concurring in part, dissenting in part, and concurring in the judgment:
I agree plaintiffs have not proven a violation of Equal Protection and, therefore, I concur in the judgment against them. I also join the rulings in connection with the motion for judgment on the pleadings. I disagree, however, on the issue of abstention. Also, I have my own view of the standard applicable to plaintiffs’ claim and whether plaintiffs proved partisanship was involved in crafting the final map.1
1. Pullman Abstention
In December 2012, defendants requested we stay this case and defer hearing plaintiffs’ federal claim until plaintiffs’ state-law claim could be resolved by the Arizona courts. At that specific time, I believed abstention was appropriate. The following explains why I reached that conclusion and why, if the motion were being *1081decided today, abstention likely would not be appropriate.
As outlined in the per curiam opinion, Pullman abstention may be appropriate when three conditions are met. “First, the complaint must touch on a sensitive area of social policy upon which the federal courts ought not to enter unless no alternative to its adjudication is open.” Cano v. Davis, 191 F.Supp.2d 1140, 1142 (C.D.Cal.2002) (quotation omitted). Second, it must be clear that the federal constitutional claim presented in the complaint “could be mooted or narrowed by a definitive ruling on the state law issues” raised by the complaint. Potrero Hills Landfill, Inc. v. Cnty. of Solano, 657 F.3d 876, 888 (9th Cir.2011) (quotation omitted). And third, “the possibly determinative issue of state law is unclear.” Id. (quotation omitted). In my view, all three conditions were met.
On the first condition, as observed by another three judge panel hearing a redistricting suit, “[rjedistricting is undoubtedly a sensitive area of state policy.” Cano, 191 F.Supp.2d at 1142. Neither plaintiffs nor the per curiam opinion disputes this condition was satisfied.
On the second condition, resolution of the state-law claim raised by plaintiffs might have removed the need to address their federal constitutional claim. In opposing the request for abstention, plaintiffs seemed to be claiming the second condition was not satisfied because it was not certain that resolving their state-law claim would end the case. But certainty is not required. As explained by the Ninth Circuit, it need not be “absolutely certain” that the state-law issue will “obviate the need for considering the federal constitutional issues.” Sinclair Oil Corp. v. Cnty. of Santa Barbara, 96 F.3d 401, 409 (9th Cir.1996). It is sufficient that the state-law issue “may” have some impact on the federal claim. C-Y Dev. Co. v. City of Redlands, 703 F.2d 375, 379 (9th Cir.1983). More importantly, however, plaintiffs’ own statements indicated that they believed resolution of their state-law claim would end this case. That is, plaintiffs argued they were certain to prevail on their state-law claim. If plaintiffs were correct, the federal claim need not have ever been addressed, meaning the second condition for abstention was satisfied.
Finally, on the third condition, and despite plaintiffs’ arguments that their state-law claim was a sure winner, there was genuine uncertainty about the meaning of the Arizona constitutional provision regarding equal population. Plaintiffs believed “the Arizona Constitution’s equal population clause is plain” and it required absolute equality of population. While defendants disagreed with plaintiffs’ reading, they conceded there was some uncertainty about the meaning of Arizona’s equal population requirement. That concession was wise given the language of the Arizona Constitution coupled with the Arizona Supreme Court’s cryptic comments in a prior redistricting case. Ariz. Minority Coal, for Fair Redistricting v. Ariz. Indep. Redistricting Comm’n, 220 Ariz. 587, 208 P.3d 676, 686 (2009). And, in any event, the required amount of “uncertainty” for Pullman purposes is not very difficult to show.
“Uncertainty for purposes of Pullman abstention means that a federal court cannot predict with any confidence how the state’s highest court would decide an issue of state law.” Pearl Inv. Co. v. City and Cnty. of San Francisco, 774 F.2d 1460, 1465 (9th Cir.1985). That uncertainty might be because of a statutory ambiguity or “because the question is novel and of sufficient importance that it ought to be addressed first by a state court.” Id. In my view, we do not know how the Arizona courts would interpret the state constitu*1082tional language. Accordingly, the third condition was met.
Because the three Pullman conditions were met, the question becomes whether some other factor rendered abstention inappropriate. The Supreme Court has recognized that a court deciding whether to abstain must be cognizant 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). And abstention is particularly troublesome in voting rights cases “because of the importance of safeguarding the right to vote.” Cano v. Davis, 191 F.Supp.2d 1140, 1142 (C.D.Cal.2002). But even in a voting rights case, the Ninth Circuit affirmed a decision to abstain when the abstention order was issued only six months before a relevant voting deadline. Badham v. U.S. Dist. Court, 721 F.2d 1170, 1174 (9th Cir.1983). In doing so, the court noted the focus should be on the risk that delay would harm the right to vote. Because, in that case, there was no substantial risk of harm to that right, abstention was appropriate. Id.
The per curiam opinion relies on the possibility of undue delay as the primary basis for rejecting the abstention request. But at the time the motion was filed it was very unlikely plaintiffs’ right to vote would have been impacted if they were sent to state court. The Commission represented that, upon arriving in state court, it would stipulate to consolidating the preliminary injunction hearing with the trial. It also agreed that the discovery performed in federal court could be used in state court. The first relevant deadline for the 2014 elections was April 28, 2014, the first day candidates could file their nomination petitions. Thus, when the abstention motion was filed in December 2012, sending the parties to state court would have given the state court approximately fourteen months to order relief before any possible harm could be suffered. Given that length of time, the state courts would have had ample time to act.2
In addition to concerns about the possible delay should the parties be sent to state court, the per curiam opinion also seems to rely on the dismissal of plaintiffs’ state-law claim as a special factor weighing against abstention.3 But the absence of a pending state-law claim should have had no impact on the abstention inquiry. In Harris County Commissioners Court v. Moore, 420 U.S. 77, 81, 95 S.Ct. 870, 43 L.Ed.2d 32 (1975), the Supreme Court found Pullman abstention appropriate even though the plaintiffs in that case “did not expressly raise a state-law claim in *1083their complaint.” In Moore, there was an issue of state law lurking in the background of the federal Equal Protection claim that, if decided a certain way, might have negated the factual premise for the federal claim. Id. at 85-88, 95 S.Ct. 870. There is no real dispute that, in this case, resolution of the state-law claim raised by plaintiffs might have had a similar impact.
Finally, now that the first important election deadline is upon us, I recognize that the abstention calculus is significantly different. If the motion were being decided today, abstention likely would not be appropriate because the state court would not have time to provide relief. Thus, today I am comfortable reaching the merits of plaintiffs’ claim. I note only that something is not quite right with plaintiffs choosing to litigate a very tenuous federal claim when they have a state-law claim they believe is guaranteed to give them a victory. Therefore, absent the looming election deadlines, I would still be inclined to send the parties to state court.4
2. Partisanship Likely Not Cognizable Basis for Suit
The per curiam opinion wisely refuses to decide whether minor population deviations, ie. deviations below ten-percent, motivated by partisanship offend the Equal Protection Clause. I doubt they do.
The redistricting process, with all its adversarial tensions, has always been recognized as a profoundly partisan process. Gaffney v. Cummings, 412 U.S. 735, 753, 93 S.Ct. 2321, 37 L.Ed.2d 298 (1973) (“Politics and political considerations are inseparable from districting and apportionment.”). The Supreme Court has repeatedly noted without condemnation that entities responsible for redistricting often act in explicitly partisan ways, such as drawing lines to protect incumbents or drawing lines to ensure a particular district elects a Democratic representative. See, e.g., Easley v. Cromartie, 532 U.S. 234, 248, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001) (plan was drawn “to protect incumbents — a legitimate political goal”); id. at 245, 121 S.Ct. 1452 (noting a legislature might draw lines to “secure a safe Democratic seat”). And while partisanship is not a terribly noble means of establishing parameters impacting the fundamental right to vote, it has long been a given, embedded in our system of government. Thus, actual use of partisanship — or at least allegations that partisanship drove redistricting decisions' — are inevitable as long as partisan entities are responsible for redistricting.
Of course, Arizona has attempted to “remove redistricting from the political process by extracting [the authority to conduct redistricting] from the legislature and governor and instead granting it to an independent commission of balanced appointments.” Ariz. Indep. Redistricting Comm’n v. Brewer, 229 Ariz. 347, 275 P.3d 1267, 1273 (2012). But the very structure of Arizona’s reformed redistricting process reflects that partisanship still plays a prominent role. In practice, the Arizona Constitution requires two commissioners be Republicans, two commissioners be Democrats, and the fifth commissioner be *1084neither a Republican nor a Democrat.5 The fact that one’s party affiliation is a qualifying characteristic to serve as a commissioner is at least an implicit acknowledgment that redistricting remains inextricably intertwined with partisan concerns.
Recognizing that partisanship remains an inevitable ingredient in Arizona’s redistricting scheme is not the same as saying redistricting decisions actually based on partisanship are immune from challenge. Under the federal constitution, it may be possible to challenge redistricting plans when partisan considerations go “too far.” See Cox v. Larios, 542 U.S. 947, 952, 124 S.Ct. 2806, 159 L.Ed.2d 831 (2004) (Scalia, J., dissenting) (noting most Justices believed partisanship “is a traditional criterion, and a constitutional one, so long as it does not go too far”). But it is presently obscure what “too far” means. It is highly improbable that any use of partisanship is “too far.” However, maybe partisanship can be used to justify population deviations below ten-percent but not above ten-percent. Or maybe it is unconstitutional to make decisions based on partisanship only if those decisions have “an actual discriminatory effect on” a particular political group. Cf. Davis v. Bandemer, 478 U.S. 109, 127, 106 S.Ct. 2797, 92 L.Ed.2d 85 (1986) (attempting to establish standard for “political gerrymandering” claim). The Supreme Court has not yet indicated which of these possibilities, if any, is correct. And the one case plaintiffs repeatedly rely upon to support their theory cannot bear nearly the weight they wish.
Plaintiffs believe Larios v. Cox, 300 F.Supp.2d 1320 (N.D.Ga.2004) “cast ex-treme doubt on whether partisanship alone ever could justify deviations from population equality.” But a brief exploration of the facts, legal holdings, and subsequent history of Larios show plaintiffs’ reliance is not well-placed.
In Larios, a three judge panel addressed the map drawn by the Democratic majority in the Georgia General Assembly. After considering the evidence, the court clearly identified the Democrat legislators as having “made no effort to make the districts as nearly of equal population as was practicable.” Id. at 1341. Instead, the Democrats had entered the redistricting process under the assumption they were free to manipulate the maps however they wished, provided the final population deviations were kept below ten percent. With that assumption in mind, the final map contained population deviations of 9.98%. Id. In addition, the Democrats refused to allow Republican legislators meaningful involvement in the process. Id.
The record made “abundantly clear that the population deviations in the Georgia House and Senate” were driven by two prohibited considerations. Id. at 1341. First, the deviations were a “concerted effort to allow rural and inner-city Atlanta regions of the state to hold on to their legislative influence (at the expense of suburban Atlanta), even as the rate of population growth in those areas was substantially lower than that of other parts of the state.” Id. at 1342. And “[sjecond, the deviations were created to protect incumbents in a wholly inconsistent and discriminatory way.” Id. In reaching these conclusions, the Larios court stressed it was not required to “resolve the issue of whether or when partisan advantage alone may justify deviations in population, because ... *1085the redistricting plans [were] plainly unlawful” on other grounds. Id. at 1352.
The Supreme Court summarily affirmed Larios. Cox v. Larios, 542 U.S. 947, 124 5.Ct. 2806, 159 L.Ed.2d 831 (2004). That summary affirmance meant the Supreme Court agreed with the judgment “but not necessarily the reasoning by which it was reached.” Mandel v. Bradley, 432 U.S. 173, 176, 97 S.Ct. 2238, 53 L.Ed.2d 199 (1977) (quotation omitted). -In other words, the summary affirmance “should not be understood as breaking new ground but as applying principles established by prior decisions to the particular facts involved.” Id. There are no prior decisions directly rejecting partisanship as a justification for minor population deviations, meaning the summary affirmance has little value on that issue. But Justice Scalia voted to set the case for argument, likely out of a concern the lower court decision would be read as addressing the issue. As explained by Justice Scalia, the Supreme Court has never made clear whether “politics as usual” is a “ ‘traditional’ redistricting criterion” that can be used to justify minor population deviations. Larios, 542 U.S. at 952, 124 S.Ct. 2806 (J. Scalia, dissenting). Justice Scalia also noted that, in a case the previous term, “all but one of the Justices agreed [partisanship] is a traditional criterion, and a constitutional one, so long as it does not go too far.” Id.
With the lower court’s explicit refusal to address the partisanship issue, and the Supreme Court’s summary affirmance, I doubt Larios offers any useful guidance on the question of partisanship.6 Absent other instructive authority supporting, their claim, we might have been better served by dismissing plaintiffs’ federal claim for failure to state a claim on which relief can be granted. See Cecere v. County of Nassau, 274 F.Supp.2d 308, 313 (E.D.N.Y.2003) (granting motion to dismiss because an allegation of “rank partisanship by the Democratic majority ... is not violative of the Fourteenth Amendment”). But having allowed plaintiffs to survive the motion to dismiss, we must now reach the merits. Fortunately, we need not decide whether partisanship can be considered in redistricting because, in fact, partisanship was not behind the final map. Unfortunately, reaching the merits required a lengthy trial and a tremendous expenditure of resources. If plaintiffs’ theory is viable, and maps containing minor deviations can be challenged as attempts to give one political party an electoral advantage, the federal courts should prepare to be deluged with challenges to almost every redistricting map. If that course is before us, a decision by the Supreme Court on whether this theory is viable, and if so when, would be welcomed.
3. Standard Applicable to Plaintiffs’ Claim
Assuming minor population deviations due to partisanship present a cognizable Equal Protection claim, the question is what standard applies to such a claim. I believe the correct standard is that plaintiffs were required to prove partisanship was the actual and sole reason for the population deviations.
In their initial filings, plaintiffs explicitly agreed they needed to show the “sole reason” behind the population deviations was partisanship.7 All three judges seemingly *1086agreed because, in resolving the motion to dismiss, we set forth the standard as requiring plaintiffs “prove that ‘the asserted unconstitutional or irrational state policy is the actual reason for the deviation.’ ” The opinion we relied on, Rodriguez v. Pataki, further explains a plaintiff must show “the deviation in the plan results solely from the promotion of an unconstitutional or irrational state policy.” 308 F.Supp.2d 346, 365 (S.D.N.Y.2004) (quoting Marylanders for Fair Representation, Inc. v. Schaefer, 849 F.Supp. 1022, 1032 (D.Md. 1994)). Thus, from the very beginning of this case, plaintiffs were on notice — and they did not seem to dispute — that they needed to establish partisanship was the actual and sole reason for the population deviations.
As the case developed, plaintiffs apparently were enlightened and rethought their stance by beginning to describe the standard as requiring they show “no constitutional goal justified” the population deviations. In connection with that softened burden, plaintiffs also, much to defendants’ frustration, began to substantively change their theory of the case such that partisanship was advanced merely as the “principal theory,” along with other prohibited characteristics such as race being implicated. But despite plaintiffs vacillations, I always understood their case as based on the allegation that partisanship drove the entirety of the redistricting process.8
By the time of trial, plaintiffs were again describing their claim as grounded on a belief that partisanship was the “sole” explanation for the population deviations. See Plaintiffs’ Proposed Findings of Fact (Final Map was created “for the sole purpose of providing Democratic candidates with partisan advantage”); Plaintiffs’ Trial Brief (“The IRC systematically underpopulated Republican plurality districts and over-populated Democratic plurality districts for the sole purpose of providing Democratic candidates with a partisan advantage ....”) (emphasis added). The Final Pretrial Order we approved accepted this framing, describing the case as requiring resolution of whether the population deviations were done “for the sole purpose of partisanship.” I am not aware of any clear request by plaintiffs that we adopt something other than the “actual and sole reason” standard. And I believe there are compelling reasons for retaining this very high standard on this type of claim.
Adopting a lower standard on this type of claim invites individuals “to challenge any minimally deviant redistricting scheme based upon scant evidence of ill will by district planners, thereby creating costly trials and frustrating the purpose of [the Supreme Court’s] ‘ten percent rule.’ ” Rodriguez, 308 F.Supp.2d at 365. Federal court challenges to redistricting plans are not only expensive and very time-consuming, they are also “a serious intrusion on the most vital of local functions.” Miller v. Johnson, 515 U.S. 900, 915, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995).
Moreover, the bright-line standard of requiring plaintiffs establish the actual and sole reason behind redistricting decisions is workable. Under this standard, a court need not engage in the formidable task of divining which reason “predominated” over the myriad of possible reasons presented *1087by those defending a new map. Instead, a court must simply determine whether the map was drawn solely for an illegitimate reason. If other reasons were involved, that ends the case.
Plaintiffs repeatedly stated they would establish partisanship as the actual and sole reason for the population deviations and we adopted that as the standard plaintiffs needed to meet. I believe that remains the appropriate standard.
4. No Evidence of Partisanship
The history of the redistricting process, as well as when and who ordered various map changes, are documented in the record and not subject to dispute. Therefore, I join most of the factual findings in the per curiam opinion. I cannot, however, join those findings pointing to partisanship as motivating certain actions. I do not believe plaintiffs carried their burden of establishing that partisanship, rather than neutral redistrieting criteria, motivated the Commission.
The final map comes to us with a “presumption of good faith.” Miller v. Johnson, 515 U.S. 900, 916, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995). It was never clear to me how plaintiffs planned to overcome this presumption. Plaintiffs made general allegations about a plan to harm the interests of the Republican party but they never specified who was allegedly behind the plan.9 At various points during the litigation, it appeared plaintiffs believed the Commission’s counsel, the Commission’s experts, the Commission’s mapping consultant, and even the Republican commissioners themselves, were all motivated by the desire to systematically harm the Republican party’s electoral chances.10 And even having sat through the trial, it remains unclear to me whether plaintiffs were trying to prove a knowing plot amongst all these actors or coincidental uncoordinated acts of partisan discrimination that occurred merely by happenstance. But regardless of who plaintiffs believed was responsible, I did not see sufficient evidence that anyone set out to harm the Republicans. And certainly not enough evidence to establish the Commission as an entity did so.
a. The Alleged Plot Failed
Before directly addressing why I believe plaintiffs failed to prove their case, it is worth noting that the 2012 election using the new map proved their theory has no basis in reality. In the 2012 elections, Republicans won 17 out of 30 (56.6%) senate seats and 36 out of 60 (60%) house seats. As of June 2012, Republicans had a statewide two party registration share of 54.4%. Thus, under the map plaintiffs believe was created to systematically harm Republican electoral chances, Republicans are overrepresented in the legislature. In other words, assuming the relevant actors drew the map to harm the Republican party’s electoral chances, the evidence shows the actors failed to achieve their goal. Because this is not a political gerrymandering case, these results are not nec*1088essarily fatal to plaintiffs’ case. See Davis v. Bandemer, 478 U.S. 109, 127, 106 S.Ct. 2797, 92 L.Ed.2d 85 (1986) (political gerrymandering claim requires proof of “actual discriminatory effect”). But it is hard to take plaintiffs’ challenge seriously given that the alleged contrivance against Republicans failed. See Adam Raviv, Unsafe Harbors: One Person, One Vote and Partisan Redistricting, 7 U. Pa. J. Const. L. 1001, 1062 (2005) (“And certainly it makes sense not to overturn a plan that, whatever the intent of the planners, did not actually hurt their political opponents.”).
b. No Explanation for Choosing Harder Path
Beyond having a theory not grounded in actual harm to a particular political party, plaintiffs also failed to offer any coherent explanation why the Commission would have chosen such an elaborate and difficult way to advantage the Democratic party. That is, assuming everyone involved in the redistricting process was driven solely by a desire to advantage Democrats over Republicans, they had a much easier path available to them than engaging in the complicated task of minor population deviations: the Commission could have set up districts of equal population but drawn the district boundaries differently. See Gaffney v. Cummings, 412 U.S. 735, 753, 93 S.Ct. 2321, 37 L.Ed.2d 298 (1973) (“[I]t requires no special genius to recognize the political consequences of drawing a district line along one street rather than another.”). That would have resulted in far greater partisan impact and the approach would have had the added benefit of being almost impossible to challenge. See, e.g., League of United Latin Am. Citizens v. Perry, 548 U.S. 399, 126 S.Ct. 2594, 165 L.Ed.2d 609 (2006) (rejecting political gerrymandering claim). It is not sensible to conclude everyone involved in the process — or at least whomever plaintiffs believe are responsible for the alleged discrimination — decided to adopt a method that was less effective and more susceptible to challenge than an obvious and available alternative.
c. Insufficient Evidence of Partisanship
Turning to the merits of plaintiffs’ claim, the evidence is overwhelming the final map was a product of the commissioners’s consideration of appropriate redistricting criteria. In particular, the commissioners were concerned with obtaining preclearance on their first attempt.11 Before this round of mapping, Arizona had never obtained preclearance on its first legislative map. Therefore, the focus on first-attempt-preclearance was reasonable given that, at that time, any failure to obtain preclearance on the first attempt would have meant Arizona could not “bail out” of Section 5 of the Voting Rights Act for another ten years. 42 U.S.C. § 1973b(a)(l)(E); Nw. Austin Mun. Util. Dist. No. 1 v. Holder, 557 U.S. 193, 199, 129 S.Ct. 2504, 174 L.Ed.2d 140 (2009) (explaining “bail out” requirements). In these circumstances, the commissioners were not content to make simply a plausible case for preclearance; rather, the commissioners set out to make the absolute strongest possible showing for preclearance.
To present the best preclearance case possible, the Commission’s counsel and consultants recommended ten minority ability-to-elect districts. The Commission agreed with that advice and the draft map contained ten districts identified by the Commission as ability-to-elect districts. Plaintiffs presented no convincing evidence this advice was the result of a conscious effort to harm Republicans. In fact, it is *1089not even clear whether plaintiffs contend' the draft map was the result of partisanship. But if partisanship actually were at the heart of the draft map, and assuming the Republican commissioners were not Democratic sleeper-agents, one would expect the record to be replete with objections by the Republican commissioners. It is not. I view the Republican commissioners’ silence as evidence that partisanship was not the driving force behind the draft map.
With no credible evidence the draft map was drawn to favor the Democratic party, the focus turns to whether the changes to the draft map were motivated by partisanship or if they can be explained on some other ground. Again, the vast majority of the changes to the draft map were agreed to by the Republican commissioners. And as observed by Commissioner Mathis, all of the commissioners are “very strong people” who would have spoken up if they had an objection. I do not believe we are in a better position to divine invidious discrimination than the partisan actors actually involved in the process.
Much more important than the relative lack of objections is that plaintiffs did not identify, with reasonable particularity, the exact changes to the final map they believe were due solely to partisanship. Plaintiffs initially seemed to be claiming every aspect of the final map was due to partisanship. However, at trial and in their post-trial briefing, they focused primarily on three districts: Districts 8, 24, and 26. The per curiam opinion explains some of the changes to Districts 24 and 26 and why the Commission believed compliance with the Voting Rights Act supported such changes. While plaintiffs disagree with those actions, I did not see any evidence that partisanship, rather than compliance with the Voting Rights Act, was the actual reason for the changes in Districts 24 and 26.
As for District 8, the per curiam opinion concludes partisanship did motivate certain changes. At trial, however, Commissioner McNulty explained those changes were meant to make District 8 more competitive. I found her explanation reasonable and credible. Also, when asked squarely whether these particular changes were due to any reason other than competitiveness and compliance with the Voting Rights Act, Commissioner McNulty said no. Again, I found her testimony credible. I would require much more evidence than what plaintiffs presented to conclude Commissioner McNulty was being untruthful in her trial testimony. More importantly, even if Commissioner McNulty did make changes to District 8 with partisanship in mind, that is not enough.
Evidence that one commissioner was motivated by partisanship is only a good starting point and it is a given that four of the five commissioners always have at least some partisan self-interest. There must be evidence that two other commissioners had that same motivation. But the Supreme Court has cautioned that “inquiry into legislative motive is often an unsatisfactory venture” because “[wjhat motivates one legislator to vote for a statute is not necessarily what motivates ... others to enact it.” Pac. Gas and Elec. Co. v. State Energy Res. Conservation, 461 U.S. 190, 216, 103 S.Ct. 1713, 75 L.Ed.2d 752 (1983). Thus, even if Commissioner McNulty was motivated by partisanship, plaintiffs would still need to show two commissioners voted with Commissioner McNulty “at least in part ‘because of,’ not merely ‘in spite of,’ ” the alleged adverse effects that particular change would have on Republicans. Pers. Adm’r of Mass. v. Feeney, 442 U.S. 256, 279, 99 S.Ct. 2282, 60 L.Ed.2d 870 (1979). I saw no such evidence.
*1090In the end, Plaintiffs’ evidence of partisanship consisted largely of pointing to the final map and asking the Court to conclude by inference only that the pattern reflected in the map established an intent to discriminate against Republicans.12 This appears to be an attempt to invoke the “disparate impact” theory of liability. But only in exceptionally rare cases is disparate impact enough to prove an Equal Protection violation. See, e.g., Washington v. Davis, 426 U.S. 229, 242, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976) (“Disproportionate impact is not irrelevant, but it is not the sole touchstone of [invidious discrimination] forbidden by the Constitution.”). Those rare cases involve situations of a clear pattern unexplainable on any legitimate grounds. See, e.g., Vill. of Arlington Heights v. Metro. Hous. Corp., 429 U.S. 252, 266, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977) (“Sometimes a clear pattern, unexplainable on grounds other than race, emerges from the effect of the state action ....”) (emphasis added).' Here, the final map’s population deviations can be explained on grounds other than partisanship.
The final map represents an attempt to satisfy legitimate redistricting criteria, especially the Voting Rights Act. As observed in the per curiam opinion, “changes that strengthened minority ability-to-elect districts were also changes that improved the prospects for electing Democratic candidates.” In other words, the changes the Commission made to strengthen its case for complying with the Voting Rights Act also had the effect of improving Democratic prospects. In light of this, the alleged pattern in the final map easily is explainable on grounds other than partisanship.
I join the judgment against plaintiffs.
. As noted in the February 22, 2013 Order, I disagreed with the resolution of the motion for protective order. The case has now proceeded to trial and the commissioners testified at length. In these circumstances, I do not believe it necessary to set forth why I would have granted the protective order in part.
. I recognize that redistricting cases pose a unique abstention problem. In the normal Pullman setting, the federal court stays the federal claim and, if the parties are not able to obtain timely relief in state court, they can return to federal court to litigate their federal claim. Cf. Harris Cnty. Comm'rs Court v. Moore, 420 U.S. 77, 84, 95 S.Ct. 870, 43 L.Ed.2d 32 (1975) (abstention not appropriate when litigation already "long delayed”). But under Supreme Court precedent applicable to redistricting suits, if plaintiffs had been forced to file in state court, we would have been absolutely barred from proceeding on the federal claim until the state court litigation con-eluded. Growe v. Emison, 507 U.S. 25, 33, 113 S.Ct. 1075, 122 L.Ed.2d 388 (1993). Plaintiffs did not provide any persuasive reason why this complication would matter because the state court would have had ample time to address the state-law claim before any harm was suffered.
. The state-law claim was formally dismissed at the same time the abstention motion was denied. Thus, even if a pending state-law claim is a necessary prerequisite to abstention, it was met at the relevant time.
. Because the Eleventh Amendment barred the state-law claim, plaintiffs’ alternative request to certify the state-law issue to the Arizona Supreme Court was correctly denied. It would have been a futile gesture to certify the question because we could not have ordered relief on the basis of state law, regardless of how the Arizona Supreme Court might have ruled. See Citizens for John W. Moore Party v. Bd. of Election Comm’rs, 781 F.2d 581, 584-86 (1986) (Easterbrook, J., dissenting) (noting that certification is not appropriate when the Eleventh Amendment means relief cannot be granted on basis of state law).
. The Arizona Constitution requires the twenty-five candidates for commissioner consist of “ten nominees from each of the two largest political parties in Arizona based on party registration, and five who are not registered with either of the two largest political par-lies.” Ariz. Const, art. IV, pt. 2, § 1(5).
. In 2006, Justice Kennedy explained that the Larios district court opinion did not give “clear guidance” on when partisanship can justify population deviations. League of United Latin Am. Citizens v. Perry, 548 U.S. 399, 423, 126 S.Ct. 2594, 165 L.Ed.2d 609 (2006).
. Plaintiffs' filings could not have made it any clearer that they conceded the issue was *1086whether partisanship was the "sole” cause for the population deviations. See Plaintiffs' Response in Opposition to Motion to Dismiss ("[Defendants] diluted Plaintiffs' votes and the votes of all citizens residing in the overpopulated districts solely to maximize the Democratic Party’s representation in the Legislature.”).
. As described on the last day of the trial, plaintiffs’ theory was that "this pattern of deviation was driven by partisanship.”
. Plaintiffs also had difficulty identifying what would be a sufficient reason for the population deviations at issue. For example, plaintiffs’ complaint recognized "compliance with the Voting Rights Act” was a "legitimate state interest.”
. Plaintiffs’ expert also had significant difficulty deciding who was behind the plan to harm Republicans. Originally, the expert stated "the individuals who were drawing the maps for the Commission were engaged in intentional political gerrymandering.” (Trial Tr. 677). At trial, the expert abandoned that position. (Trial Tr. 677, 685). Later, the expert agreed that one of the Republican commissioners had "engaged in invidious discriminatory vote dilution” to benefit the Democratic party. (Trial Tr. 719).
. Plaintiffs' counsel conceded obtaining pre-clearance was a legitimate state interest.
. Plaintiffs repeatedly claimed the partisan breakdown of the final population deviations could not be explained by chance. Of course, there is no claim that the map was designed at random, meaning the argument that the deviations could not have occurred by chance is trivial. More importantly, plaintiffs fail to take account of a basic problem always presented in cases where the court is asked to infer intent based on statistics: “statistics demonstrating that chance is not the more likely explanation are not by themselves sufficient to demonstrate that [reliance on the prohibited characteristic] is the more likely explanation.” Gay v. Waiters' and Dairy Lunchmen's Union, 694 F.2d 531, 553 (9th Cir.1982). In other words, a statistical aberration negating chance is very different from a statistical aberration establishing invidious intent.