority in Marylanders observed: “We ... deem it extremely unlikely that in the future private citizens would refuse to serve on a prestigious gubernatorial committee because of a concern that they might subsequently be deposed in connection with actions taken by the committee.” 144 F.R.D. at 305 n. 23.
The parties dispute the relevance of some of plaintiffs’ requested discovery. But to the extent that plaintiffs have requested information not relevant to the central disputes in this litigation, the Commission need not rely on legislative privilege for protection. As stated in our order dated February 22, 2013, the court will not permit “discovery that is not central to the federal claims or any other inappropriate burden under Federal Rule of Civil Procedure 26(c).”
In conclusion, the rationale supporting the legislative privilege does not support extending it to the members of the Arizona Independent Redistricting Commission in this ease.
IV. Conclusions of Law
A. Burden of Proof
The Equal Protection Clause of the Fourteenth Amendment requires that state legislative districts “must be apportioned on a population basis,” meaning that the state must “make an honest and good faith effort to construct districts ... as nearly of equal population as is practicable.” Reynolds v. Sims, 377 U.S. 533, 577, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964). Some deviation in the population of legislative districts is constitutionally permissible, so long as the disparities are based on “legitimate considerations incident to the effectuation of a rational state policy.” Id. at 579, 84 S.Ct. 1362. Compactness, contiguity, respecting lines of political subdivisions, preserving the core of prior districts, and avoiding contests between incumbents are examples of the legitimate criteria that can justify minor population deviations, so long as these criteria are “nondiscriminatory” and “consistently applied.” Karcher v. Daggett, 462 U.S. 725, 740, 103 S.Ct. 2653, 77 L.Ed.2d 133 (1983).
Before requiring the state to justify its deviations, plaintiffs must make a prima facie case of a one-person, one-vote violation. By itself, the existence of minor deviations is insufficient to make out a prima facie case of discrimination. Brown v. Thomson, 462 U.S. 835, 842, 103 S.Ct. 2690, 77 L.Ed.2d 214 (1983). With respect to state legislative districts, the Supreme Court has said that, as a general matter, a “plan with a maximum population deviation under 10% falls within this category of minor deviations.” Id. at 842, 103 S.Ct. 2690. Although courts rarely strike down plans with a maximum deviation of less *1072than ten percent, a maximum deviation below ten percent does not insulate the state from liability, but instead merely keeps the burden of proof on the plaintiff. See Cox v. Larios, 542 U.S. 947, 124 S.Ct. 2806, 159 L.Ed.2d 831 (2004) (summarily affirming the invalidation of a plan with a 9.98 percent maximum population deviation).
Because the maximum deviation here is below ten percent, the burden is on plaintiffs to prove that the deviations did not result from the effectuation of legitimate redistricting policies. The primary way in which plaintiffs seek to carry their burden is by showing that the Commission deviated from perfect population equality out of a desire to increase the electoral prospects of Democrats at the expense of Republicans. Plaintiffs argue that partisanship is not a legitimate redistricting policy that can justify population deviations.
The Supreme Court has not decided whether or not political gain is a legitimate state redistricting tool. See Cox, 542 U.S. at 951, 124 S.Ct. 2806 (Scalia, J., dissenting) (noting that the Court has not addressed whether a redistricting plan with a maximum deviation under ten percent “may nevertheless be invalidated on the basis of circumstantial evidence of partisan political motivation”). Because we conclude that the redistricting plan here does not violate the Fourteenth Amendment whether or not partisanship is a legitimate redistricting policy, we need not resolve the question. For the purposes of this opinion, we assume, without deciding, that partisanship is not a valid justification for departing from perfect population equality.
Even assuming that small deviations motivated by partisanship might offend the Equal Protection Clause, plaintiffs will not necessarily sustain their burden simply by showing that partisanship played some role. The Supreme Court has not specifically addressed what a plaintiff must prove in a one-person, one-vote challenge when population deviations result from mixed motives, some legitimate and some illegitimate.
This panel has not reached a consensus on what the standard should be.10 We *1073conclude, for purposes of this decision, that plaintiffs must, at a minimum, demonstrate that illegitimate criteria predominated over legitimate criteria.
Finally, we reject plaintiffs’ argument that strict scrutiny applies to the extent that the Commission claims that racial motivations drove the deviations from population equality. All of the cases cited in support of this argument involve racial gerrymandering claims. See, e.g., Abrams v. Johnson, 521 U.S. 74, 117 S.Ct. 1925, 138 L.Ed.2d 285 (1997). As plaintiffs concede, this is not a racial gerrymandering case. Nor have plaintiffs specifically articulated how, in the absence of a claim of racial discrimination, strict scrutiny helps their case. Suppose that, applying strict scrutiny, we concluded that the Commission employed race as a redistricting factor in a manner not narrowly tailored to advance a compelling governmental interest. That may establish a racial gerrymandering violation, but it would not establish a one-person, one-vote violation. We decline to reduce plaintiffs’ burden by importing strict scrutiny into the one-person, one-vote context, a context in which the Supreme Court has made clear we owe state legislators substantial deference. See Gaffney v. Cummings, 412 U.S. 735, 749, 93 S.Ct. 2321, 37 L.Ed.2d 298 (1973).
In sum, plaintiffs must prove that the deviations were not motivated by legitimate considerations or, if motivated in part by legitimate considerations, that illegitimate considerations predominated over legitimate considerations. Because we have found that the deviations in the Commission’s plan were largely motivated by efforts to gain preclearance under the Voting Rights Act, we turn next to whether compliance with Section 5 of the Voting Rights Act is a permissible justification for minor population deviations.
B. Compliance with the Voting Rights Act as a Legitimate Redistricting Policy
The Supreme Court has not specifically spoken to whether compliance with the Voting Rights Act is a redistricting policy that can justify minor population deviations. The Court has not provided an exhaustive list of permissible criteria. Among the legitimate criteria it has approved are compactness, contiguity, respecting municipal lines, preserving the cores of prior districts, and avoiding contests between incumbents. Karcher, 462 U.S. at 740, 103 S.Ct. 2653. In the context of racial gerrymandering cases, the Court has assumed, without deciding, that the Voting Rights Act is a compelling state interest. Vera, 517 U.S. at 977, 116 S.Ct. 1941 (plurality opinion).
We conclude that compliance with the Voting Rights Act is among the legitimate redistrieting criteria that can justify minor population deviations. If compliance with the Voting Rights Act is not a legitimate, rational state policy on par with compactness and contiguity, we doubt that the Court would have assumed in Vera that it is a compelling state interest. Neither plaintiffs nor the dissenting opinion have offered a sensible explanation.
More importantly, we fail to see how compliance with a federal law concerning voting rights — compliance which is mandatory for a redistricting plan to take effect — cannot justify minor population *1074deviations when, for example, protecting incumbent legislators can. This is, perhaps, our primary disagreement with the dissenting opinion. It too narrowly defines the reasons that may properly be relied upon by a state to draw state legislative districts with wider variations in population.
The dissenting opinion correctly notes, at 1103-04, that states are required to establish congressional districts of essentially equal population. It acknowledges, as it must, that state legislative districts are not subject to as strict a standard. A state legislative plan may include some variation in district population in pursuit of legitimate interests.
The dissenting opinion also acknowledges, at 1101 & 1105, that obtaining pre-clearance under the Voting Rights Act was a legitimate objective in redistricting. But it contends that pursuit of that objective could not justify even minor variations in population among districts. In practical terms, the dissenting opinion would apparently permit the Commission to consider the preclearance objective only in drawing lines dividing districts of equal sizes.
The Supreme Court has made it clear, however, that states have greater latitude when it comes to state legislative districts. The Equal Protection Clause does not require exact equality. In drawing lines for state legislative districts, “[a]ny number of consistently applied legislative policies might justify some variance.” Karcher, 462 U.S. at 740, 103 S.Ct. 2653. Obtaining preclearance under the Voting Rights Act appears to us to be as legitimate a reason as other policies that have been recognized, such as avoiding contests between incumbents and respecting municipal lines.
Plaintiffs and the dissenting opinion, at 1103, attempt to reframe the inquiry, arguing that the text of the Voting Rights Act itself does not specifically authorize population deviations. That is correct; there is no specific authorization for population deviations in the text of the legislation. But neither is there specific, textual authorization for population deviations in any of the other legitimate, often uncodified legislative policies that the Supreme Court has held can justify population deviations. For example, the Supreme Court’s conclusion that compactness can justify population deviations does not turn on the existence of a Compactness Act that specifically authorizes population deviations for the sake of compact districts. The question is not whether the Voting Rights Act specifically authorizes population deviations, but whether seeking preclearance under the Voting Rights Act is a legitimate, rational state goal in the redistricting process. We are satisfied that it is.
The dissenting opinion, at 1103, goes a step further and argues that the Voting Rights Act itself prohibits any deviation in exact population equality for the purpose of complying with the Voting Rights Act. No court has so held, and we note that plaintiffs themselves have alleged that the Arizona redistricting plan violates the Equal Protection Clause, not that it violates the Voting Rights Act. We do not read the Act in the same way that the dissenting opinion does.11
*1075Plaintiffs also argue that the Department of Justice does not purport to be able to force jurisdictions to depopulate districts to comply with Section 5. In a document entitled “Guidance Concerning Redistricting Under Section 5 of the Voting Rights Act,” the Department advises: “Preventing retrogression under Section 5 does not require jurisdictions to violate the one-person, one-vote principle.” 76 Fed. Reg. 7470, 7472 (Feb. 9, 2011). But the Guidance goes on to make clear that, in the Department’s view, Section 5 might in some cases require minor population deviations in state legislative plans. When a jurisdiction asserts that it cannot avoid retrogression because of population shifts, the Department looks to see whether there are reasonable, less retrogressive alternatives, as the existence of these alternatives could disprove the jurisdiction’s assertion that retrogression is unavoidable. For state legislative redistricting, “a plan that would require significantly greater overall population deviations is not considered a reasonable alternative.” Id. (emphasis added). The implication is that the Department would consider a plan with slightly greater population deviation to be a reasonable plan that would avoid retrogression — in other words, the Department might hold a state in violation of Section 5 if it could have avoided retrogression with the aid of minor population deviations. To be clear, we do not base our understanding of the law upon the Department’s interpretation, but plaintiffs have cited the Department’s Guidance as supporting its position, and we do not agree. In our view, the Department’s Guidance expresses a conclusion that avoiding retrogression can justify minor population deviations. That is our conclusion, as well, based on our own view of the law, separate and apart from the Department’s position.
This conclusion is not altered by the Supreme Court’s recent decision in Shelby County v. Holder, — U.S. -, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013), which was decided after the legislative map in question here was drawn and implemented.12 In Shelby County, the Court held that Section 4(b) of the Voting Rights Act, which contained the formula determining which states were subject to the preclearance requirement, was unconstitutional. Id. at 2631. The Court did not hold that the preclearance requirement of Section 5 was unconstitutional, but its ruling rendered the preclearance requirement inapplicable to previously covered jurisdictions, at least until Congress enacts a new coverage formula that passes constitutional muster. See id.
Plaintiffs and the dissenting opinion, at 1100-02, argue that this ruling applies retroactively to this case, such that the Commission was not required to obtain pre-clearance for the legislative map at issue, thereby nullifying the pursuit of preclearance as a justification for population deviations. See Harper v. Va. Dep’t of Taxation, 509 U.S. 86, 113 S.Ct. 2510, 125 L.Ed.2d 74 (1993) (requiring that a rule of federal law announced by the Court and applied to the parties in that controversy “be given full retroactive effect by all courts adjudicating federal law”).
But that approach reads too much into Shelby County. The Court did not hold that Section 5 of the Voting Rights Act, the section that sets out the preclearanee process, was unconstitutional. The Court’s *1076opinion stated explicitly to the contrary: “We issue no holding on § 5 itself, only on the coverage formula.” Shelby Cnty., 133 S.Ct. at 2631. The Court did not hold that Arizona or any other jurisdiction could not be required to comply with the preclearance process, if a proper formula was in place for determining which jurisdictions are properly subject to the preclearance process. To the contrary, the Court’s opinion expressly faulted Congress for not updating the coverage formula, implying that a properly updated coverage formula that “speaks to current conditions” would withstand challenge. Id.
If we had before us a challenge to the coverage formula set forth in Section 4 of the Voting Rights Act, we would unquestionably be expected to apply Shelby County “retroactively,” and we would do so. That is, however, not the issue before us. Neither is the issue before us whether the legislative map violated or complied with the Voting Rights Act.
Rather, the issue is whether the Commission was motivated by compliance with that law in deviating from the ideal population. In other contexts, where the issue is not whether the actions of public officials actually complied with the law but instead whether they might have reasonably thought to have been in compliance, we do not expect those public officials to predict the future course of legal developments.
For example, in the qualified immunity context, the issue is whether the actions of public officials “could reasonably have been thought consistent with the rights they are alleged to have violated.” Anderson v. Creighton, 483 U.S. 635, 638, 107 S.Ct. 3034, 97 L.Ed.2d 523 (1987). There, we assess their actions based on law “clearly established” at the time their actions were taken. Id. at 639, 107 S.Ct. 3034. Similarly, in the Fourth Amendment context, we decline to apply the exclusionary rule when a police officer conducts a search in reasonable reliance on a later invalidated statute. Illinois v. Krull, 480 U.S. 340, 348-19, 107 S.Ct. 1160, 94 L.Ed.2d 364 (1987). We generally decline to require the officer to predict whether the statute will later be held unconstitutional, unless the statute is so clearly unconstitutional that a reasonable officer would have known so at the time. Id. at 355, 107 S.Ct. 1160; see also Davis v. United States, — U.S. -, 131 S.Ct. 2419, 2431-32,180 L.Ed.2d 285 (2011) (noting that even though a new Fourth Amendment rule applies retroactively, “the exclusion of evidence does not automatically follow” because of the good-faith exception).
Arizona was not the only state that drew new district lines following the 2010 census. The other states and jurisdictions subject to preclearance under the Voting Rights Act engaged in the same exercise. Nothing in Shelby County suggests that all those maps are now invalid, and we are aware of no court that has reached such a conclusion, despite the concern expressed in the dissenting opinion, at 1100, that leaving the maps in place “would give continuing force to Section 5.” To repeat, Shelby County did not hold Section 5 to be unconstitutional. Neither did it hold that any effort by a state to comply with Section 5 was improper.
In redistricting, we should expect states to comply with federal voting rights law as it stands at the time rather than attempt to predict future legal developments and selectively comply with voting rights law in accordance with their predictions. Accordingly, so long as the Commission was motivated by the requirements of the Voting Rights Act as it reasonably understood them at the time, compliance with the Voting Rights Act served as a legitimate justification for minor population deviations.
*1077 C. Application to 2012 Legislative Map
Plaintiffs argue that Districts 8, 24, and 26 could not have been motivated by compliance with the Voting Rights Act. They argue that only eight ability-to-elect districts existed in the benchmark plan. Because the Commission had created eight ability-to-eleet districts even without Districts 8, 24, and 26, and avoiding retrogression only requires creating as many ability-to-elect districts as are in the benchmark plan, plaintiffs argue that the Voting Rights Act could not have motivated the creation of these three districts. In essence, plaintiffs urge us to determine how many ability-to-elect districts were strictly necessary to gain preclearance and to hold that deviations from the creation of purported ability-to-elect districts above that number cannot be justified by Voting Rights Act compliance.
This argument runs into several problems. First of all, plaintiffs have not given the court a basis to independently determine that there existed only eight ability-to-eleet districts in the benchmark plan. Plaintiffs point to the fact that the Commission argued that there were eight benchmark districts in its submission to the Department of Justice. But the submission to the Department was an advocacy document. The Commission was motivated to make the strongest case for preclearance by arguing for a low number of benchmark ability-to-elect districts and a high number of new ability-to-elect districts. The Commission’s consultants and counsel, in public meetings, had advised the Commission that their analysis suggested the existence of ten benchmark districts. The discrepancy between the advice given in meetings and the arguments put forth in the submission to the Department of Justice is not a sufficient basis for the court to conclude that there were only eight ability-to-elect districts in the benchmark plan. Moreover, while plaintiffs criticize elements of the functional analysis performed by the Commission’s consultants, plaintiffs have not provided the court with any functional analysis of their own or from any other source showing which districts provided minorities with the ability to elect in either the benchmark plan or the current plan that they challenge. In short, even if we were inclined to independently determine how many ability-to-elect districts existed in the benchmark plan, plaintiffs have not carried their burden to show that there were only eight.
In any event, we need not determine whether the minor population deviations were strictly necessary to gain preclearance. Plaintiffs presented testimony from an expert witness, Thomas Hofeller, to demonstrate that a plan could have been drawn with smaller population deviations. Dr. Hofeller prepared such a map, but he acknowledged that he had not taken other state interests into account, including in-' terests clearly identified as legitimate, nor had he performed a racial polarization or functional analysis, so that map did not necessarily present a practical alternative. Because he concluded, contrary to the Commission and its counsel and consultants, that the benchmark number for minority ability-to-elect districts in the prior plan was only eight (seven Hispanic districts and one Native American district), his belief that his alternative map would have been precleared by the Justice Department was disputed. More importantly, evidence that a map could have been drawn with smaller population deviations does not prove that illegitimate criteria motivated the deviations. See Marylanders for Fair Representation, Inc. v. Schaefer, 849 F.Supp. 1022, 1035 (D.Md.1994).
*1078Rather, it is enough that the minor population deviations are “based on legitimate considerations.” Reynolds v. Sims, 377 U.S. 533, 579, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964). In other words, we will invalidate the plan only if the evidence demonstrates that the deviations were not the result of reasonable, good-faith efforts to comply with the Voting Rights Act. We will not invalidate the plan simply because the Commission might have been able to adopt a map that would have precleared with less population deviation if we determine that in adopting its map the Commission was genuinely motivated by compliance with the Voting Rights Act.
This approach is in accord both with the deference federal courts afford to states in creating their own legislative districts and the realities of the preclearance process. The Department of Justice does not inform jurisdictions of the number of districts necessary for preclearance ahead of time. Nor could the Commission be certain which districts in any tentative plan would be recognized by the Department as having an ability to elect. These determinations are complex and not subject to mathematical certainty. For us to determine the minimum number of ability-to-elect districts necessary to comply with the Voting Rights Act and then to strike down a plan if minor population deviations resulted from efforts that we concluded were not strictly necessary for compliance would create a very narrow target for the state. It would also deprive states of the flexibility to which the Supreme Court’s one-person, one-vote jurisprudence entitles them in legislative redistricting. See Gaffney v. Cummings, 412 U.S. 735, 749, 93 S.Ct. 2321, 37 L.Ed.2d 298 (1973) (“Nor is the goal of fair and effective representation furthered by making the standards of reapportionment so difficult to satisfy that the reapportionment task is recurringly removed from legislative hands and performed by federal courts”).
That deviations from perfect population equality in this case resulted in substantial part because of the Commission’s pursuit of preclearance is evidenced both by its deliberations and by advice given to the Commission by its counsel and consultants. Plaintiffs cite Larios v. Cox for the proposition that advice of counsel is not a defense to constitutional infirmities in a redistricting plan. 300 F.Supp.2d 1320 (N.D.Ga.2004), aff'd, 542 U.S. 947, 124 S.Ct. 2806, 159 L.Ed.2d 831 (2004). In Larios, state legislators mistakenly believed that any plan with a maximum deviation below ten percent was immune from a one-person, one-vote challenge and then created a plan with a maximum deviation of 9.98 percent deviations in the pursuit of illegitimate objectives. See id. at 1328. In holding that the plan violated the one-person, one-vote principle, the court held that reliance on faulty legal advice did not remedy the constitutional infirmity in the plan. Id. at 1352 n. 16. But in Larios, there was no question that the legislature had pursued illegitimate policies. The legislature had taken counsel’s advice to mean that it did not need to have legitimate reasons for deviating. The court held that they did need legitimate reasons for deviating, and the Supreme Court affirmed.
Here, by contrast, what motivated the Commission is at issue. Counsel’s advice does not insulate the Commission from liability, but it is probative of the Commission’s intent. That is not to say that reliance on the advice of counsel will in all cases demonstrate the good-faith pursuit of a legitimate objective. The advice might be so unreasonable that the Commission could not reasonably have believed it, or other evidence may show that the Commission was not acting pursuant to the advice. But the Commission’s attorneys *1079gave reasonable advice as to how to pursue what they identified as a legitimate objective, and the Commission appeared to act in accordance with that advice. That is strong evidence that the Commission’s actions were indeed in the pursuit of that objective, one that we have concluded for ourselves was legitimate.
With respect to the ten districts presented to the Department of Justice as ability-to-elect districts, including Districts 24 and 26, the evidence before us shows that the population deviations were predominantly based on legitimate considerations. The Commission was advised by its consultants and counsel that it needed to create at least ten districts. Given the uncertainty in determining the number of districts, and that one of the Commission’s highest priorities was to preclear the first time, the Commission was not unreasonable in acting pursuant to this advice. As noted in our findings of fact, the target of ten districts was not controversial and had bipartisan support. All commissioners, including the Republican appointees, believed that ten districts were appropriate.
A somewhat closer question is presented by the changes to the district boundaries, including Districts 24 and 26, made between the draft map and the final map. The draft racial polarization analysis prepared by King and Strasma indicated that minorities would be able to elect candidates of them choice in all ten proposed ability-to-elect districts' in the draft map. Plaintiffs argue that no further changes could be justified by the Commission’s desire to obtain preclearance because the draft map met that goal. The preclearance decision was not going to be made by King and Strasma, however, and the Commission could not be sure what it would take to satisfy the Department of Justice. The Commission was advised to try to strengthen the minority ability-to-elect districts even further, and it was not unreasonable under the circumstances for the Commission to undertake that effort. With regard to the ten ability-to-elect districts, we conclude that plaintiffs have not carried their burden of demonstrating that no legitimate motive caused the deviations or that partisanship predominated. Creation of these districts was primarily a consequence of the Commission’s good-faith efforts to comply with the Voting Rights Act and to obtain preclearance.
District 8 presents an even closer question, because the evidence clearly shows that partisanship played some role in its creation. Commissioner McNulty presented the possible change to Districts 8 and 11 as an opportunity to make District 8 into a more competitive district. We do not doubt that the creation of competitive districts is a rational, legitimate state interest. But to justify population deviations, legitimate state criteria must be “nondiscriminatory” and “consistently applied.” Karcher v. Daggett, 462 U.S. 725, 740, 103 S.Ct. 2653, 77 L.Ed.2d 133 (1983). Commissioner McNulty’s competitiveness proposal was neither applied consistently nor in a nondiscriminatory fashion. It was applied to improve Democratic prospects in one single district. It was not applied to districts favoring Democrats as well as to those favoring Republicans, so competitiveness cannot justify the deviation. We have found that partisanship motivated the Democratic commissioners to support this change, since both expressed support for it before there was any mention of presenting District 8 to the Department of Justice for the sake of preclearance.
But while partisanship played some role, plaintiffs have not carried their burden to demonstrate that partisanship predominated over legitimate factors. Because Commissioner McNulty’s change only slightly *1080increased the level of population inequality in District 8 and the other affected districts, let alone the plan as a whole, plaintiffs must make a particularly strong showing to carry their burden. Cf. Karcher, 462 U.S. at 741, 103 S.Ct. 2653 (“The showing required to justify population deviations is flexible, depending on the size of the deviations, [etc.]”). As noted in our findings, the changes in population inequality from draft map to final map that can be attributed to the vote on Commissioner McNulty’s proposed change is an increase of 0.7 percent deviation in District 8, a decrease of 1.6 percent in District 11, an increase of 2.4 percent in District 12, and an increase of 1.4 percent in District 16. Altogether, the change resulted in a small decrease in deviation in one district and small increases in deviation in three districts. While there is some increase in deviation that can be attributed in part to partisanship, it is not a particularly large increase.
We have also found that the preclearance goal played a role in the change to District 8. Consultants and counsel suggested pursuing it for the sake of preclearance, and only then did Chairwoman Mathis endorse the idea. Without her vote, there would not have been a majority to adopt that change. In light of the small deviations resulting from this change order and because legitimate efforts to achieve preclearance also drove the decision, plaintiffs have not proved that partisanship predominated over legitimate reasons for the Commission as a whole.
We have concluded that compliance with the Voting Rights Act is a legitimate state policy that can justify minor population deviations, that the deviations in the map in large part resulted from this goal, and that plaintiffs have failed to show that other, illegitimate motivations predominated over the preclearance motivation. Therefore, plaintiffs’ challenge to the map under the one-person, one-vote principle fails.
V. Conclusion
We find in favor of the Commission on plaintiffs’ claim that the Commission’s legislative redistricting plan violated the one-person, one-vote principle of the Equal Protection Clause of the Fourteenth Amendment of the United States Constitution. We order the entry of judgment for the Commission.
. This per curiam opinion speaks for a majority of the court in all but one respect. On the issue of the burden of proof that plaintiffs must bear, there is not a majority opinion. See the specific discussion on that subject below, at 1072-73 n. 10.
Judge Silver concurs in the result and joins this opinion in all but three respects. One is the burden of proof requirement just mentioned. There is no majority conclusion on that subject. Her second difference is with the factual finding that partisanship played some part in the drafting of the legislative district maps, primarily discussed below in section II.I, at 1060-63, and to some extent in section IV.C, at 1079-80. She finds that partisanship did not play a role. The finding on that subject expressed in this opinion represents a majority consisting of Judge Clifton and Judge Wake. The third disagreement, previously announced, was from the majority's denial prior to trial of defendants' motion for abstention under Railroad Commission of Texas v. Pullman Co., 312 U.S. 496, 61 S.Ct. 643, *104785 L.Ed. 971 (1941), discussed below in section III.B, at 1066-69. That motion was denied by a majority consisting of Judge Clifton and Judge Wake. Judge Silver's separate views are expressed in a separate opinion, concurring in part, dissenting in part, and concurring in the judgment, filed together with this per curiam opinion.
Judge Wake dissents from the result reached in this opinion, though he joins portions of it. In addition to the finding that partisanship played some role, identified in the preceding paragraph, he specifically joins in section III of this opinion, at 1063-71, discussing our resolution of pretrial motions. His views are expressed in his separate opinion, concurring in part, dissenting in part, and dissenting from the judgment, also filed together with this opinion.
. Arizona Secretary of State Ken Bennett, sued in his official capacity, is also a nominal defendant in the action. When we refer to "defendants” in this opinion, however, we refer collectively to the Commission and commissioners.
. The Democratic candidate in District 8 won with 49 percent of the vote; the Republican received 46 percent of the vote, and the Libertarian candidate received the remaining 5 percent. Republicans won both of the state house races in District 8.
. Van Jones served as a special advisor to President Obama in 2009. He resigned that position after criticism from conservatives and Republicans.
. In a case decided after the implementation of the Commission's new redistricting plan, the Supreme Court held unconstitutional the coverage formula used to determine which states are subject to the Section 5 preclearance requirement. See Shelby County v. Holder, - U.S. -, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013). We discuss the impact of Shelby County on this case in our conclusions of law, at 1075-77.
. In order to better understand the factual findings that follow, some understanding of the requirements of the Voting Rights Act is useful. We include this discussion as background, acknowledging that it incorporates conclusions of law, albeit for the most part *1054conclusions that do not appear to us to be controversial.
. As discussed below, at 1072-73 n. 10, we have not reached agreement on the legal standard to be applied. A majority of the court has concluded that plaintiffs have failed to demonstrate that illegitimate considerations predominated over legitimate ones. Necessarily, therefore, plaintiffs have not proven that illegitimate considerations were the actual and sole reasons for the population deviations. By either test adopted by the two judges that make up a majority of the court, plaintiffs have failed to carry their burden.
. The portion of the complaint referenced during oral argument seeks fees "against the IRC only." In light of this language, it is unclear why plaintiffs believed they had requested fees from the individual commissioners.
. This case commenced in April 2012. We set a schedule with the intent that our decision would be filed in time for the 2014 election cycle, even leaving time for review by the Supreme Court. That is why trial was scheduled and held in March 2013, even though the parties requested a later trial date. The subsequent filing by the Supreme Court in June 2013 of its decision in Shelby County v. Holder, — U.S. -, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013), put that goal in jeopardy. The parties were subsequently ordered to brief the impact of Shelby County and, as illustrated by the dialogue between this opinion and the dissenting opinion, it has taken us some time to determine that impact. When we denied the motion for abstention, however, we did not know that Shelby County was coming. The denial of the motion for abstention was based on our belief that we would reach a conclusion in time for the 2014 election cycle and that it would be highly unlikely that a similarly timely result could be achieved if we abstained in favor of state court adjudication.
We note that even with the unanticipated delay to consider the impact of Shelby County, our decision is filed about twenty-four months after the commencement of this action, about on par with the twenty-two months that it took the Arizona trial court to resolve the challenge to the legislative redistricting maps drawn following the 2000 census. We remain of the view that abstaining in favor of state court litigation, which would likely have entailed an appeal following a state trial court decision, would have taken even more time.
. As expressed in her separate concurring opinion, at 1085-87, Judge Silver concludes that plaintiffs must show that the "actual and sole reason” for the challenged population deviation was improper. See Rodriguez v. Pataki, 308 F.Supp.2d 346, 366 (S.D.N.Y.2004) (holding that plaintiffs must show that the "deviation results solely from an unconstitutional or irrational state purpose” (emphasis added)).
Judge Clifton is not persuaded that the bar ought to be set that high. Some Supreme Court authority suggests that plaintiffs must show that illegitimate criteria at least predominated over legitimate considerations. For example, while government programs that draw classifications on the basis of race are typically subject to strict scrutiny, redistricting plans challenged for racial gerrymandering are not subject to strict scrutiny "if race-neutral, traditional districting considerations predominated over racial ones.” Bush v. Vera, 517 U.S. 952, 964, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996) (plurality opinion). Requiring a showing that illegitimate criteria predominated over legitimate criteria appears appropriate to him in light of the deference courts afford states in constructing their legislative districts and because multiple motives will frequently arise in any deliberative body. Cf. Miller v. Johnson, 515 U.S. 900, 915-16, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995) (noting that courts must be "sensitive to the complex interplay of forces that enter a legislature’s redistricting calculus” and afford states the "discretion to exercise the political judgment necessary to balance competing interests”).
Judge Wake, as discussed in his separate opinion, at 1106-07, concludes that both the "only motive” and the “predominant motive” standards are unsatisfactory.
For decision purposes, a majority of the panel, made up of Judge Clifton and Judge Silver, have concluded that plaintiffs have not *1073demonstrated that partisanship predominated over legitimate redistricting considerations, applying the lower standard favored by Judge Clifton. Though Judge Silver concludes that the standard should be higher, if the predominance standard is not met, the "actual and sole reason” standard cannot be met. For discussion purposes, therefore, this per cu-riam opinion will speak in terms of the predominance standard.
. Similarly, the dissenting opinion contends, at 1103, that the Department of Justice “has never required unequal population for pre-clearance in the 48 years of administering Section 5.” That assertion is not proven. More importantly, it is an irrelevant straw man. For preclearance purposes, any variation in population is a means, not an end. There would never be reason for the Department to "require!] unequal population.” That is not the Department’s goal. The question is whether a state might improve its chances of obtaining preclearance by presenting a plan that includes minor population *1075variations. The evidence presented to us supported that proposition, and neither plaintiffs nor the dissenting opinion deny that fact.
. As noted above, the decision was announced after the trial of this case. We ordered and obtained supplemental briefing from the parties on the impact of the decision on this case.