MEMORANDUM OPINION AND ORDER
WILLIAM PRYOR, Circuit Judge:The Alabama Legislature faced a difficult task in 2012. The Fourteenth Amendment requires state legislative districts of roughly equal population and prohibits racial gerrymandering. But the Voting Rights Act required Alabama to avoid retrogressing the ability of black voters to elect candidates of their choice. In other words, the legislature had to draw districts of roughly equal population that were conscious enough of race to comply with the Voting Rights Act, but not so conscious of race that they violated the Fourteenth Amendment. In the process, the legislature had to resolve conflicts between traditional districting criteria and secure enough votes to pass both houses. And to further complicate matters, most of the existing majority-black districts were underpopulated by at least five percent.
After the legislature enacted a plan, the Alabama Legislative Black Caucus and the Alabama Democratic Conference sued Alabama for violating the Fourteenth Amendment and the Voting Rights Act. We granted judgment for Alabama after a four-day bench trial. The Supreme Court vacated that judgment and remanded to allow the plaintiffs to reargue their claims of racial gerrymandering and present new evidence.
*1033The plaintiffs now challenge all thirty-five majority-black districts and House District 85 as racial gerrymanders. We accepted new evidence, ordered briefing, and held oral argument. At our request, the plaintiffs agreed to draw alternative plans that complied with federal and state law and to submit briefing on the plans. Alabama deposed the plaintiffs’ experts and submitted its own briefing. We imposed no page limits on any of the briefing.
We have readopted our earlier decisions resolving all claims that the Supreme Court did not address, (Doc. 242), and we now decide the claims of racial gerrymandering. To succeed on a claim of racial gerrymandering, the plaintiffs must prove that “race [was] the ‘dominant and controlling’ or ‘predominant’ consideration in deciding ‘to place a significant number of voters within or without a particular district.’ ” Ala. Legislative Black Caucus v. Alabama, — U.S. -, 135 S.Ct. 1257, 1264, 191 L.Ed.2d 314 (2015) (quoting Miller v. Johnson, 515 U.S. 900, 913, 916, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995)). Race predominated over traditional districting criteria if it “was the criterion that, in the State’s view, could not be compromised.” Shaw v. Hunt, 517 U.S. 899, 907, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996). If the plaintiffs prove that race predominated, then the defendants must prove that they had a “strong basis in evidence,” Ala. Legislative Black Caucus, 135 S.Ct. at 1274, that the use of race was “narrowly tailored to serve a compelling state interest,” Shaw, 517 U.S. at 907-08, 116 S.Ct. 1894. A strong basis in evidence consists of “good reasons to believe such use is required, even if a court does not find that the actions were necessary for statutory compliance.” Ala. Legislative Black Caucus, 135 S.Ct. at 1274.
The plaintiffs argue that race predominated when the drafters kept the black population percentage in a district the same as it was before redistricting, but more is necessary under Supreme Court caselaw. It is possible to hit a supposed target solely by considering traditional dis-tricting criteria, as the plaintiffs concede when their alternative plans match the previous black population percentage in a district. The plaintiffs instead must provide evidence of how the drafters subordinated traditional districting criteria to race. We consider all of the evidence offered by the parties on remand, and we have no mechanical formula or system of weights for considering this evidence.
We find that race did not predominate in 22 of the 36 districts, and with respect to those districts, our inquiry ends there. We also find that race predominated in 14 of the 36 districts, and we must next decide whether those districts survive strict scrutiny.
We conclude that Alabama has satisfied strict scrutiny in two of the districts where race predominated. Alabama asserts an interest in complying with the Voting Rights Act, and it relies primarily on statements by two incumbent members of the Black Caucus at public meetings of the redistricting committee. This evidence is sufficient in those members’ districts. As we explain, the Supreme Court does not require that the legislature conduct studies. It instead requires only that the legislature had a strong basis in evidence for its use of race. The statement of Senator Hank Sanders in particular is detailed and based on his experience as an influential longtime incumbent. This kind of testimony constitutes a “strong basis in evidence.” And despite the plaintiffs’ insistence to the contrary, the record does not establish that the drafters had an incorrect understanding of section 5 in these two districts.
We GRANT judgment for the plaintiffs with respect to Senate District' 20, Senate *1034District 26, Senate District 28, House District 32, House District 53, House District 54, House District 70, House District 71, House District 77, House District 82, House District 85, and House District 99, and we ENJOIN the use of these districts in future elections. With respect to the other 24 districts, we GRANT judgment for the defendants.
TABLE OF CONTENTS
I. BACKGROUND ...1034
A. The Parties ... 1034
B. Relevant Factual Background ...1035
C. Decision of the Supreme Court ...1041
D. Subsequent Proceedings ... 1043
II. FINDINGS OF FACT AND CONCLUSIONS OF LAW .. .1044
A. Standing ... 1045
B. Majority-Black Districts ... 1045
C. The Plaintiffs’ Court-Ordered Alternative Plans ... 1046
D. Racial Predominance Generally ...1049
a. The ±1% Deviation Does Not Prove that Race Predominated. ... 1049
b. Alabama Had a Statewide Policy of Racial Targets, but the Plaintiffs Still Must Prove that the Policy Caused Race to Predominate in Individual Districts ...1050
c. The Dissent Misstates the Test for Racial Predominance. ... 1056
d. The Plaintiffs Must Prove that the Precinct Splits Cannot Be Explained by Traditional Districting Criteria. ... 1058
E. Strict Scrutiny Generally .. .1061
F. District>-By-Distriet Analysis of the Challenged Plan ... 1064
a. Senate Districts 18, 19, and 20 (Birmingham) ...1065
b. Senate District 23 (West Black Belt) ...1082
c. Senate District 24 (West Black Belt) ...1111
d. Senate District 26 (Montgomery) ...1129
e. Senate District 28 (East Black Belt) .. .1140
f. Senate District 33 (Mobile) ... 1158
g. House Districts 19 and 53 (Madison County) .. .1165
h. House District 32 ... 1185
i. House Districts 52, 54, 55, 56, 57, 58, 59, and 60 (Jefferson County) ...1192
j. House Districts 67, 68, 69, 70, 71, and 72 (West Black Belt) ... 1221
k. House Districts 76, 77, and 78 (Montgomery) ... 1279
l. House Districts 82, 83, 84, and 85 (East Black Belt) ... 1297
m. House Districts 97, 98, 99, and 103 (Mobile County) ... 1320
CONCLUSION .. .1348
APPENDIX ...1349
I. BACKGROUND
We divide our discussion of the background in four parts. First, we identify the parties. Second, we explain the history of the redistricting process as it relates to the claims of racial gerrymandering. Third, we review the decision of the Supreme Court in Alabama Legislative Black Caucus v. Alabama, — U.S. -, 135 S.Ct. 1257, 191 L.Ed.2d 314 (2015). Fourth, we discuss the proceedings that occurred after the decision of the Supreme Court.
A. The Parties
In this opinion, we divide the parties in three groups. The first group is the Black Caucus plaintiffs: the Alabama Legislative Black Caucus, the Alabama Association of *1035Black County Officials, Fred Armstead, George Bowman, Rhondel Rhone, Senator Bobby Singleton, Albert F. Turner, and Jiles Williams Jr. The second group is the Democratic Conference plaintiffs: the Alabama Democratic Conference, Lynn Pett-way, Stacey Stallworth, Rosa Toussaint, and Framon Weaver Sr. We refer to the third group as “Alabama” or “the defendants,” and they are Alabama, Governor Robert J. Bentley, Representative Randy Davis, Senator Gerald Dial, Representative Jim McClendon, and Secretary of State John H. Merrill.
B. Relevant Factual Background
This litigation has a deeply partisan backstory. After the 2000 Census, the Democrat-controlled legislature adopted redistricting plans that were expressly partisan. Montiel v. Davis, 215 F.Supp.2d 1279, 1283 (S.D. Ala. 2002) (“Plaintiffs have proffered no evidence to refute the abundant evidence ... that [the redistricting plans] were the product of the Democratic Legislators’ partisan political objective to design Senate and House plans that would preserve their respective Democratic majorities.”). The redistricting criteria in 2001 required that the population in a given district be within ± 5% of the ideal population of a district. Id. Within that range, the 2001 redistricting plans systematically underpopulated Democratic districts, including majority-black districts. Out of the 26 majority-black House districts, 21 were underpopulated, and 11 were underpopulated by greater than 4 percent. (Doc. BO-42 at 3-4). Six of the eight majority-black Senate districts were underpopulated, and four of those districts were underpopulated by greater than four percent. (Doc. 30-44 at 2).
The Democratic leaders boasted about their partisan strategy. They filed an ami-cus brief in the Supreme Court of the United States that described the districts as an example of a successful partisan gerrymander. See Brief for Leadership of the Alabama Senate and House of Representatives as Amici Curiae Supporting Ap-pellees, Vieth v. Jubelirer, 541 U.S. 267, 124 S.Ct. 1769, 158 L.Ed.2d 546 (2004) (No. 02-1580) (Def. Ex. 448). The brief explained that, during the redistricting process after the 2000 Census, “the Democratic leadership pursued a biracial strategy aimed at safeguarding its governing majorities in both houses of the Legislature.” Id. The brief pronounced that the partisan strategy had succeeded: “The 2002 general election returned Democratic candidates to 71% of the Senate seats and 60% of the House seats, with 52% of the statewide vote supporting Democrats in Senate races and 51% supporting Democrats in House races.” Id.
Unsurprisingly, Republicans were not enthused. They challenged the 2001 districts as racial gerrymanders in federal court, but Alabama successfully defended the population deviations as “the product of the Democratic Legislators’ partisan political objective to design Senate and House plans that would preserve their respective Democratic majorities.” Montiel, 215 F.Supp.2d at 1283. After the Republicans’ complaint of racial gerrymandering failed, they filed another complaint that challenged the population deviations as an unlawful partisan gerrymander, but that complaint failed because it was barred by res judicata, Gustafson v. Johns, 434 F.Supp.2d 1246, 1255, 1267 (S.D. Ala. 2006).
The 2010 Census revealed severe malap-portionment of population among the districts, especially in the majority-black House districts that the Democrat-controlled legislature had drawn in 2001. In the 2010 Census, all of the 26 House districts that were majority-black in 2001 were underpopulated. (Doc. 30-37). Twen*1036ty-four of those districts were underpopulated by more than 5 percent, the maximum deviation allowed under the 2001 plans, and nine were underpopulated by more than 20 percent. (Id.). All eight of the Senate districts that were majority-black in 2001 were underpopulated, seven of them by more than 5 percent and two of them by more than 20 percent. (Doc. 30-41). Many of these malapportioned districts were located in the Black Belt, a south-central region of the State named for its black soil. Many black Alabamans reside there due to the region’s history of agriculture and slavery. (Doc. 203 at 18). The following tables illustrate the severity of the underpopulation in the challenged districts:
Population Deviation in Challenged Senate Districts
[[Image here]]
(Doc. 30-41).
Population Deviation in Challenged House Districts
*1037[[Image here]]
(Districts with a * were not majority-black in 2001).
(Doc. 30-37).
The 2001 partisan gerrymander failed to save the Democrats in 2010, when Republicans won supermajorities in both houses. Because the Alabama Constitution requires the Alabama Legislature to update its districts after each decennial census, see Ala. Const. Art. IX, §§ 199-200, the task of responding to the population ma-lapportionment in the districts fell to the newly elected Republican-controlled legislature. The Alabama Code provides for a Permanent Legislative Committee on Reapportionment to address any problems of malapportionment that arise after a new census. See Ala. Code §§ 29-2-50, 29-2-51. The Committee is charged with developing new reapportionment plans for the state. See id. § 29-2-50(2). Two Republicans, Senator Gerald Dial and Representative Jim McClendon, co-chaired the Committee, (Corr. Joint Stip. of Facts at 3), which eventually produced the legislative redistricting plans that the plaintiffs now challenge: Act No. 2012-602 (House) and Act No. 2012-603 (Senate).
The Republican-controlled legislature used the same criteria as previous legislatures, with one exception: they tightened the limit on population deviation to ±1%. Republicans in the legislature used the *1038narrow deviation in part to avoid a judgment of liability under the Fourteenth Amendment as had occurred in Georgia in Larios v. Cox, 300 F.Supp.2d 1320 (N.D. Ga.), aff'd, 542 U.S. 947, 124 S.Ct. 2806, 159 L.Ed.2d 831 (2004). (Doc. 217 at 205-06). They also benefited politically from a tighter deviation, as it would undo the partisan gerrymander that the Democrats had previously enacted. But in all other respects, the Republicans used the same guidelines as the previous plan and attempted to avoid change when possible. (Doc. 215 at 29-30; Doc. 134-4 at 25-26).
Senator Dial and Representative McClendon worked with Randy Hinaman to draw the new districts for the legislature'. (Doc. 125-10 at 2). Hinaman is a political consultant with experience working in Alabama. (Doc. 217 at 115). He drew the congressional districts in Alabama after the 2010 Census, (id. at 116); worked with Democrat leaders after the 2000 Census to draw the congressional districts that were adopted by the legislature and pre-cleared by the United States Department of Justice, (id. at 115); and drew congressional districts that were adopted by a three-judge district court in 1992 and affirmed by the Supreme Court, see Wesch v. Hunt, 785 F.Supp. 1491, 1499 (S.D. Ala.), aff'd sub nom. Camp v. Wesch, 504 U.S. 902, 112 S.Ct. 1926, 118 L.Ed.2d 535 (1992); (Doc. 217 at 114-15).
Hinaman used a computer program called Maptitude to draw the plans. Mapti-tude allows the user to draw districts based on census data. (Doc. 134-4 at 15). It also allows the user to load additional data into the program to assist with the drawing of the districts. (Id.) Hinaman collected political data from the Republican National Committee for every election in Alabama between 2002 and 2010 and imported that data into Maptitude. (Id.). Hinaman also collected and imported information from the Reapportionment Office about the residences of incumbents. (Id. at 36). As he drew the districts, Hinaman had political data down to the precinct level and census data, including racial data, down to the census-block level. (Id. at 110-12).
The Committee gave Hinaman written guidelines for drawing the new district lines, (Corr. Joint Stip. of Facts at 3), which we attach as an appendix to this opinion. As already discussed, the Committee changed the allowable population deviation for the State Board of Education and the legislature from ± 5%, which had been used in the 2001 plans, to ± 1%. (Id. at 3; Doc. 30-4 at 2). The guidelines required the districts to be drawn in accordance with the Voting Rights Act, to be contiguous and reasonably compact, to be composed of as few counties as practicable, to avoid contests between incumbent members whenever possible, and to respect communities of interest. (Doc. 30-4 at 2-4). The guidelines defined a community of interest as “an area with recognized similarities of interest, including but not limited to racial, ethnic, geographic, governmental, regional, social, cultural, partisan, or historic interests; county, municipal, or voting precinct boundariés; and commonality of communications.” (Id. at 3-4). Finally, the guidelines acknowledged that not all of the redistricting goals could be accomplished and provided that, in cases of conflict, priority would be given to the requirements of one person, one vote and the Voting Rights Act. (Id. at 4).
Section 2 of the Act prohibits any “standard, practice, or procedure” that “results in a denial or .abridgement of the right of any citizen of the United States to vote on account of race or color.” 52 U.S.C. § 10301(a). The Supreme Court has applied section 2 to redistricting. See Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). To decide a *1039claim under section 2, a court must first decide whether “(i) ‘[the racial minority] is sufficiently large and geographically compact to constitute a majority in a single-member district’; (ii) the group is ‘politically cohesive’; and (iii) ‘the white majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate.’ ” Reno v. Bossier Parish Sch. Bd., 520 U.S. 471, 479, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997) (alteration in original) (quoting Gingles, 478 U.S. at 50-51, 106 S.Ct. 2752). If these factors, known as the Gingles factors, are present in a district, a court then looks to whether “the totality of the circumstances supports a finding that the voting scheme is dilutive,” id. at 480, 117 S.Ct. 1491, which is to say that members of a protected minority group “have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice,” 52 U.S.C. § 10301(b).
Section 5 of the Act requires a jurisdiction covered under section 4 to obtain preclearance of a new voting “standard, practice, or procedure” by either the Attorney General of the United States or the United States District Court for the District of Columbia. M§ 10304. Such a change can have neither “the purpose nor ... the effect of denying or abridging the right to vote on account of race or color.” Id. The Supreme Court has applied section 5 to redistricting, see, e.g., Beer v. United States, 425 U.S. 130, 133, 96 S.Ct. 1357, 47 L.Ed.2d 629 (1976), and Alabama was a covered jurisdiction in 2012, see Shelby Cty. v. Holder, — U.S. -, 133 S.Ct. 2612, 2621, 186 L.Ed.2d 651 (2013). “Whether a voting procedure change should be precleared depends on whether the change ‘would lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise.’ ” Georgia v. Ashcroft, 539 U.S. 461, 466, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003) (quoting Beer, 425 U.S. at 141, 96 S.Ct. 1357). When the Attorney General evaluated whether the 2012 plan had a retrogressive effect, he compared the new districts with the old districts in the light of the updated census data. See Guidance Concerning Redistricting Under Section 5 of the Voting Rights Act, 76 Fed. Reg. 7471 (Feb. 9, 2011).
The Committee co-chairs, Senator Dial and Representative McClendon, both believed that avoiding retrogression was a priority. Representative McClendon testified that a district “is retrogressed if the minorities in that district, whether by race or language, are worse off after redistricting than they were before redistricting.” (Doc. 217 at 221). Senator Dial testified that, as each majority-black district was repopulated, “to keep from regressing the district and increasing that population, we had to increase it percentagewise on the same number of minority votes that we had.” (Doc. 215 at 36). When asked whether “that included bringing the African American populations of those districts up to approximately equal as best you could with what it had been in 2001,” Dial agreed. (Id. at 37). He was “committed not to regress” the “numbers that had been established under the last 'redistricting plan.” (Id. at 28-29). According to Dial, it was “fair” to say that he tried to “maintain” the black percentage in a given district relative to the 2010 Census population under the 2001 district lines. (Id. at 136). Dial was not concerned with creating higher percentages of black population within a district. (Id. at 56).
Hinaman was also concerned with retrogression. He “look[ed] at [the] 2010 census as applied to 2001 lines, [and] whatever that number was, [he] tried to be as close to that as possible.” (Doc. 217 at 145-46). If he “was significantly below that, [he] *1040was concerned about that being retrogression that would be looked upon unfavorably by the Justice Department under Section 5.” (Id.). Hinaman explained that this inquiry was “somewhat of a subjective thing, but ... if you took a district that was somewhere in the 60 to 65 percent black majority district and you brought it down into the low 50s,” he thought “people would be concerned whether that population would then have the opportunity to elect a candidate of their choice.” (Doc. 134 — 4 at 101-02). When asked about a hypothetical district “in the upper 70s” that was redrawn as 70 percent black, he answered that he “would be less concerned.” (Id.). Hinaman also explained that no one instructed him that such a reduction was forbidden. (Id.).
At the beginning of the reapportionment process, the Committee conducted public hearings at 21 locations throughout Alabama. (Corr. Joint Stip. of Facts at 3). Senator Dial and Representative McClen-don attended all of the hearings. (Id. at 4). At the public hearing in Dallas County, Senator Hank Sanders — a black Democrat who represents majority-black Senate District 23 — asked Dial to use 62 percent as a minimum for the majority-black districts because often the population statistics for a district do not reflect the actual voters in that district. (Doc. 30-28 at 6). At the public hearing in Clarke County, Representative Thomas Jackson — a black Democrat who represents majority-black District 68 — explained that a majority-black district should be 62 percent to 65 percent black. (Doc. 30-23 at 8).
When Hinaman began working on the plans, he drew the majority-black districts first. (Doc. 217 at 146^47). Because every majority-black district was underpopulated, he needed to add precincts to each district until it satisfied the ±1% deviation. (Id. at 142-43). Hinaman explained how he drew those districts:
[I]n toto, whatever I added to a various district, I would look to see what change that made to the overall black percentage in that district. And so in some districts I could add in anything I wanted, and it didn’t matter because they were — you know, either they didn’t need that much population, or the changes I added didn’t matter.
(Id. at 143). Hinaman “tried to look at the additions en masse, not just a precinct.” (Id. at 144). He might “add a white precinct, a majority white precinct and a majority African American precinct; but if you look at the end number, if it did not retrogress the overall end number for that precinct, then they were added in.” (Id.) Hinaman would consider splitting precincts to increase black population percentage only when he was concerned that he had significantly reduced the black population percentage in a given district. (Id. at 144-46).
Hinaman also tried to avoid putting incumbents in conflict with one another, (id. at 119), to accommodate the wishes of incumbents about them districts, (id. at 139), to maintain each district along similar lines, (id. at 162), and to comply with the guidelines set forth by the Committee, (id. at 139). According to Hinaman, “It was also a goal to change each district to some extent the least amount possible.” (Doc. 134^4 at 25-26). Some of these goals had higher priority than others. For example, counties were split in some instances to comply with requests from incumbent legislators, (Doc. 217 at 135-36), or because they were split similarly in previous plans, (Doc. 134 — 4 at 34). Although maintaining each district along similar lines to the previous plan was “a goal,” it was “down on the list” from complying with the Constitu*1041tion and “[separating incumbents.” (Doc. 217 at 162).
Significant portions of the plan were based on suggestions from incumbent legislators. Hinaman traveled to Alabama to meet in person with many of the Republican legislators. (IcL at 120-21). Although he did not meet with Democratic legislators before the plans were introduced in the legislature, he incorporated suggestions that Senator Dial and Representative McClendon received from Democratic legislators. (Doc. 217 at 121). Dial gave Hina-man a proposed map for the three majority-black Senate districts in Jefferson County that Senator Rodger Smitherman, a black Democrat from Jefferson County, had given him. (Doc. 134-4 at 43). Dial instructed Hinaman to incorporate those maps into the Senate plan to the extent possible because they represented the wishes of the three senators from those districts. (Id.) Hinaman drew the majority-black districts in Jefferson County to be substantially the same as the maps provided to him by Senator Dial. (Id.) McClen-don gave Hinaman proposed maps for Montgomery County from Representative Thad McClammy, a black Democrat who represented that county. (Id. at 44-45). McClendon told Hinaman to adopt as many of McClammy’s ideas as possible, and Hinaman followed that instruction. (Id. at 45-47).
There were also alternative plans proposed in the legislature, including the McClammy Plan for the House, (Common Ex. 45), the Sanders Plan for the Senate, (Common Ex. 47), the Reed-Buskey Plans for the House and Senate, (Common Exs. 42, 48), and the Knight Plan for the House, (Common Ex. 46). None of them complied with the ± 1% deviation, and at least some of the plans put numerous incumbents in conflict. For example, the Reed-Buskey Plan for the House put four Republicans and two Democrats in conflict, and the Knight Plan put 26 Republicans and six Democrats in conflict. (Def. Supp. Ex. 16). By contrast, the plans adopted by the legislature created only two incumbent conflicts. Two black incumbent Democrats, Representatives Juandalynn Givan and Demetrius Newton, lived in the new House District 60. But Hinaman had been told that Representative Newton planned to retire, (Doc. 134-4 at 132), and he has since died, (Doc. 203 at 52). Another black incumbent Democrat, Representative John Knight, and a white incumbent Democrat, Representative Joe Hubbard, were both placed in the new House District 77. (Def. Supp. Ex. 16).
Some of the alternative plans also drastically reduced the black population percentage in certain districts. Senator Dial testified that he rejected two alternative plans for the Senate because they did not keep certain majority-black districts at or above 62 percent, as Senator Sanders had requested. Dial explained that he rejected Sanders’s own Senate plan because “[i]t didn’t even meet the requirements [he] had said would keep them 62 percent at least.” (Doc. 215 at 77). Dial likewise rejected the Reed-Buskey Senate Plan because it retrogressed some districts and fell short of 62 percent black population. (Id. at 126).
The legislative plans that the plaintiffs now challenge, Act 2012-602 and Act 2012-603, were introduced, considered, and approved during a special session of the legislature. All of the proposed substitutes were defeated, and Governor Bentley signed the Acts into law on May 31, 2012. The votes to approve the plans fell largely along party lines. (NPX 314-15; Def. Supp. Exs. 21, 25).
C. Decision of the Supreme Court
In 2015, the Supreme Court vacated our previous final judgment upholding the dis*1042tricts and remanded the case. Ala. Legislative Black Caucus, 135 S.Ct. at 1262-63. The Supreme Court “focus[ed] upon four critical District Court determinations”: whether the plaintiffs could bring a statewide claim of racial gerrymandering, whether the Democratic Conference plaintiffs had standing, whether race predominated, and whether the districts survived strict scrutiny. Id. at 1264. We explain each issue in turn.
First, the Supreme Court held that the plaintiffs’ “undifferentiated statewide” claims of racial gerrymandering were “insufficient.” Id. at 1266. It remanded so that the plaintiffs could challenge individual districts as racial gerrymanders. Id. We were instructed to review each challenged district individually, but we were also instructed to consider “statewide evidence to prove that race predominated in the drawing of individual district lines.” Id. at 1267. The Supreme Court further stated that Alabama “expressly adopted and applied a policy of prioritizing mechanical racial targets above all other districting criteria (save one-person, one-vote)” and remanded for a determination of how that policy affected individual districts. Id. at 1267-68.
Second, the Supreme Court remanded with instructions to “reconsider the [Democratic] Conference’s standing by permitting the [Democratic] Conference to file its list of members and permitting the State to respond, as appropriate.” Id. at 1270.
Third, the Supreme Court held that Alabama’s “requirement that districts have approximately equal populations is a background rule against which redistricting takes place.” Id. at 1271. As a result, it “is not one factor among others to be weighed against the use of race to determine whether race ‘predominates.’ ” Id. at 1270. Instead, “it is part of the redistricting background, taken as a given, when determining whether race, or other factors, predominate in a legislator’s determination as to how equal population objectives will be met.” Id. The Supreme Court suggested that had we “not taken a contrary view of the law, [our] ‘predominance’ conclusions, including those concerning the four districts that the Democratic Conference specifically challenged, might well have been different.” Id. at 1271.
Fourth, the Supreme Court clarified the test for strict scrutiny when a state asserts that it had a compelling interest in complying with the Voting Rights Act. The Supreme Court explained that a “mechanical interpretation of § 5 can raise serious constitutional concerns.” Id. at 1273. Instead of asking, “How can we maintain present minority percentages in majority-minority districts?” Alabama must ask, “To what extent must we preserve existing minority percentages in order to maintain the minority’s present ability to elect the candidate of its choice?” Id. at 1274. Alabama must prove that it had “a ‘strong basis in evidence’ in support of the (race-based) choice that it has made.” Id. (quoting Brief for the United States as Amicus Curiae Supporting Neither Party 29, Ala. Legislative Black Caucus, 135 S.Ct. 1257 (Nos. 13-895, 13-1138)). “[Legislators ‘may have a strong basis in evidence to use racial classifications in order to comply with a statute when they have good reasons to believe such use is required, even if a court does not find that the actions were necessary for statutory compliance.’ ” Id (quoting Brief for the United States as Amicus Curiae Supporting Neither Party 29).
The Supreme Court also discussed Senate District 26 in detail. The Court found “strong, perhaps overwhelming, evidence that race did predominate as a factor when the legislature drew the boundaries of Senate District 26.” Id. at 1271. It observed that “[t]he legislators in charge of creating the redistricting plan believed, and told *1043their technical adviser, that a primary redistricting goal was to maintain existing racial percentages in each majority-minority district, insofar as feasible.” Id. And it found “considerable evidence that this goal had a direct and significant impact on the drawing of at least some of District 26⅛ boundaries.” Id. But it made no finding that race predominated in District 26 and instead remanded that issue to this Court, Id. at 1272.
D. Subsequent Proceedings
After we received the mandate from the Supreme Court, we directed the parties to provide us with several pieces of information. First, we directed the plaintiffs to identify the districts that they intended to challenge as racial gerrymanders. (Doc. 225 at 2). They challenged every district with a majority-black total population— House Districts 19, 32, 52-60, 67-72, 76-78, 82-85, 97-99, and 103, and Senate Districts 18-20, 23, 24, 26, 28, and 33. (Doc. 229 at 2). The Black Caucus plaintiffs later abandoned their challenge to House District 84. (Doc. 300-1 at 107).
Second, we directed the Democratic Conference plaintiffs to file a list of members residing in the challenged districts. (Doc. 225 at 2). They did so, (Doc. 238), and Alabama did not file a response to that list.
Third, we asked the parties to identify any new evidence that they hoped to introduce and to consider whether an evidentia-ry hearing would be beneficial. (Doc. 225 at 2). They submitted hundreds of supplemental exhibits based on evidence already introduced at trial. (Doc. 229 at 2-3). The plaintiffs also asked us to take judicial notice of census data, several legislative acts, and the preclearance submissions that Alabama sent to the Department of Justice regarding its 2001 redistricting plans. (Id.). All parties declined the opportunity for an evidentiary hearing. (Id.).
We have readopted our earlier findings of fact and conclusions of law to the extent that the Supreme Court did not address them. We readopted our orders on one person, one vote; partisan gerrymandering; and the claim that the redistricting acts violated the Equal Protection Clause based on the interaction between the Acts and the local legislative system in Alabama. (Doc. 242 at 2 (readopting Docs. 53 and 174)). We also readopted those portions of our previous final judgment that decided the claim of vote dilution brought under section 2 and the claim of intentional discrimination brought under section 2, the Fourteenth Amendment, and the Fifteenth Amendment. (Id. (readopting parts of Doc. 203)). And we readopted our previous findings of fact on all claims other than racial gerrymandering, as well as our previous findings about historical fact and witness credibility. (Doc. 242 at 3). The Black Caucus plaintiffs moved for reconsideration of our previous orders in the light of Arizona State Legislature v. Arizona Independent Redistricting Commission, — U.S. -, 135 S.Ct. 2652, 192 L.Ed.2d 704 (2015), and we denied that motion. (Docs. 261, 265).
Finally, we received briefs in support of final judgment on liability. (Docs. 256, 258, 263, 271-72). Before oral argument, we requested additional information from-the parties, (Doc. 275), which they provided, (Docs. 276-80). At oral argument, we asked the plaintiffs whether they would be willing to submit an alternative statewide plan that complies with federal law and the redistricting criteria adopted by Alabama. They agreed, and we ordered them to do so, (Doc. 283), over the objection of Alabama. The Black Caucus and Democratic Conference plaintiffs each filed a .plan, along with explanatory briefs and several hundred more exhibits. (Docs. 285-87, 294). Alabama deposed both mapmakers and filed their depositions, and it submit*1044ted a response brief and eighteen exhibits. (Docs. 295-97). The plaintiffs then submitted reply briefs with more exhibits. (Docs. 298-301). We imposed no page limits on any briefs on remand.
II. FINDINGS OF FACT AND CONCLUSIONS OF LAW
The Democratic Conference plaintiffs allege that all of the districts that are majority-black by total population are the product of a racial gerrymander; the Black Caucus plaintiffs challenge the same districts, with the exception of House District 84. In the plaintiffs’ view, the drafters maintained a policy of meeting racial targets or floors in each of the districts. According to the plaintiffs, the drafters attempted to replicate the percentage of black population that lived in each challenged district before redistricting. They argue that the use of these targets caused each of these districts to be a racial gerrymander.
To make out a claim of racial gerrymandering, the burden is initially on the plaintiffs to prove that “race was the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.” Miller v. Johnson, 515 U.S. 900, 916, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995). The Supreme Court has explained that the plaintiffs “must prove that the legislature subordinated traditional race-neutral districting principles ... to racial considerations.” Ala. Legislative Black Caucus, 135 S.Ct. at 1270. Traditional considerations include factors such as protecting incumbents, respecting communities of interest, maintaining contiguity and compactness, conforming to political subdivisions, and sorting based on political affiliation. Id. Race predominates over these factors if “[r]ace was the criterion that, in the State’s view, could not be compromised.” Shaw v. Hunt, 517 U.S. 899, 907, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996). When the plaintiffs proceed with only indirect evidence that race predominated and the design of a district can be explained by traditional districting criteria, the plaintiffs have not satisfied their burden of proof.
If the plaintiffs meet their burden, then the defendants must prove that the district satisfies strict scrutiny. “Strict scrutiny does not apply merely because redistricting is performed with consciousness of race. Nor does it apply to all cases of intentional creation of majority-minority districts.” Bush v. Vera, 517 U.S. 952, 958, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996) (citation omitted). The elusive distinction between “being aware of racial considerations and being motivated by them ..., together with the sensitive nature of redistricting and the presumption of good faith that must be accorded legislative enactment, requires [us] to exercise extraordinary caution.” Miller, 515 U.S. at 916, 115 S.Ct. 2475. We “must be sensitive to the complex interplay of forces that enter a legislature’s redistricting calculus.” Id. at 915-16, 115 S.Ct. 2475. “[T]he Constitution does not place an affirmative obligation upon the legislature to avoid creating districts that turn out to be heavily, even majority, minority. It simply imposes an obligation not to create such districts for predominantly racial, as opposed to political or traditional, districting motivations.” Easley v. Cromartie, 532 U.S. 234, 249, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001).
We have already ruled, (Doc. 203 at 160), and the parties do not dispute, that compliance with federal election law, including sections 2 and 5 of the Voting Rights Act, was a compelling governmental interest at the time of redistricting. The Supreme Court has explained that a legislature must have a “strong basis in evi*1045dence” that its district is narrowly tailored to comply with the Act:
This standard ... ‘does not demand that a State’s actions actually be necessary to achieve a compelling state interest in order to be constitutionally valid.’ And legislators ‘may have a strong basis in evidence to use racial classifications in order to comply with a statute when they have good reasons to believe such use is required, even if a court does not find that the actions were necessary for statutory compliance.”
Ala. Legislative Black Caucus, 135 S.Ct. at 1274 (citation omitted) (quoting Brief for United States as Amicus Curiae in Support of Neither Party 29). Put succinctly, this inquiry should not devolve into “a fight over the ‘best’ racial quota.” Id. at 1281 (Thomas, J., dissenting).
The plaintiffs have proved that race predominated in 14 of the 36 districts that they challenge: Senate District 20, Senate District 23, Senate District 26, Senate District 28, House District 32, House District 53, House District 54, House District 68, House District 70, House District 71, House District 77, House District 82, House District 85, and House District 99. Of those 14 districts, only Senate District 23 and House District 68 survive strict scrutiny. We explain our reasoning in the rest of this section, which we divide in six parts. First, we conclude that the Democratic Conference plaintiffs have standing. Second, we make findings about the number of majority-black districts in the challenged plan and the benchmark — the 2010 Census population in the 2001 districts. Third, we discuss the court-ordered alternative plans and the parties’ arguments about them. Fourth, we address the general arguments and evidence about racial predominance. Fifth, we address the general arguments and evidence about strict scrutiny. Sixth, we examine each challenged district individually.
A.Standing
The Supreme Court remanded with instructions to “reconsider the Conference’s standing by permitting the Conference to file its list of members and permitting the State to respond, as appropriate.” Id. at 1270 (majority opinion). In accordance with that mandate, we ordered the Democratic Conference to file its list of members, (Doc. 225 at 2), and the Democratic Conference complied, (Doc. 238). Alabama elected not to file a response to that list, although we gave it the opportunity to do so. (Doc. 237 at 2.) We find that this list establishes that the Democratic Conference has members who reside in all of the challenged districts, and we conclude that this list “is sufficient to meet the Conference’s burden of establishing standing.” Ala. Legislative Black Caucus, 135 S.Ct. at 1269.
B.Majority-Black Districts
Throughout this litigation, the parties often have relied on total population statistics. We have used those statistics because the plaintiffs argued that they prove that race predominated. We will continue using those numbers in this opinion when the parties argue about racial predominance using total population statistics.
But for purposes of the Voting Rights Act, the relevant statistic is voting-age population. See, e.g., 52 U.S.C. §§ 10301, 10304 (referring to the right to vote); Bartlett v. Strickland, 556 U.S. 1, 18, 129 S.Ct. 1231, 173 L.Ed.2d 173 (2009) (opinion of Kennedy, J., joined by Roberts, C.J., and Alito, J.) (applying section 2); Georgia v. Ashcroft, 539 U.S. 461, 485-90, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003) (applying section 5), abrogated on other grounds by 52 U.S.C. § 10304. To avoid confusion, we use “majority-black district” *1046to refer to a district with a majority-black voting-age population, not a majority-black total population. When we refer to a district with a majority-black total population, we will add the words “total population.”
To the extent we earlier readopted any findings about the number of majority-black districts, we now substitute the following findings. The legislature increased the number of majority-black House districts from 26 in 2010 to 27 in Act 602. In 2010, House Districts 19, 32, 52-60, 67-72, 76-78, 82, 83,. 97-99, and 103 were majority-black. (Doc. 35-2). All of these districts are majority-black in Act 602, as is House District 84. (Id.). On the Senate side, the legislature increased the number of majority-black districts from 7 in 2010 to 8 in Act 603. In 2010, Senate Districts 18, 19, 20, 23, 24, 26, and 33 were majority-black. (Doc. 35-3). All of those districts are majority-black in Act 603, as is Senate, District 28. (Id.).
C. The Plaintiffs’ Court-Ordered Alternative Plans
After an initial round of briefs and supplemental exhibits, the plaintiffs had produced several alternative plans, but none complied with the ±1% population deviation set by the Committee. At oral argument, the plaintiffs agreed to draw plans that complied with the state redistricting criteria and federal law. We ordered them to do so over Alabama’s objection. The Black Caucus plaintiffs submitted their 1% Plan, and the Democratic Conference plaintiffs submitted their Plan A. The parties agree that both plans adhere to the ± 1% deviation and have no more precinct splits or incumbent conflicts than the enacted plan. The plaintiffs assert that race did not predominate in their plans and that their districts comply with federal law and the Committee guidelines, but the plaintiffs failed to prove that their plans comply with all of the requirements of federal law. For this reason, we cannot treat the plaintiffs’ districts as conclusive evidence that race predominated in a district or that a district was not narrowly tailored.
The plaintiffs’ mapmakers came dangerously close to admitting that race predominated in at least some of the districts in their plans. The Black Caucus plaintiffs’ expert, William Cooper, testified in his deposition that he intentionally increased the black population in Senate District 9 to create an influence district “because there was an interest that was expressed during that time period in 2012 when I was doing HB — or SB-5 to create a district in — a Senate District in north Alabama, specifically in Madison County, that would have at least some influence.” (Doc. 297-1 at 93-94). He reiterated later in his testimony that he “created an influence district consciously” in Senate District 9. (Id. at 126).
Cooper’s use of race in this manner was not limited to Senate District 9. The black population percentage in District 26 would be over 17 points lower in the 1% Plan than it was in 2010. (Doc. 30-41 at 1; Doc. 296-1 at 4). He explained that he “did consciously lower the black population percentage [in Senate District 26] by extending it out into Lowndes [County] and Au-tauga [County],” (Doc. 297-1 at 129). And in District 32, he testified that he “chose to create” a majority-black district with a black voting-age population of 50 percent and split precincts to do so:
A: Well, according to the court order, I needed to create 27 majority-black districts, and [House District 32] is where I chose to create one of the ones. That, logically, is a little more difficult—
Q: And how did you know when you created a black district?
A: When I had more than 50 percent black voting age. Generally, I would *1047stop at that point because I was working at the precinct level.
Q: Okay.
A: In this case though, precincts had to be split.
(Id. at 124).
The Democratic Conference plaintiffs’ mapmaker, Anthony Fairfax, testified that he also drew some district lines on the basis of race:
A: ... There were certain circumstances where I felt that the districts had an exceedingly high [black population] percentage.... There were some that were 70 percent and higher, and so in that particular case, it bordered packing.
Q: Okay. And so if you came to a district that was 70 percent, would you look at the racial composition of the district as you were drawing it?
A: It was done probably on and off. Yeah, I would say at some times you would have to look at the racial composition. There’s no other way to actually stop from packing if you don’t look at the racial composition.
Q: Would you tend to look at it as you go, as you were adding precincts, or would you draw a district and then look back to see what the composition was?
A: Usually there [were] no racial indicators, but if I saw that we’re getting to 75, 78, 80 percent or something thereabouts, then you have to look at the racial indicators for the voting districts.
Q: Okay. And as you looked at it, you know, what would you do if you get a district that was getting into the high 70s?
A: You would have to move into areas, neighborhoods, that were less black, less African-American, let’s say.
Q: Okay. So let’s say hypothetically you were drawing a district, and you noticed that it was getting to 75 percent or above. You would then start looking for precincts that had a lower percentage of African-American voters?
A: Areas, yes. You move into those areas, not necessarily specific precincts that you grab, but those areas that would actually offset that percentage. There really isn’t any other way to do that.
[[Image here]]
Q: ... Can you tell me now any Senate districts where you got to a high percentage and so you started looking for areas with a lower percentage of African-American voters?
A: There were. I can’t tell you specifically which numbers because it’s a process that you’re working on. It’s not necessarily just one.
(Id. at 78-76).
He later testified that race may have dictated his redistricting choices in House District 68 as well:
Q: So you weren’t looking for population on the borders to try to make sure [House District] 68 stayed above 50 percent [black]?
A: There could have been consideration to make sure that stayed majority minority. Yeah, there could have been consideration to do that.
Q: Are you saying you just don’t remember one way or the other?
A: Right, right. And the reason why is, again, the process of collecting ... the voting districts, there may have been some time where if I did and it dropped below 50, then I want to *1048get it back up. If it dropped into the 40s, then I want to get back up.
Q: What would you do to get back up?
A: Well, I’d have to go into areas that would increase the African-American population.
(Doc. 296-7 at 191-92). But neither of the mapmakers admitted that race predominated over traditional districting criteria in any district.
Moreover, neither of the plaintiffs proved that their alternative plans satisfied the Voting Rights Act. A cursory look at the black voting-age population percentages illustrates why we cannot assume that the alternative districts avoided retrogression and satisfied section 2. For example, the Black Caucus plaintiffs dropped Senate District 23 from 61.79 percent black in the benchmark to 51.06 percent black in the 1% Plan, and they dropped House District 32 from 56.62 percent black to 50.82 percent black. (Doc. 203 at 47-48; Doc. 295 at 18-19). The Democratic Conference plaintiffs took Senate District 26 from 70.87 percent black in the benchmark to 57.70 percent black in Plan A, and they took House District 19 from 67.70 percent black to 50.13 percent black. (Doc. 203 at 47-48; Doc. 295 at 18-19). These districts may in fact satisfy section 2 and section 5 of the Voting Rights Act, but the plaintiffs have failed to prove that they do. Although plaintiffs are not required to produce alternative plans, those plans cannot conclusively prove the unconstitutionality of the challenged plans when we are uncertain whether they violate federal law.
The Black Caucus plaintiffs offer no arguments about how their 1% Plan complies with the Voting Rights Act. The Democratic Conference plaintiffs make several arguments, but they fail to persuade us. First, they cite Dr. Alan Lichtman’s testimony at trial that a bare majority of black voters provides the ability to elect, as well as his testimony about Montgomery County and Madison County. (Doc. 287 at 22-23). We discredited his statewide testimony because “Lichtman did not conduct any statistical analysis to determine whether factors other than race were responsible for the voting pattern,” such as “affluence, strength of a political campaign, or party loyalty.” (Doc. 203 at 79). He “also did not conduct any analysis of Democratic primaries between black and white candidates, which might have offered further evidence about whether white voters are more likely to support white Democrats and black voters are more likely to support black Democrats.” (Id.). The Democratic Conference plaintiffs’ reliance on election returns in uncontested general elections does not remedy that problem. Although we credited the testimony about specific alternative districts in Madison and Montgomery Counties, Lichtman’s testimony does not prove that this new plan — especially the districts in other parts of the state — satisfies the Voting Rights Act. The plaintiffs argue that the districts comply by dint of a coalition with Hispanics, but they fail to prove that Hispanics and blacks form a cohesive coalition.
Second, the Democratic Conference plaintiffs argue for the first time in the brief that explains their plan that the correct statistical category is any-part black, not single-race black. (Doc. 287 at 24). They fail to explain why we should change our metric at this late date, after they provided us with single-race black statistics when attacking the enacted plan. More importantly, they fail to explain how the any-part black statistics would prove that Plan A avoids retrogression; if we use any-part black statistics for Plan A, then we would have to do so for the benchmark as well.
Third, the Democratic Conference plaintiffs argue that, “[e]ven if small increases *1049in black population were proven to be necessary, it would be a simple matter to increase those numbers slightly, consistent with traditional districting criteria.” (Id. at 19). They do not explain why we can be sure that they could raise the black population percentage without subordinating traditional districting principles to race, and their own expert warned of a potential “domino effect” across the state when the shape of one district changes. (Doc. 296-7 at 164, 202). We reject this argument as well.
We also disagree with the argument made by Alabama that “where Plaintiffs’ plans present districts with very similar lines or features to those in the Legislature’s plans, these districts should be affirmed for that reason alone.” (Doc. 295 at 38). If an alternative district has identical lines, we take the plaintiffs’ offering of that plan as a concession that race did not predominate in the enacted district. But where there are significant differences, we must review the record for evidence of racial gerrymandering in the enacted district.
A district is constitutional even if the drafters were concerned about race, so long as race was not “the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.” Ala. Legislative Black Caucus, 135 S.Ct. at 1270 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475). That is, we must determine whether “the legislature subordinated traditional race-neutral districting principles ... to racial considerations.” /¿(alteration in original) (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475). The plaintiffs tell us that they did not subordinate traditional districting criteria to race, so if they made the same choices as the legislature in a district, it is strong evidence that the legislature did not subordinate traditional dis-tricting criteria to race. Alternative plans also can establish that a certain black population percentage was unavoidable based on demographics and traditional districting criteria.
D. Racial Predominance Generally
The parties and the dissent make four general arguments about racial predominance, and we address each one in turn. First, the plaintiffs yet again attack the ± 1% population deviation by arguing that it caused the drafters to subordinate traditional districting principles to racial considerations, but they ignore our opinions and those of the Supreme Court. Second, the plaintiffs argue that the drafters explicitly prioritized the use of racial targets above other considerations in the challenged districts, but this argument does not prove that race predominated in every district. Third, the dissent offers its own version of the standard for racial predominance, but its method is too inflexible. Fourth, Alabama argues that the plaintiffs have failed to establish that any county split or precinct split was the result of racial predominance, but once again this argument does not prove that race predominated in every district,
a. The ±1% Deviation Does Not Prove that Race Predominated.
The Black Caucus plaintiffs once again challenge the population deviation. They contend that splitting counties is evidence of racial gerrymandering because the provision of the Alabama Constitution that requires counties to be kept whole, Ala. Const. Art. IX, § 200, is the “most important” districting principle in Alabama, (Doc. 256 at 12, 15-16), and Alabama split more counties than it would have with a looser population deviation. The Black Caucus plaintiffs argue that “[wjhether a strict ± 1% rule is an ‘appropriate[ ]’ apportionment ... depends on how it interacts with race and traditional districting *1050principles.” (Id. at 17). Because a ±5% deviation is the bare minimum to satisfy federal law, the Black Caucus plaintiffs believe that the decision to better realize one person, one vote is evidence of racial gerrymandering.
This argument is frivolous. Five percent is the constitutional floor, not a ceiling. The Constitution does not protect a right to less equal districts, and more equal districts are not proof of racial gerrymandering. Our opinion on this issue, (Doc. 53 at 4-10), and the decision of the Supreme Court in this case, 135 S.Ct. at 1263, have repeatedly explained that under federal law, Alabama is entitled to use a ±1% population deviation. One person, one vote is merely “part of the redistricting background, taken as a given, when determining whether race, or other factors, predominate in a legislator’s determination as to how equal population objectives will be met.” Id. at 1270.
b. Alabama Had a Statewide Policy of Racial Targets, but the Plaintiffs Still Must Prove that the Policy Caused Race to Predominate in Individual Districts.
The Supreme Court found that Alabama pursued a policy of keeping the black population in each majority-black district at or above the percentage in that district in 2010, but it remanded for this Court to determine whether the legislature subordinated traditional districting criteria to race in individual districts. See id. at 1267. The plaintiffs argue that the targets were applied statewide in such a way that race predominated in every challenged district. The Democratic Conference plaintiffs also argue that Hinaman and Senator Dial explicitly subordinated the goals of maintaining communities of interest, preserving county boundaries and precinct lines, and avoiding changes in the districts.
Without further district-specific inquiry, these arguments fail to prove that race necessarily predominated in the design of any challenged district. As the plaintiffs’ own mapmakers explained on remand, there are “places where you just literally cannot avoid” a certain black population percentage. (Doe. 297-1 at 30). That is, the legislature might have matched certain percentages of black population because they followed traditional districting criteria.
The testimony of McClendon, Dial, and Hinaman establishes that the drafters did not necessarily prioritize racial targets over all other traditional districting criteria in every single district. None of the three gave any indication that “[r]ace was [a] criterion that ... could not be compromised.” Shaw, 517 U.S. at 907, 116 S.Ct. 1894. Representative McClendon testified that a “district is retrogressed if the minorities in that district, whether by race or language, are worse off after redistricting than they were before redistricting.” (Doc. 217 at 221). This statement does not mean that he prioritized maintaining the black population percentage over all other criteria. Senator Dial testified that “to keep from regressing [a] district and increasing [its] population, we had to increase it per-centagewise on the same number of minority votes that we had.” (Doc. 215 at 36). But his goal was “bringing the African American populations of those districts up to approximately equal as best [we] could with what it had been in 2001.” (Id. at 37). Senator Dial did not testify that he subordinated other districting criteria to a racial target.
Hinaman testified that he “look[ed] at [the] 2010 census as applied to 2001 lines, [and] whatever that number was, [he] tried to be as close to that as possible. And if [he] was significantly below that, [he] was concerned about that being retrogression *1051that would be looked upon unfavorably by the Justice Department under Section 5.” (Doc. 217 at 145-46). Hinaman explained that “if you took a district that was somewhere in the 60 to 65 percent black majority district and you brought it down into the low 50s, [he thought] people would be concerned whether that population would then have the opportunity to elect a candidate of their choice.” (Doc. 134-3 at 102). But if “a district that was in the upper 70s [was taken] down to 70 percent black” he “would be less concerned.” (Id.) No one instructed Hinaman that such a reduction would “be a matter of concern.” (Id.) This testimony, together with that of Dial and McClendon, means that statewide evidence alone cannot answer the question of racial predominance.
The Democratic Conference plaintiffs argue that, in every challenged district, Hinaman and Senator Dial prioritized racial considerations over the goals of maintaining communities of interest, preserving county boundaries and precinct lines, and avoiding changes in the districts, (Doc. 258 at 15), but they are. mistaken. Hinaman testified that changing each district as little as possible “was a goal” but it was “down on the list from one person, one vote,” “not retrogressing the minority districts,” and “[separating incumbents.” (Doc. 217 at 162). That some priorities (for example, minimizing change) were less important than other priorities (for example, separating incumbents) is unsurprising. It does not prove that race predominated over traditional districting criteria. Hina-man also testified that he would sometimes split precincts for several reasons: avoiding retrogression, maintaining the ±1% population deviation, and accommodating incumbents. (Doc. 134-3 at 117-18). Again, Hinaman’s testimony does not suggest that race was a “criterion that ... could not be compromised.” Shaw, 517 U.S. at 907, 116 S.Ct. 1894. But even if he split some precincts for the sole purpose of increasing the black population percentage in a majority-black district, Hinaman did not testify that he did so in every precinct, and the plaintiffs must prove their claims on a district-by-district basis. See Ala. Legislative Black Caucus, 135 S.Ct. at 1265.
To be sure, Senator Dial believed that compliance with the Voting Rights Act was a higher priority than maintaining communities of interest and preserving county boundaries. (Doc. 215 at 28). But he did not directly draft any of the districts, and he did not testify that he used an impermissible racial target in any particular district. Moreover, Dial was correct that compliance with federal law must be a higher priority.
In any event, we know that the policy of racial targets was often disregarded because most of the districts did not match the previous black population percentage. Only 13 of the 28 challenged House districts and 3 of the 8 challenged Senate districts maintained a black population percentage within 1 point of the percentage in 2010. In the House, 11 of the 28 districts were at least 4 points away from their prior percentages; in the Senate, 4 of 8 were at least 4 points away from their prior percentages. In the House, the 2012 plans varied by as much as -8.54% to + 9.76% from the target; in the Senate, they varied by as much as -14.58% to + 8.91%. Moreover, if the drafters tried to maintain a policy of not reducing the black population percentage in any of the challenged districts, they failed spectacularly: in a quarter of the districts, the black population percentage decreased between 4 and 15 points. The following tables illustrate the extent to which the state repeatedly, and often badly, missed its alleged targets:
*1052Change in Total Black Population Percentage in Challenged Senate Districts Under Enacted Plans
[[Image here]]
Change in Total Black Population Percentage in Challenged House Districts Under Enacted Plans
[[Image here]]
*1053Nor is there anything necessarily suspicious about a district that maintains the same black population percentage that it had in 2001. In fact, the plaintiffs drew several districts in their court-ordered alternative plans that came closer to the previous percentage than the state did. In the Black Caucus 1% Plan, House District 70 was only 0.07 points off, and House District 72 came within one point. (Doc. 295 at 31 & n.7). In Democratic Conference Plan A, Senate District 18 hit the target exactly, and House District 82, House District 85, Senate District 28 came within one point. (Id).
Even the earlier alternative plans have similar black population percentages in many of the majority-black districts. The following tables show the black population percentages using data from the 2010 Census under the 2001 district lines, under the enacted districts, and under the alternative plans.
Total Black Population Percentages of Challenged Senate Districts Across Plans
[[Image here]]
*1054Total Black Population Percentages of Challenged House Districts Across Plans
[[Image here]]
(* = Statistic for House District 6) (. = Statistic for House District 19)
Such similarities in black population percentages are unsurprising when the plan was meant to maintain the characteristics of the preexisting districts to the extent possible. (Doc. 134-4 at 25-26; Doc. 217 at 162). Even in the majority-white districts, thirty-five districts have a black population percentage within one point of the old lines. (Doc. 263-2). The plaintiffs do not argue that the drafters maintained racial targets for the majority-white districts, and yet we see the same pattern of similar percentages of black population. The plaintiffs have presented no evidence that the 2001 district lines were racial gerrymanders, and the only evidence we have suggests that those lines were based on partisan gerrymanders, as the state and Democratic leaders previously argued. Where Alabama chose to follow existing district lines, we cannot infer that their decision to avoid change necessarily created a racial gerrymander.
*1055Some districts also will have high percentages of black population no matter how the district is drawn because the black population in Alabama is not evenly dispersed throughout the state. It is concentrated in counties along the Black Belt in the south-central part of the state, as well as the counties that contain major metropolitan areas: Madison County in the north (Huntsville), Jefferson County in the north-central (Birmingham), Montgomery County in the south-central (Montgomery); and Mobile County in the southwest (Mobile). The following map shows how the counties with high percentages of black population are concentrated:
Black Population Percentages in Alabama Counties in 2010 Census
[[Image here]]
(U.S. Census Bureau, American FactFin-der, http://factfinder.eensus.gov/faces/nav/ jsfipages/community_facts.xhtml).
Within those counties, many cities and neighborhoods are racially concentrated. For instance, Birmingham, with a population of 212,237, is 73.4 percent black, (id.), and small towns to the west of Birmingham have high concentrations of black population. Bessemer, southwest of Birmingham, is 71.2 percent black. (Id.) Fair-field, west of Birmingham, is 94.6 percent black. (Id.) Midfield, west of Birmingham, is 81 percent black. (Id.) Brighton, southwest of- Birmingham, is 81 percent black. (Id.) In contrast, small towns and neighborhoods to the south and southeast of Birmingham are predominantly white. Mountain Brook, southeast of Birmingham, is 97.2 percent white. (Id.) Home-*1056wood, south of Birmingham, is 74.6 percent white. (Id.) Vestavia Hills, south of Birmingham, is 90.4 percent white. (Id.) And Hoover, south of Birmingham, is 75.1 percent white. (Id.). For these reasons, that a black population percentage is similar to the old map does not prove by itself that race predominated in every challenged district. The plaintiffs must prove district-by-district that race predominated, including with evidence about targets where appropriate.
We also reject the related argument advanced by the plaintiffs that Alabama should bear the burden of proving that race did not predominate in each district because of the plaintiffs’ evidence about a statewide policy of racial targets. This argument is another attempt by the plaintiffs to make out a statewide claim of racial gerrymandering. As the Supreme Court explained, a “racial gerrymandering claim ... applies to the boundaries of individual districts. It applies district-by-district. It does not apply to a State considered as an undifferentiated ‘whole.’ ” Ala. Legislative Black Caucus, 135 S.Ct. at 1265 (citation omitted). A “plaintiff must show that ‘race was the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.’” Id. (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475). The plaintiffs “can present statewide evidence in order to prove racial gerrymandering in a particular district,” id. but this evidence does not transform the inquiry. The plaintiffs still must prove, in each district, that race predominated over traditional districting factors.
c. The Dissent Misstates the Test for Racial Predominance.
Our dissenting colleague misapplies Supreme Court precedent and, like the plaintiffs, uses a rebuttable presumption of racial predominance. The key to the dissent’s test is whether the legislature hit the alleged racial target in a particular district. (Dissent at 1370-71, 1377-78, 1378). If the alleged target was met, our colleague infers that race predominated. (Id. at 1375, 1376-77). This inference makes the dissent’s task easy. (Id. at 1390). Once the inference is applied, our colleague asks whether any district-specific drafting choices corroborate the inference that race predominated and, if so, presumes that it did. (Id. at 1371, 1378). Only then — after inferring that race predominates and confirming that inference with circumstantial evidence — does our colleague engage in the inquiry required by the Supreme Court: whether race-neutral factors predominated over racial considerations in the “overall design of the district.” (Id. at 1378). But, tellingly, our colleague comes to this required inquiry with his mind made up, because “evidence that the drafters hit, within one percentage point, their admitted target of maintaining existing racial percentages paints a clear picture that the racial percentage in each district was specifically intended.” (Id. at 1378 (emphasis added)). This test manifests two errors — it effectively flips the burden of proof, and it excludes evidence that we must consider.
By starting with this inference of racial predominance, the dissent gets the test exactly backwards. The plaintiffs have the burden of proving racial predominance, and they “must prove that the legislature subordinated traditional race-neutral dis-tricting principles ... to racial considerations.” Ala. Legislative Black Caucus, 135 S.Ct. at 1265 (alteration in original) (emphasis omitted) (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475). To be sure, the Supreme Court explained “[t]hat Alabama expressly adopted and applied a policy of prioritizing mechanical racial targets above all other districting criteria (save one-per*1057son, one-vote),” but that Alabama did so only “provides evidence that race motivated the drawing of particular lines in multiple districts in the State.” Id. at 1267 (emphasis added). That is, we weigh Alabama’s policy in our analysis, but our consideration of this significant evidence does not end our inquiry as it effectively does the inquiry of our colleague. {See Dissent at 1383 (stating that the achievement of an alleged racial quota “creates a strong inference of intent”)). Instead, we follow the approach of the Supreme Court and ask whether the plaintiffs have introduced additional evidence “that this goal had a direct and significant impact on the drawing of at least some of’ the boundaries of a district, or whether “ ‘traditional’ factors” offer a better explanation. Ala. Legislative Black Caucus, 135 S.Ct. at 1271.
In contrast to the balanced approach of the Supreme Court, the test of our dissenting colleague leads to the rebuttable presumption of unconstitutionality of every majority-minority district in which the drafters came close to their alleged target. (Dissent at 1355, 1378, 1404). Take our dissenting colleague’s analysis of proposed House District 67. He concedes that race-neutral principles explain why the drafters incorporated portions of Perry County into the district. (Id. at 1396). But our colleague cannot join our “ultimate” conclusion that race did not predominate because the black population remained close to the 2001 levels and some of the boundary lines could have been “smoother.” (Id. at 1394, 1396). In other words, race predominated because the alleged target was met. This ostensible but-for test of racial predominance is not the test of the majority in Alabama Black Legislative Caucus. It is closer to the view of the dissenting justices, first articulated in an earlier decision. See League of United Latin Am. Citizens v. Perry, 548 U.S. 399, 517, 126 S.Ct. 2594, 165 L.Ed.2d 609 (2006) (Scalia, J., concurring in the judgment in part and dissenting in part, joined by Roberts, C.J., Thomas, J., and Alito., J.) (“In my view, however, when a legislature intentionally creates a majority-minority district, race is necessarily its predominant motivation and strict scrutiny is therefore triggered.”). True, our dissenting colleague’s test would make this case easy, but it might also call into question “the constitutionality of the Voting Rights Act” itself. See Bethune-Hill v. Va. State Bd. of Elections, 141 F.Supp.3d 505, 527 (E.D. Va. 2015) (“If the use of a [racial target to avoid retrogression] is sufficient to trigger strict scrutiny in the absence of a facial manifestation in the lines themselves through the subordination of traditional redistricting principles, then the constitutionality of the Voting Rights Act — as applied to redistricting — would be drawn into question.”).
In addition, the district court decisions that the dissent relies on do not support its view of racial predominance. (Dissent at 1372-73). In Bethune-Hill, the majority opinion rejected a test for racial predominance that is similar to the dissent’s inflexible test. 141 F.Supp.3d at 528 (rejecting the test of the dissent that “views the 55% [black voting age population] floor as a ‘filter through which all line-drawing decisions had to pass’ ” (citation omitted)). In Covington v. North Carolina, 316 F.R.D. 117 (M.D.N.C. 2016), the district court emphasized that a “racial gerrymandering claim ‘applie[d] district-by-district,’ and not to the state ‘as an undifferentiated whole.’ ” Id. at 140 (quoting Alabama Legislative Black Caucus, 135 S.Ct. at 1265). The analysis of the district court therefore weighed “district-specific evidence” together with statewide evidence of racial predominance. Id. And in Harris v. McCrory, 159 F.Supp.3d 600 (M.D.N.C. 2016), the' district court found that statewide evi*1058dence of a racial quota proved that race predominated because the drafter of the redistricting plan at issue expressly testified that he subordinated race to traditional districting criteria. Id. at 612. Such direct evidence is absent from the record before us.
By inferring that the drafters acted in bad faith if they hit their alleged target, contra Miller, 515 U.S. at 915, 115 S.Ct. 2475 (“[T]he good faith of a state legislature must be presumed.”), our dissenting colleague ignores evidence that the legislature did not subordinate race-neutral dis-tricting criteria to racial considerations. For example, our dissenting colleague would have us disregard the alternative plans produced by the plaintiffs. (Dissent at 1373-75). Our consideration of the alternative plans, according to our dissenting colleague, is an attempt to hunt down after-the-fact justifications for the drafting decisions of the defendants, (id. at 1373), but we disagree. We consider the alternative plans because the plaintiffs urged us to consider them over the strong objection of the defendants, (Doc. 284 at 9-10,12-13, 35-38), and the plaintiffs bear the burden of proving racial predominance. Ala. Legislative Black Caucus, 135 S.Ct. at 1265.
Our colleague would also have us cast every race-neutral drafting choice in a district that hit its alleged target as evidence that race predominated. (Dissent at 1375). Once again, this assertion is the result of presuming the unconstitutionality of the majority-minority districts in which the drafters hit their alleged target. The Supreme Court, by contrast, commands us to weigh evidence that the drafters met a racial target against other factors. Ala. Legislative Black Caucus, 135 S.Ct. at 1271.
Finally, the dissent argues that we take a “categorical[ ]” approach to racial predominance because we refuse to rule that race predominated when the evidence establishes that race-neutral factors offer the best explanation for the composition of a district. (Dissent at 1369, 1369-70). Our colleague asserts that we “insist[] that race predominates only when a district contains direct evidence of race-based de-cisionmaking or explanations based on traditional districting criteria are impossible.” (Id. at 1370). Not true. We, unlike our dissenting colleague, hold the plaintiffs to their burden of proof. Although the dissent’s inquiry effectively begins and ends with evidence that the drafters met their racial target, we, weigh all the evidence the parties presented. And if race-neutral districting factors offer a better explanation for the drafting of a district, we find that race did not predominate. Our approach comports with the longstanding recognition by the Supreme Court of “the sensitive nature of redistricting and the presumption of good faith that must be accorded legislative enactments.” Miller, 515 U.S. at 916, 115 S.Ct. 2475. Specifically, we exercise the “extraordinary caution in adjudicating claims that a State has drawn district lines on the basis of race” commanded by the Supreme Court. Id.
We follow the instruction of the Supreme Court to consider the evidence presented by the parties on remand. As part of the totality of that evidence, we consider the black population percentage in the challenged districts. And although our dissenting colleague would have us consider this evidence to the exclusion or diminution of other evidence produced by the parties, we cannot do so. Were it as simple as seeing whether Alabama hit its alleged target, there would have been no need for remand in this case.
d. The Plaintiffs Must Prove that the Precinct Splits Cannot Be Explained by Traditional Districting Criteria.
We agree with Alabama that the plaintiffs must prove that individual precincts, *1059which are a community of interest in Alabama, were split on the basis of race and not some other traditional districting criterion. We also agree, with respect to most of the splits, that the plaintiffs have failed to do so. Because these arguments are made about many districts, we address them now.
That the legislature split a precinct does not necessarily prove that race predominated. The plaintiffs present no direct evidence that any precinct was split because of racial considerations. In fact, the Democratic Conference plaintiffs admitted on remand that the presence of split precincts in the majority-black districts is not probative, (Doc. 258 at 28-29), because many precincts were split between majority-white districts as well. When they tried to draw an alternative plan, the Black Caucus plaintiffs conceded that not all precinct splits are suspicious. (-See, e.g., Doc. 801 at 11). And Hinaman testified that communities of interest may cross the lines of political subdivisions. (Doc. 217 at 210).
Further, we cannot infer that race predominated in a district simply because the legislature put a higher percentage of black population into one district or the other. (-See Doc. 256 at 31; Doc. 258 at 26). Many of the differences are negligible. (-See, e.g., Def. Ex. 405 (Second Mount Zion Ch precinct, placing population that was 72.9 percent black into majority-black House District 68 and population that was 71.4 percent black into majority-white House District 90)). Several of the precinct splits placed a higher black population percentage in a majority-white district than a majority-black district. (-See, e.g., Def. Ex. 405 (Westlawn Mid. Sch precinct, placing population that was 18.6 percent black in majority-black House District 53 and population that was 24.5 percent black in majority-white House District 6)). And several of the precincts give so little population to a district that the split can hardly be called evidence of anything. Moreover, in many of the splits where the majority of the population is black on one side and white on the other side, the drafters could have chosen to distribute census blocks on more racially polarized lines but did not do so. (See, e.g., APSX 55 (placing majority-white census blocks along the border into House District 72)).
Racial disparities in the precinct splits may also be the result of geography. For example, if a majority-white district abuts a majority-black district, a precinct that connects the two may naturally have more white population on one side and more black population on the other. A sensible line between the two districts would unavoidably have disparate racial percentages. Only a racial gerrymander could do otherwise. As Fairfax testified, racial disparities in precinct splits can be “just natural, unfortunately, population patterns, demo-graphical population patterns.” (Doc. 296-7 143-44).
For these reasons, the plaintiffs have failed to prove that race predominated because the legislature split precincts. We also observe that Alabama submitted an exhibit in which it redrew each district with only whole precincts and, where possible, a population within the ± 1% deviation. Of the 33 hypothetical districts that could comply with the ±1% deviation, 18 had a black population percentage within two points of the corresponding enacted district. In fact, some of the hypothetical districts came closer to the percentage for the 2010 Census population under the 2001 lines than the enacted district did. The following tables compare the figures for the hypothetical and enacted districts:
Black Population Percentages in Hypothetical Senate Districts with No Split Precincts
*1060[[Image here]]
Black Population Percentages in Hypothetical House Districts with No Split Precincts
[[Image here]]
*1061(For districts marked with an asterisk, there was no combination of precincts that complied with the ± 1% deviation.)
(Def. Ex. 3, Doc. 263-3).
Although the plaintiffs do not object to the method used in this exercise, we make no findings about individual districts based solely on this exercise because the state redrew each district individually and some precincts are included in more than one district. (See Doc. 263-3 (House Districts 70 and 71)). We also observe that four districts do not comply with the population deviation, and unsplitting some of precincts could cause incumbent conflicts.
E. Strict Scrutiny Generally
The parties do not dispute that, for each district where racial considerations predominated over traditional districting criteria, Alabama must establish that the district was narrowly tailored to achieve a compelling governmental interest. See Vera, 517 U.S. at 976, 116 S.Ct. 1941. Here, the only asserted interest is complying with the Voting Rights Act. Alabama must have a “strong basis in evidence” that its race-based choice was necessary to achieve that compelling interest. Ala. Legislative Black Caucus, 135 S.Ct. at 1274.
As an initial matter, Alabama had a compelling governmental interest in complying with sections 2 and 5 of the Voting Rights Act. We previously ruled that it did, (Doc. 203 at 160), and the Supreme Court has not held otherwise. No party disputes that, at the time of redistricting, section 5 required Alabama at least to maintain the overall number of majority-black districts. Nor does any party dispute that section 5 required Alabama to avoid a “discriminatory purpose” in redistricting. 52 U.S.C. § 10304. The plaintiffs initially argued that the requirements of section 5 are irrelevant to the constitutionality of the 2012 redistricting because the Supreme Court held the preclearance formula in section 4 unconstitutional in 2013, see Shelby County, — U.S. -, 133 S.Ct. 2612, 186 L.Ed.2d 651, but the plaintiffs have since abandoned this argument and the Supreme Court rejected it as applied to a different state in Harris v. Ariz. Indep. Redistricting Comm’n, No. 14-232, slip op. at 10-11, — U.S. -, 136 S.Ct. 1301, 194 L.Ed.2d 497 (U.S. Apr. 20, 2016). Although the Democratic Conference plaintiffs argue that the Legislature did not understand the requirements of section 5, that argument relates to narrow tailoring.
Under section 5, the legislature must “preserve existing minority percentages” to the extent necessary “to maintain the minority’s present ability to elect the candidate of its choice.” Ala. Legislative Black Caucus, 135 S.Ct. at 1274. The legislature need not “guess precisely what percentage reduction a court or the Justice Department might eventually find to be retrogressive,” id. at 1273, but it must have a “strong basis in evidence” for its use of race, id. at 1274. “[L]egislators ‘may have a strong basis in evidence to use racial classifications in order to comply with a statute when they have good reasons to believe such use is required, even if a court does not find that the actions were necessary for statutory compliance.’ ” Id. (quoting Brief of United States as Ami-cus Curiae 29).
Section 2 of the Voting Rights Act overlaps in some ways with section 5. Under section 2, Alabama must avoid diluting the voting strength of a racial minority where that “racial group is sufficiently large and geographically compact to constitute a majority in a single-member district,” “the group is politically cohesive,” and the white population “votes sufficiently as a bloc to enable it ... usually to defeat the *1062minority’s preferred candidate,” Bossier Parish, 520 U.S. at 479, 117 S.Ct. 1491 (quoting Gingles, 478 U.S. at 50-51, 106 S.Ct. 2752). If these factors, known as the Gingles factors, are present in a district, a court looks to whether “the totality of the circumstances supports a finding that the voting scheme is dilutive.” Id. at 480, 117 S.Ct. 1491.
Alabama makes three arguments about why the districts are narrowly tailored to comply with section 5, but none of them prove that every challenged district survives strict scrutiny. First, Alabama argues that the districts with a total black population percentage of 62 to 65 percent are narrowly tailored because that was the percentage that two black legislators told the Committee was necessary to maintain black voters’ ability to elect their preferred candidates. At the public hearing in Thom-asville, Representative Thomas Jackson — a member of the Black Caucus — explained that majority-black districts in his area should be “sixty-two percent or sixty-five percent.” (Doc. 30-23 at 8). At the public hearing in Selma, Senator Sanders — also a member of the Black Caucus — told Senator Dial that none of the majority-black districts should be less than 62 percent black. (Doc. 30-28 at 6). Senator Sanders explained why he thought this minimum was necessary:
Sometimes a lot of people don’t vote. Sometimes a lot of people can’t vote. They might be in prisons or other kinds of institutions. Sometimes a lot of folks are discouraged for one reason or another. So I would hope that 62 percent is a minimal [percentage] for the majority African-American districts].
(Id.). Senator Dial testified that, if he had told the black leadership in the Senate that it could have no more than 55 percent black population, and that this lower number was better for their communities, “Senator Sanders and my other good friends in the Senate ... would simply have glazed over and asked me when I was going to the mental institute.” (Doc. 215 at 44-45).
The Democratic Conference plaintiffs are correct that the use of the 65 percent threshold was not narrowly tailored for every district. The drafters relied on the statements of black legislators in rural districts in the western Black Belt. Because they had no evidence about other districts, especially urban areas or districts outside of the western Black Belt, Alabama cannot use 65 percent across the entire state without any further inquiry.
Second, Alabama argues that caselaw from the Supreme Court and multiple circuits establishes that 65 percent is a reasonable threshold for minority voting percentages. In United Jewish Organizations of Williamsburgh, Inc. v. Carey, 430 U.S. 144, 97 S.Ct. 996, 51 L.Ed.2d 229 (1977), the Supreme Court explained that “it was reasonable for the Attorney General to conclude in this case that a substantial nonwhite population majority in the vicinity of 65% would be required to achieve a nonwhite majority of eligible voters.” Id. at 164, 97 S.Ct. 996. The Seventh Circuit has explained that “a guideline of 65% of total population (or its equivalent) has achieved general acceptance in redistricting jurisprudence.” Ketchum v. Byrne, 740 F.2d 1398, 1415 (7th Cir. 1984); see also Latino Political Action Comm., Inc. v. City of Boston, 784 F.2d 409, 414 (1st Cir. 1986) (“Where voting is highly polarized, a 65 percent figure is a generally accepted threshold which has been used by the Department of Justice and reapportionment experts.”). The Eighth Circuit also has concluded that “either 60% of the voting age population or 65% of the total population . is reasonably sufficient to provide black voters with an effective majority.” *1063African Am. Voting Rights Legal Def. Fund, Inc. v. Villa, 54 F.3d 1345, 1348 n.4 (8th Cir. 1995).
But as the Democratic Conference plaintiffs argue, this rule of thumb must yield to specific evidence. Alabama asserted in its 2001 submissions to the Department of Justice for preclearance that 55 percent black voting-age population was sufficient to avoid retrogression, and Alabama submitted a study by Professor Richard L. Engstrom that concluded “a black voting age population of about 55% provides African-Americans with a reasonable opportunity to elect the representative of their choice.” (ADC Supp. Ex. 1 at 7; accord ADC Supp. Ex. 2 at 9). None of the cases that Alabama cites in its brief were decided about Alabama after this submission, so none can provide a strong basis in evidence for the legislature to use a higher percentage in 2012.
Third, Alabama argues that the enacted plans are the only plans in the record that satisfy section 2, section 5, and the Committee guidelines with regard to every majority-black district. We put no weight on the argument of Alabama that its plans satisfy strict scrutiny because the plaintiffs have not offered any alternative plans that comply with the Committee guidelines. Alabama has the burden to prove that its plans are narrowly tailored. See Miller, 515 U.S. at 920, 115 S.Ct. 2475. To be sure, the absence of a better alternative is evidence that Alabama narrowly tailored its plan. But the lack of a workable alternative is not dispositive. In some instances, we do not need to see an alternative plan to conclude that a district fails strict scrutiny.
Our dissenting colleague distorts the ev-identiary burden of Alabama to prove narrow tailoring. Although our colleague uses the terms “narrow tailoring” and “strong basis in evidence,” he looks to affirmative action caselaw — not caselaw about claims of racial gerrymandering — to define those terms. (Dissent at 1357, 1357-58, 1362, 1363, 1364, 1368). Because this error transforms the burden of Alabama from one of proving “good reasons” into a burden of proving actual necessity, Alabama Legislative Black Caucus, 135 S.Ct. at 1274 (citation omitted), we respectfully disagree.
In the context of a racial gerrymandering claim, the Supreme Court has long held that the requirement of narrow tailoring gives states “leeway,” although to a “limited degree,” to comply with the Voting Rights Act. Bush v. Vera, 517 U.S. 952, 977, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996). This flexible approach to narrow tailoring is unique to the racial gerrymandering context because the Supreme Court “adhere[s] to [a] longstanding recognition of the importance in our federal system of each State’s sovereign interest in implementing its redistricting plan.” Id. at 978, 116 S.Ct. 1941. Against this interest weighs compliance with the Voting Rights Act. To balance these interests, and to avoid placing “state actors ... [in a] trap[ ] between ... competing hazards of liability’ by the imposition of unattainable requirements under the rubric of strict scrutiny,” the Court instructs that we take a “flexible]” approach to narrow tailoring. Id. at 977-78, 116 S.Ct. 1941 (quoting Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 291, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (O’Connor, J. concurring in part and concurring in the judgment)).
The Supreme Court emphasized on remand that we were to adhere to this longstanding standard of narrow tailoring:
[W]e do not insist that a legislature guess precisely what percentage reduction a court or the Justice Department might eventually find to be retrogressive. The law cannot insist that a state *1064legislature, when redistricting, determine precisely what percent minority population § 5 demands. The standards of § 5 are complex; they often require evaluation of controverted claims about voting behavior; the evidence may be unclear; and, with respect to any particular district, judges may disagree about the proper outcome. The law cannot lay a trap for an unwary legislature, condemning its redistricting plan as either (1) unconstitutional racial gerrymandering should the legislature place a few too many minority voters in a district or (2) retrogressive under § 5 should the legislature place a few too few.
Alabama Legislative Black Caucus, 135 S.Ct. at 1273-74 (citations omitted). The distinctions that the Court drew in this passage make clear that narrow tailoring does not require an exact connection between the means and ends of redistricting. Instead, as we have explained, narrow tailoring requires that the legislature had a “strong basis in evidence,” which in turn means that the legislators had “good reasons” to draft a district in which race predominated over traditional districting criteria. Id. at 1274.
Our dissenting colleague argues that this standard is the same standard that the Supreme Court “applie[s] ... [in the] racial-preference eontext[ ].” (Dissent at 1357). To support this argument, our colleague latches- onto a citation in Alabama Legislative Black Caucus to an amicus brief of the United States that in turn cited Ricci v. DeStefano, 557 U.S. 557, 129 S.Ct. 2658, 174 L.Ed.2d 490 (2009), and several citations in other racial-gerrymandering decisions. (Dissent at 1357, 1358). But this argument misconstrues the decisions of the Supreme Court.
The Supreme Court prescribes different approaches to narrow tailoring depending upon the context. In the context of affirmative action, the Court asks whether the “race-based action is necessary to further a compelling governmental interest.” Grutter v. Bollinger, 539 U.S. 306, 327, 123 S.Ct. 2325, 156 L.Ed.2d 304 (2003) (emphasis added); see also Ricci, 557 U.S. at 582, 129 S.Ct. 2658 (explaining that race-based remedies for past discrimination “are constitutional only where there is a ‘strong basis in evidence’ that the remedial actions were necessary” (emphasis added) (quoting Richmond v. J.A. Croson Co., 488 U.S. 469, 500, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989))). In the context of voting rights, by contrast, narrow tailoring “does not demand that a State’s actions actually be necessary to achieve a compelling state interest in order to be constitutionally valid.” Ala. Legislative Black Caucus, 135 S.Ct. at 1274 (emphasis added). In other words, the standard of proof is different in the voting rights context. And this difference makes sense because a state must navigate the Scylla and Charybdis of compliance with the Voting Rights Act, on one hand, and the demands of the Equal Protection Clause, on the other. See id. at 1273-74 (“The law cannot lay a trap for an unwary legislature.... ”).
Finally, contrary to the dissent, (Dissent at 1364-65, 1366-67), the Supreme Court has never required particular studies. Cf. Shaw v. Hunt, 517 U.S. 899, 915, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996) (“[W]e have not always provided precise guidance on how closely the means (the racial classification) must serve the end (the justification or compelling interest).”). All the Court “insist[s]” upon is “a ‘strong basis in evidence.’” Ala. Legislative Black Caucus, 135 S.Ct. at 1274 (quoting Brief for United States as Amicus Curiae 29). We evaluate the proposed districts in which race predominated based on this standard.
F. Districh-By-District Analysis of the Challenged Plan
With these factual findings and conclusions of law in mind, we turn to the dis*1065trict-specific evidence and arguments. We begin with the Senate districts, discussing them in numerical order. We then discuss the House districts, beginning with the Madison County districts and continuing in numerical order.
a. Senate Districts 18, 19, and 20 (Birmingham)
We find that race did not predominate over traditional districting criteria in the design of Senate Districts 18 and 19, but we find that race predominated in the design of District 20. All three districts were severely underpopulated in 2010 — by 17.64, 20.06, and 21.37 percentage points, respectively. (Doc. 1 at 10). Senator Smith-erman, a black Democrat who represents District 18, provided the drafters with a one-page map that Hinaman “endeavored to duplicate.” (Doc. 203 at 31-32; Doc. 217 at 121). The drafters maintained the core of each district, echoing their stated policy of minimizing change, (Doc. 134-4 at 25-26), and each district has grown to expand in population:
[[Image here]]
(Ala. Reapportionment Office, Alabama Legislative Policymaker’s Dashboard, http://policymaker.alabama.gov/Districts. aspx.)
The total black population percentages decreased in each of these districts, undermining the plaintiffs’ claim that race predominated over traditional districting factors because of a policy of racial targets. In Districts 19 and 20, the black population percentage dropped significantly. The following table shows the change for each district:
Total Black Population Percentage in 2010 Census
[[Image here]]
If the drafters intended to meet targets, they missed them — and badly in Districts 19 and 20.
The Democratic Conference plaintiffs drew a different configuration of Districts 18, 19, and 20, but they kept all three districts majority-black and entirely within Jefferson County. The following map superimposes Democratic Conference Plan A (darker dotted lines) on the enacted plan *1066(red lines and shaded colors) and the county boundaries (lighter dotted lines):
Senate Districts 18-20 in Democratic Conference Plan A and Act 603
[[Image here]]
(Doc. 287-20 at 1). Plan A District 18 trades the northern and southern additions of the enacted district for a western addition, and Plan A District 19 pushes west to rural Jefferson County and the county line. Plan A District 20 reaches into territory that Act 603 gave to District 18, as well as farther north. The Democratic Conference plaintiffs did not draw significantly more compact or regular districts, and they did not explain why their choices were required by traditional districting criteria. They provided us with quantitative measurements of the compactness of their alternative districts, but they did not provide measurements for each enacted district, so we cannot make any findings based on the measurements they have provided. That said, the Democratic Conference plaintiffs split fewer precincts in these districts. Plan A provides little evidence overall that race predominated in the design of Districts 18,19, or 20.
The Black Caucus plaintiffs drew a different configuration of the three districts. They shifted District 18 westward in an odd hook that follows precinct lines, moved District 19 slightly northward, and kept District 20 largely in the same place but with different irregular lines that follow precinct boundaries. The following maps superimpose the Black Caucus 1% Plan (shaded colors and numbers in black boxes) on the enacted plan (purple lines and numbers in purple circles), with 2012 incumbent locations marked by a blue star:
Senate District 18 in Black Caucus 1% Plan and Act 603
*1067[[Image here]]
(APSX 534).
Senate District 19 in Black Caucus 1% Plan and Act 603
[[Image here]]
*1068(APSX 535).
Senate District 20 in Black Caucus 1% Plan and Act 603
[[Image here]]
(APSX 536). As with Democratic Conference Plan A, the shape and compactness of these districts do not establish that race predominated in the design of Act 603. The plaintiffs fail to explain why their choices were required by traditional districting criteria other than precinct splits, which we discuss later. We cannot say that this plan tends to establish that race predominated.
The plaintiffs argue that, although there “were not enough black residents to maintain the existing super-majorities,” the drafters came as “close[ ]” as they could to hitting their “targets.” (Doc. 258 at 43; see also Doc. 256 at 145). This argument fails because the plaintiffs present no evidence to support their assertion. The shapes of the districts are not so bizarre as to give rise to an inference of gerrymandering, nor are they noticeably more bizarre than the alternative plans or the 2001 plan. And the total black population percentages in the final alternative plans are not so different from the enacted plan that we can infer that race predominated. In the Black Caucus 1% Plan, the black population was 3.14 percentage points lower in District 18, a mere 0.37 points lower in District 19, and 0.17 points higher in District 20. (Doc. 296-1 at 4). In Democratic Conference Plan A, District 18 was 0.4 points higher, District 19 was 3.2 points lower, and District 20 was 0.9 points lower. (Id.). In fact, the black population percentages in these districts are similar across every alternative plan, even the ones that ignore the Committee guidelines:
2010 Census Total Black Population Percentages Under Various Plans
*1069[[Image here]]
On this record, we find that the black population percentages in the enacted plan were the result of demographics and race-neutral choices, not the unsuccessful pursuit of numerical targets.
The plaintiffs also argued initially on remand that the “hook” in the northwestern part of District 20 is proof that race predominated, but their own evidence suggests otherwise. Both plaintiffs’ maps establish that the “hook” takes in majority-white population as well as majority-black population. (ADC Supp. Ex. 36H; APSX 319). And both Plan A and the 1% Plan put substantial portions of the hook in District 20. (APSX 533B; Doc. 287-20 at 1).
The absence of county splits in the enacted districts has prompted odd and contradictory arguments that do not persuade us that race predominated. The Black Caucus plaintiffs argue on remand that the drafters were forced to split county boundaries in the adjoining majority-white districts so that the majority-black districts could meet their “targets.” (Doc. 256 at 146). But Senator Dial testified that it was also the avoidance of incumbent conflicts that required the splitting of counties in the majority-white districts. (Doc. 125-3 at 14). And the plaintiffs’ argument is at odds with Supreme Court precedent. The Supreme Court has made clear that a plaintiff has standing to challenge only his own district as a racial gerrymander. See United States v. Hays, 515 U.S. 737, 745, 115 S.Ct. 2431, 132 L.Ed.2d 635 (1995) (“Even assuming (without deciding) that [the Act] causes injury sufficient to invoke strict scrutiny[,] ... appellees have pointed to no evidence tending to show that they have suffered that injury.”); see also Sinkfield v. Kelley, 531 U.S. 28, 121 S.Ct. 446, 148 L.Ed.2d 329 (2000). Whether the challenged districts are viable does not depend on the borders of other districts.
The Democratic Conference plaintiffs suggest that the drafters should have crossed county lines with the majority-black districts in Birmingham. They contended initially on remand that “there is no explanation for the decision that the black-majority districts alone had to be confined to [Jefferson] County and could not be extended into adjacent counties.” (Doc. 258 at 42 n.20). They evidently changed their minds when they drew Plan A because they kept the majority-black Senate districts in Jefferson County entirely within county lines. The plaintiffs cannot have it both ways: splitting county lines and not splitting county lines cannot both be evidence of racial predominance, at least without more explanation than they provide.
Finally, the plaintiffs point to the number and characteristics of the precinct splits. Both plaintiffs split fewer precincts than the state did, (Doc. 300-1 at 84-90). In Districts 18 and 19, these splits do not provide any evidence that race predominated. In District 20, they do. We discuss each district in turn.
In District 18, the drafters split six precincts, five of which are not suspicious. In Birmingham Botanical Gardens precinct, the drafters split the precinct along a *1070smooth line and an area of zero-population blocks.
Birmingham Botanical Gardens Precinct in Act 603
[[Image here]]
*1071(APSX 317). In Mountain Brook City Hall precinct, the drafters split a precinct of exclusively majority-white blocks between Districts 18 and 16.
Mountain Brook City Hall Precinct in Act 603
[[Image here]]
(APSX 324). The split put 927 people in District 18, 6 percent of whom were black, and 3,975 people in District 16, 0.2 percent of whom were black. (Def. Ex. 475). In three other precincts — Muscoda Community Center precinct, (APSX 326), Pleasant Hill United Methodist Church precinct, (APSX 328), and Robinson Elementary School precinct, (APSX 329) — the drafters split the precinct with another majority-black block, and the plaintiffs fail to explain the significance of these splits.
The sixth split, of Homewood Public Library precinct, is insufficient to prove that race predominated. The legislature put 963 people in District 18, 41 percent of whom were black, and 6,327 people in District 18, 3 percent of whom were black. (Def. Ex. 475).
Homewood Public Library Precinct in Act 603
*1072[[Image here]]
(APSX 321). But the number of black people placed in District 18 from this split amounts to less than a third of a percent of the population of the district. If we remove this precinct from the district entirely, the black • population percentage increases from 59.12 percent to 59.23 percent, still shy of the alleged target of 59.93 percent. We find that race did not predominate in the design of District 18.
In District 19, the drafters again split six precincts, five of which are not suspicious. In Valley Creek Baptist Church precinct, the drafters drew an irregular line that included many majority-white blocks that were unnecessary to reach majority-black blocks.
Valley Creek Baptist Church Precinct in Act 603
*1073[[Image here]]
(APSX 331). The split put 461 people in District 19, 28 percent of whom were black, and 2,717 people in majority-white District 5, 11 percent of whom were black. (Def. Ex. 475). In Johns Community Center Precinct, the drafters put several populous majority-white blocks in District 12 that were unnecessary to reach majority-black blocks.
Johns Community Center Precinct in Act 603
*1074[[Image here]]
(APSX 322). The split put 805 people in District 19, 16 percent of whom were black, and 683 people in District 5, 4 percent of whom were black. (Def. Ex. 475). In Hillview Fire Station # 1 precinct, the drafters put 366 people in District 19, 48 percent of whom were black, and 2,433 people combined in Districts 17 and 20, 66 percent of whom were black. (Def. Ex. 475).
Hillview Fire Station # 1 Precinct in Act 603
*1075[[Image here]]
(APSX 320). And in Pleasant Hill United Methodist Church precinct, (APSX 328), and Muscoda Community Center precinct, (APSX 326), the drafters split the precinct between two majority-black districts with no racial pattern.
The final split, of Maurice L West Community Center precinct, does not prove that race predominated in the placement of a significant number of people. The drafters used a suspicious line to put 1,579 people in District 19, 31 percent of whom were black, and 581 people in District 17, 5 percent of whom were black. (Def. Ex. 475).
Maurice L West Community Center Precinct
*1076[[Image here]]
(APSX 323). But the black people placed in District 19 from this split amount to less than half of a percent of the population of the district. If we include the entire precinct, which would keep the precinct within the ±1 percent population deviation, the black population percentage decreases negligibly from 65.39 percent to 65.05 percent, still missing the alleged target of 71.65 percent by over six points. If we remove the entire precinct, which would underpopulate the district, the black population percentage increases negligibly to 65.72 percent, still about six points under the supposed target. We find that race did not predominate in the design of District 19.
In District 20, we find that one precinct split is not suspicious. The drafters split Robinson Elementary School between two majority-black districts, (APSX 329), and the plaintiffs fail to explain how this split proves that race predominated.
But on the basis of the other six precinct splits in District 20, we find that race predominated. In Trussville Baptist Church precinct, the drafters used a suspicious line to put 796 people in District 20, 44 percent of whom were black, and 9,300 people in majority-white District 17, 4 percent of whom were black.
Trussville First Baptist Church Precinct in Act 603
*1077[[Image here]]
(APSX 330). In Mountain View Baptist Church precinct, the drafters drew a line that does not wind around looking for majority-black blocks. But the line did put 7,325 people in District 20, 25 percent of whom were black, and 129 people in District 17, only one of whom was black. (Def. Ex. 475).
Mountain View Baptist Church Precinct in Act 603
*1078[[Image here]]
*1079(APSX 325). In Gardendale Civic Center, the drafters used an irregular line to put 1,500 people in District 20, 33 percent of whom were black, and 12,863 people in District 17, 5 percent of whom were black. (Def. Ex. 475).
Gardendale Civic Center Precinct in Act 603
[[Image here]]
(APSX 319). In Pinson United Methodist Church precinct, the drafters used a somewhat irregular line, but the split assigned 4,260 people to District 20, 65 percent of whom were black, and 2,457 people to Pin-son United Methodist Church, 10 percent of whom were black. (Def. Ex. 475).
Pinson United Methodist Church Precinct in Act 603
*1080[[Image here]]
(APSX 327). In Fultondale First Baptist Church precinct, the drafters used an irregular line to split a precinct of almost all majority-white blocks.
Fultondale First Baptist Church Precinct in Act 603
*1081[[Image here]]
(APSX 318). But the split put 2,268 people in District 20, 39 percent of whom were black, and 1,637 people in District 17, only 6 percent of whom were black. (Def. Ex. 475). And in Hillview Fire Station # 1 precinct, the drafters used an irregular line to put 1,957 people in District 20, 81 percent of whom were black, and 842 people in Districts 17 and 19 combined, only 23 percent of whom were black. (Id.).
Hillview Fire Station # 1 Precinct in Act 603
*1082[[Image here]]
(APSX 320). On the basis of these precinct splits, we find that race predominated in the design of District 20.
We further conclude that District 20 does not survive strict scrutiny. Alabama makes no district-specific arguments about why this district was narrowly tailored to achieve its interest in complying with the Voting Rights Act. It makes a statewide argument based on the comments of Senator Sanders and Representative Jackson, but these statements do not provide a strong basis in evidence in District 20. Sanders and Jackson would be familiar with their own districts in the west Black Belt, but Senate District 20 is in Jefferson County. District 20 is also more urban than Sanders’s and Jackson’s districts — 51 percent of the district lives in Birmingham. (Def. Supp. Ex. 6). Because the state has not provided a strong basis in evidence for its use of race, we must enjoin the use of Senate District 20 in future elections.
b. Senate District 23 (West Black Belt)
Although we view it to be a close question, we find that race predominated over traditional districting criteria in the design of Senate District 23. We base this finding primarily on the increased number of counties in the district and the way in which irregular lines include areas with high percentages of black population, especially through the use of precinct splits.
*1083The district lines are not so irregular on their face that we find that race predominated. District 23 was severely underpopulated before redistricting, and it bordered the severely underpopulated District 24 and the residence of the incumbent of District 22. As compared with the 2001 lines, the borders of Senate District 23 are cleaner along the northern and eastern sides, and slightly more irregular to the southwest:
[[Image here]]
(Ala. Reapportionment Office, Alabama Policymaker’s Dashboard, supra).
The Democratic Conference plaintiffs did not draw a significantly more regular or compact district. The following map shades areas only in Plan A in yellow, areas only in Act 603 in red, and areas of overlap in orange:
Senate District 23 in Conference Plan A and Act 603
[[Image here]]
(Doc. 287-24 at 1).
The Black Caucus plaintiffs, in contrast, drew a worse overall shape for District 23. Instead of being relatively compact, the district in the 1% Plan wraps around part of District 22. The following map shows the Black Caucus district in orange and the enacted district with a purple line and the number 23 in a purple circle:
Senate District 23 in the Black Caucus 1% Plan and Act 603
*1084[[Image here]]
(APSX 537).
The plaintiffs have more success with their evidence about counties. The 2001 plan included all or part of nine counties, two of them whole (Dallas and Wilcox) and seven of them split (Autauga, Clarke, Co-necuh, Lowndes, Marengo, Monroe, and Perry). (Doc. 30-44 at 11). Act 603 increased the total number of counties to ten, five of them whole (Butler, Dallas, Lowndes, Perry, and Wilcox) and five of them split (Clarke, Conecuh, Marengo, Monroe, and Washington). (Doc. 30-40 at 8). Both plaintiffs drew a district with no more than eight counties overall and no more than two split counties, raising the possibility that race trumped the Committee guideline about counties in District 23. The Democratic Conference plaintiffs drew a district with eight counties overall and one split by filling out Conecuh and Monroe; removing Clarke, Marengo, and Washington; and adding part of Autauga. (Doc. 300-1 at 93; Doc. 287-19 at 2). Although we do not put much weight on the Black Caucus plaintiffs’ unseemly design, they drew a district with six counties overall and two splits by filling out Conecuh and Monroe, and removing Lowndes, Mar-engo, Perry, and Wilcox. (APSX 537). Even in a district with eight (Plan A), nine (2001), or ten (Act 603) counties, a difference of two counties is substantial.
We now discuss each of the counties. We find no evidence that race predominated in Butler, Dallas, Lowndes, Perry, or Wilcox Counties. Butler County was whole in District 30 in the 2001 plan, (Doc. 30-44 at 12), and is whole in District 23 in Act 603, Plan A, and the 1% Plan. Dallas County is whole in District 23 in all four plans. (Id.). Lowndes County was partially in District 23 in the 2001 plan, (id.), and is entirely in District 23 in Act 603 and Plan A. Perry County was partially in District 23 in the 2001 plan, (id.), and is entirely in District 23 in Act 603 and Plan A. Wilcox County is whole in District 23 in all four plans.
But the evidence about Clarke County suggests that race predominated. The drafters kept the core of the Clarke County portion of the district and pushed to the border of Washington County. They then split Washington County, instead of taking *1085more of Clarke County. They gave no explanation about why this suspicious choice was the result of traditional districting criteria. Moreover, Clarke County has a much lower black population percentage overall (43.88 percent) than the District 23 portion of Clarke County does (70.67 percent). (Def. Ex. 475; Doc. 297-4 at 6). We acknowledge that the District 23 portion of Clarke County was 77.69 percent black in 2001, (Doc. 30-44 at 11), but the legislature drew new lines in 2012 that must be evaluated on their own merit.
The drafters split six precincts in Clarke County, and five of them are evidence that race predominated. First, in Thomasville National Guard Armory precinct, District 23 uses an irregular line to take all of the majority-black blocks along the border and none of the majority-white blocks.
Thomasville National Guard Precinct in Act 603
[[Image here]]
*1086(APSX 300). The split placed no majority-black blocks in District 22. (Id.). Second, in Fulton City Hall precinct, the legislature used irregular lines to put all but one of the majority-black blocks in District 23.
Fulton City Hall Precinct in Act 603
[[Image here]]
(APSX 295). The resulting split assigned 146 people to District 23, 40 percent of whom were black, and 859 people in District 22, only 4 percent of whom were black. (Def. Ex. 475). Third, in Jackson City Hall precinct, District 22 pushes across the border at two points and stops before reaching any majority-black blocks.
Jackson City Hall Precinct in Act 603
*1087[[Image here]]
(APSX 296). Fourth, in Overstreet Grocery precinct, the split is formed by a line of majority-white blocks on the District 22 side and almost exclusively majority-black blocks on the District 23 side.
^ „„„ Overstreet Grocery Precinct in Act 603
[[Image here]]
*1088(APSX 298). The split placed all of the majority-black blocks in District 23. (Id.). Overall, it put 368 people in District 23, 78 percent of whom were black, and 286 people in District 22, only 17 percent of whom were black. (Def. Ex. 475). Fifth, in Skipper Fire Station-Jackson National Guard-Jackson Fire Dept, precinct, the legislature put into District 23 all of the populous majority-black blocks along the border.
Skipper Fire Station-Jackson National Guard-Jackson Fire Dept. Precinct in Act 603
[[Image here]]
(APSX 299A).
The sixth precinct split in Clarke County has no clear racial pattern. In Old Engineers Building-Antioeh Fire Station-Hell-western Fire Dept-Grove City Hall-Old Engineers precinct, the legislature drew an irregular shape that put 539 people in District 23, 55 percent of them black, and 3,067 people in District 22, only 25 percent of them black. (Def. Ex. 475). But the incursions sometimes pick up majority-black blocks and sometimes pick up majority-white blocks with no apparent pattern. (APSX 297).
Old Engineers Building-Antioeh Fire Station-Hellwestern Fire Dept-Grove City Hall-Old Engineers in Act 603
*1089[[Image here]]
(Id.).
After splitting Clarke County to reach the border with Washington County, the drafters put an 82 percent black population in District 23 from Washington County, which was only 25 percent black overall. (Def. Ex. 475; Doc. 297-4 at 6). District 23 did not enter Washington County in the 2001 plan, and the Democratic Conference plaintiffs did not include Washington County in their alternative plan. The defendants offered no specific explanation about why they entered and split Washington County.
The drafters also split five precincts in this county, and three of them are evidence that race predominated. In Carson-Preswick precinct, the drafters put almost exclusively majority-white blocks from two non-eontiguous areas into District 22. The border is composed of majority-black blocks and zero-population blocks on the District 23 side and majority-white blocks and zero-population blocks on the District 22 side.
Carson-Preswick Precinct in Act 603
*1090[[Image here]]
(APSX 372). The drafters assigned 241 people to District 23, 86 percent of them black, and 329 people to District 22, 17 percent of them black. (Def. Ex. 475). In Cortelyou precinct, the drafters drew an irregular split in the northern corner of the precinct that put only majority-white blocks in District 22.
Cortelyou Precinct in Act 603
*1091[[Image here]]
(APSX 373). The drafters placed 412 peo-pie in District 23, 66 percent of them black, and 86 people in District 22, none of them black. (Def. Ex. 475). In McIntosh Voting House Voting District precinct, the drafters carved a bent leg into the southeast portion of the precinct that picked up every majority-black block in the area and only three majority-white blocks,
McIntosh Voting House Voting District Precinct in Act 603
*1092[[Image here]]
(APSX 376). The split placed 523 people in District 23, 73 percent of them black, and 1,435 people in District 22, 4 percent of them black. (Def. Ex. 475).
In the other two split precincts, there was no apparent pattern of racial sorting. In Malcolm Voting House Voting District precinct, the drafters carved off a mostly unpopulated corner of the irregularly shaped precinct.
Malcolm Voting House Voting District Precinct in Act 603
*1093[[Image here]]
(APSX 374). They put 24 people in District 23, 75 percent of them black, and 558 people in District 22, 61 percent of them black. (Def. Ex. 475). Neither the shape nor the statistics are suspicious. In McIntosh Community Center Voting District precinct, instead of following the straight precinct line, the drafters included a narrow zero-population block in District 22.
Zoom of McIntosh Community Center Voting District Precinct in Act 603
*1094[[Image here]]
(APSX 375A). We are at a loss as to why the drafters made this choice, but race could not have been the reason.
In Conecuh County, we find weak evidence that race predominated. The legislature kept the core of the Conecuh County portion of the district but changed the shape slightly. This design put a slightly higher black population percentage in the district (60 percent) than the county had as a whole (46 percent). (Def. Ex. 475; Doc. 297-4 at 6).
The drafters split five precincts in Cone-cuh County, and the splits give us only mixed evidence. First, in Bermuda Community House precinct, the drafters opted for an irregular line instead of a straight line and placed all of the majority-black blocks in District 23.
Bermuda Community House Precinct in Act 603
*1095[[Image here]]
(APSX 302). The split put 170 people in District 23, 54 percent of them black, and 262 people in District 22, only 19 percent of them black. (Def. Ex. 475). Second, in Castleberry Fire Dept.-1 precinct', the drafters put a significant cluster of majority-black blocks and a majority-white block of 49 people in District 23. (APSX 303). They put almost no majority-black blocks in District 22. (Id,.). But we cannot say that the line was irregular:
Castleberry Fire Dept.-1 Precinct in Act 603
[[Image here]]
*1096(Id.). The legislature assigned 225 people to District 23, 85 percent of them black, and 736 people to District 22, only 7 percent of them black. (Def. Ex. 475). Third, in Herbert FD precinct, the drafters split the precinct roughly down the middle with a sensible line and no apparent racial sorting.
Herbert FD Precinct in Act 603
[[Image here]]
*1097(APSX 304). Fourth, in Paul Fire Dept. precinct, the drafters used an irregular line in two places to put all three of the significant majority-black blocks and only one majority-white block in District 23.
Paul Fire Dept. Precinct in Act 603
[[Image here]]
(APSX 305). The drafters assigned 137 people to District 23, 58 percent of them black, and 122 people to District 22, only 2 percent of them black. (Def. Ex. 475).
The choices in Marengo County provide almost no evidence that race predominated. The drafters preserved the core of the district in Marengo County and smoothed out the line between Districts 23 and 24, as compared to 2001. The portion of Marengo County in District 23 barely had a higher black population percentage (56 percent) than the county as a whole (52 percent). (Def. Ex. 475; Doc. 297-4 at 7). The drafters split only one precinct, Cornerstone Church, and they split it with another majority-black district using no racial pattern. (APSX 337).
The choices in Monroe County more clearly reflect the predominance of race. The district noticeably retreated in 2012, ceding territory to overpopulated District 22 — a counterintuitive choice.
*1098Monroe County in 2001
Monroe County in 2012
(Ala. Reapportionment Office, supra). We recognize that sometimes an underpopulated district must give up population in one place to gain it in another, but the defendants gave no explanation why they did so here. This shape put a significantly higher percentage of black population in the district (68 percent) than the county as a whole (42 percent). (Def. Ex. 475; Doc. 297-4 at 6). In contrast, the black population percentage under the 2001 lines using 2000 Census data (43.9 percent) more closely matched the county as a whole in 2001 (40.4 percent). (Doc. 30-44 at 11; Def. Ex. 475). Based on these facts, we infer that race drove the drafters’ choices in Monroe County.
Four split precincts in Monroe County provide further evidence that race predominated. First, in Mexia Fire Station precinct, the drafters put a narrow majority-black census block running parallel to the border in District 23. That was the only majority-black block in the precinct, although the number of people moved was small and the shape is not any more irregular than the precinct line.
Mexia Fire Station Precinct in Act 603
[[Image here]]
*1099(APSX 346). Second, the split in Monroe Beulah Church precinct follows a straight line, except where doing so would put two small majority-black blocks in District 22. This irregular line put all but one majority-black block in District 23.
Monroe Beulah Church Precinct in Act 603
[[Image here]]
(APSX 347). The drafters assigned 71 people to District 23, 72 percent of them black, and 141 people to District 22, only 13 percent of them black. (Def. Ex. 475). Third, in Monroeville Armory, the split puts all but one majority-black block in District 23. The District 23 side of the irregular border is formed almost entirely by majority-black blocks, while the District 22 side of the border is formed entirely by majority-white blocks.
Monroeville Armory Precinct in Act 603
*1100[[Image here]]
(APSX 348). The drafters assigned 1,524 people to District 23, 51 percent of them black, and 1,247 people to District 22, only 23 percent of them black. (Def. Ex. 475). In Perdue Hill Masonic Lodge precinct, District 22 took all of the reasonably accessible majority-white blocks in the east.
Perdue Hill Masonic Lodge Precinct in Act 603
*1101[[Image here]]
(APSX 350).
The other precinct splits in Monroe County are less suspicious. In Bethel Bapt House precinct, the drafters put a majority-white block in District 22 and mostly majority-black blocks in District 23, but this decision smoothed out an irregular precinct shape.
Bethel Bapt House Precinct in Act 603
*1102[[Image here]]
(APSX 343). In Chrysler-Eliska-McGill precinct, they drew a smooth line that placed a sparsely populated majority-black area in District 23 but left majority-black blocks along the border in District 22.
Chrysler-Eliska-McGill Precinct in Act 603
[[Image here]]
(APSX 344). In Days Inn-Ollie precinct, the drafters again smoothed out irregular precinct lines in this and neighboring precincts with a split along mostly zero-population areas,
Days Inn-Ollie Precinct in Act 603
*1103[[Image here]]
*1104(APSX 345). That straight line, continued into Monroeville Housing Auth precinct.
Monroeville Housing Auth Precinct in Act 603
[[Image here]]
*1105(APSX 349). In Purdue Hill precinct, one majority-black block falls on the District 23 side of a roughly even break.
Purdue Hill Precinct in Act 603
[[Image here]]
(APSX 351). And in Shiloh Grimes pre-cinet, the split shaved off an irregular bulge in the precinct.
Shiloh-Grimes Precinct in Act 603
[[Image here]]
*1106(APSX 352).
The drafters removed Autauga County in 2012, but the Democratic Conference plaintiffs kept part of the county in Plan A. (Doc. 287-19 at 2). The plaintiffs have not explained why keeping Autauga County was required by traditional districting criteria, and there is reason to think that removing this county, which is not part of the Black Belt, creates a stronger community of interest in the district.
The plaintiffs argue with some force that, given the extent of the changes to the districts, the only way the drafters could have maintained black population percentages as close as they did was through a policy of racial targets. The enacted district had a black population percentage of 64.84 percent, almost the same as the 64.76 percent black population in the benchmark. (Doc. 30-39 at 2; Doc. 30-41 at 1). In the hypothetical district with no precinct splits submitted by Alabama, the black population percentage would be 1.2 points lower, causing the district to miss the purported target. (Doc. 263-3 at 2). The Democratic Conference plan, which contained fewer counties and fewer split counties, had a total black population percentage of only 58.9 percent, nearly six points lower. (Doc. 296-1 at 4). Alabama has not explained why the Democratic Conference district violates the Committee guidelines or federal law, and that plan suggests that some of the county and precinct splits may have been race-based.
Whether race predominated in the design of District 23 is a close question. The shape is not bizarre. The legislature added one county overall, bringing the total to ten, in a district where it needed to add population. In the process, it also decreased the number of split counties. Some of the choices about particular counties are not suspicious, and neither are some of the precinct splits. But there are also suspicious split counties and suspicious split precincts. The legislature managed to meet its target almost exactly, which it would not have accomplished without the precinct splits. And the Democratic Conference plaintiffs managed to draw a sensibly shaped district with fewer county splits, two fewer counties overall, and a significantly lower black population percentage. We find that race predominated in the design of District 23.
Because we find that race predominated, we must decide whether District 23 survives strict scrutiny. We conclude that the district satisfies strict scrutiny because the state had a strong basis in evidence to believe that ability to elect for purposes of the Voting Rights Act required black population percentages of 62 to 65 percent in this area. The drafters drew a district in that range.
Senator Sanders, the longtime incumbent in District 23, urged the Committee to maintain at least 62 to 65 percent black majorities in the majority-black districts. He gave clear testimony with several reasons for his conclusion:
One of many concerns is we are not to have any less African-American — the majority African-American districts than you have, and that those districts ought not be less than 62 percent. And I just want to say why 62 percent, ought not to be less than 62 percent. Many times a population of a district is not reflective of the voters at all in that district. Sometimes a lot of people don’t vote. Sometimes a lot of people can’t vote. They might be in prison or other kinds of institutions. Sometimes a lot of folks are discouraged for one reason or another. So I would hope that 62 percent is a minimal for the majority African-American district.
*1107(Doc. 30-28 at 6). Representative Thomas Jackson, whose House district in the western Black Belt overlaps with Senate District 23, likewise told the Committee that the district should be “sixty-two percent or sixty-five percent.” (Doc. 30-23 at 8). Senator Dial testified that, if he had told the black leadership in the Senate that it could have no more than 55 percent black population percentage, “Senator Sanders and my other good friends in the Senate ... would simply have glazed over and asked me when I was going to the mental institute.” (Doc. 215 at 44-45). The legislature could reasonably treat Senator Sanders and Representative Jackson as informed about voting patterns in their home districts.
The plaintiffs’ evidence at trial and the testimony of one of their experts in 2000 confirmed what Sanders and Jackson told the Committee. Dr. Joe Reed testified that a majority-black district needs to be at least 60 percent black to allow minority voters to elect the candidate of their choice. (Doc. 216 at 159-60). And in 2000, Dr. Theodore Arrington, an expert witness for the plaintiffs in this litigation, testified that a 61 percent majority in nearby Dallas County did not guarantee black voters the ability to elect a candidate of their choice for county commission. (Doc. 203 at 84; Doc. 217 at 80-81). See Wilson v. Jones, 130 F.Supp.2d 1315, 1326 (S.D. Ala.) (“Dr. Arrington’s position [is] that at least a 62% black voting age population was needed to assure blacks’ an opportunity to elect their choices in a district.”), aff'd sub nom. Wilson v. Minor, 220 F.3d 1297 (11th Cir. 2000). Although we agree with our dissenting colleague that the legislature did not use this evidence as its basis fot District 23, (Dissent at 1363-64 n.9), it does confirm the reliability of the district-specific evidence that it did have.
The Supreme Court has explained that “the narrow tailoring requirement insists only that the legislature have a ‘strong basis in evidence’ in support of the (race-based) choice that it has made.” Ala. Legislative Black Caucus, 135 S.Ct. at 1274 (quoting Brief of the United States as Amicus Curiae 29). A strong basis in evidence consists of “good reasons to believe such use [of race] is required,” but it need not “actually be necessary to achieve a compelling state interest.” Id. (quoting Brief of the United States as Amicus Curiae 29). If the detailed comments of an influential longtime incumbent in the district and a consistent statement by the incumbent in an overlapping House district, provided to the Committee at a hearing for the purpose of gathering input, do not fit this description, then the burden of proving “good reasons” has been transformed into a burden of proving actual necessity.
We reject our dissenting colleague’s argument that the comments of Senator Sanders and Representative Jackson are “exactly the type of stereotyping about black voting behavior that strict scrutiny is intended to prohibit.” (Dissent at 1368). Our colleague makes this charge repeatedly. (Id. at 1356-57, 1358-59, 1362, 1364-65, 1367-68). Although we agree that strict scrutiny forbids the use of racial stereotypes, it denies reality to suggest, as the dissent does, that the drafters relied on racial stereotypes instead of “demograph-ie[ ]” support, (id. at 1368), when they followed the suggestions of Senator Sanders and Representative Jackson. The drafters relied on these comments, as they were urged to do, because Sanders and Jackson possessed intimate knowledge of the concerns of the constituents of District 23.
In their public comments for this redistricting, both Sanders and Jackson, as members of the Alabama Legislative Black *1108Caucus, complained that the legislature had failed to listen to their concerns about redistricting in the last cycle. At a public hearing about redistricting, Representative Jackson protested the drafters’ disregard for his constituents’ interests during the previous redistricting effort and his hope that this time around the drafters would listen to his constituents: “You know, during the last reapportionment hearing, ... [y]ou ... made Montgomery County three different districts and you brought the school board all the way down to Mobile. We fought that. And I’m asking today, ... are we going to be heard?” (Doc. 30-23 at 7). He later observed, “[n]obody listens too well to us.” (Id. at 8). Similarly, Senator Sanders voiced his “concerns” that the drafters would not heed the interests of his constituents and would instead either retrogress or pack his district. (Doc. 30-28 at 6-7).
As already explained, Senator Dial, one of the Republican leaders of the redistricting process, testified that he worried that his colleagues would react with disdain— they “would simply have glazed over” — if he refused to follow their advice and imposed a lower black population percentage on their districts. (Doc. 215 at 44-45). But the dissent suggests that the Republican leaders of the reapportionment hearings should have retorted, to Jackson and Sanders: “We appreciate our colleagues’ concern, but your comments are ‘generic and conclusory’; they ‘lack[] precise recommendations’; and ‘they constitute exactly the type of stereotyping about black voting behavior that strict scrutiny is intended to prohibit.’ ” (Dissent at 1359,1368).
The dissent’s contention that the drafters were wrong to rely on this evidence and, worse, that they engaged in pernicious racial discrimination by doing so 'highlights the predicament of the drafters. Either the drafters could have adhered to the suggestions and, under the view of the dissent, engage in racial stereotyping, or the drafters could have disregarded these comments and then faced accusation of not “listen[ing] too well.” (Doc 30-23 at 8).
The dissent faults Senator Sanders in particular for imprecision in his use of “sometimes,” “could be,” and other figures of speech, (Dissent at 1356, 1359-60,1368), but Sanders’s testimony was powerful, concise, on point, and given by an expert in the politics of the Black Belt and Senate District 23. Sanders, a Harvard-educated lawyer, has served in the Alabama Senate for 34 years; his district includes parts of ten counties, and has in the past included part of one other; he has had an active trial practice throughout the region for many years, e.g. Sellers v. Lowndes Cty. Bd. of Educ., 550 So.2d 1021, 1021 (Ala. Civ. App. 1989) (listing “Hank Sanders” of the law firm “Chestnut, Sanders, Sanders, Turner, Williams & Pettaway” as attorney for the appellee), including in this court; and during his decades of public service, Senator Sanders has spoken at countless meetings on topics of public concern. (See Doc. 30-28). Nor is he a redistricting neophyte. He has participated in redistricting after the 1990, 2000, and now 2010 censuses. E.g., Gustafson v. Johns, 434 F.Supp.2d 1246, 1251 (S.D. Ala. 2006) (explaining that “Senator Hank Sanders” was a “[d]efen-dant-[i]ntervenor” in an earlier action that challenged a 2001 redistricting plan). As Senator Dial testified at trial, Senator Sanders is a highly respected senator in the State of Alabama. (Doc. 215 at 37-38).
The Legislature had strong reasons for relying on Senator Sanders’ testimony. Senator Sanders had no need for demographic studies to form his political opinions, and he did not require exact voting behavior on a precinct level. And we reject the dissent’s contention that Senator Sand*1109ers engaged in racial stereotyping. (Dissent at 1356-57, 1358-59, 1362, 1364-65, 1368).
The dissent also argues that the public comments are a litigating position concocted by Alabama after the fact, (Dissent at 1363-64), but the evidence proves otherwise. Senator Sanders and Representative Jackson gave public statements at an information-gathering hearing held by the redistricting committee. (See Docs. 30-23, 30-28). Senator Dial testified at trial that he considered Senator Sanders’s opinion:
Q. Do you remember the hearing in Selma?
A. Yes.
Q. Okay. And did Senator Hank Sanders attend that hearing?
A. Yes.
Q. Did he at any time give you any instructions about his district or about African American districts in general?
A. He did both. I had talked to Senator Sanders. He realized his district had to grow.... So he gave me some instructions on how he thought his district should grow. He also told me, and it’s public record, that he felt like that the minority districts should be at a minimum 62 percent minorities.
(Doc. 215 at 37). Senator Dial also testified that he considered Senator Sanders a credible source: “I’ve worked with Senator Sanders for years. We’ve been together on issues and opposed on issues, and I consider him a viable member of the Alabama Senate and basically a spokesman for the minorities in the state of Alabama.” (Id.). The defendants rely on this testimony about Senator Sanders’s credibility, and the plaintiffs fail to rebut it.
Nor are the dissent’s other objections to Senator Sanders’s and Representative Jackson’s testimony persuasive. The dissent argues that Senator Sanders’s “comments were not geographically specific,” (Dissent at 1362), but Senator Sanders was the longtime incumbent in District 23 — a specific geographic area — and he testified at the hearing in that capacity. (Doc. 30-28 at 1). Indeed, he made specific geographic suggestions about his district to the drafters. (Id. at 13). The dissent also asserts that Senator Sanders and Representative Jackson proposed “imprecise remedies.” (Dissent at 1362). But Jackson’s and Sanders’s comments were specific — they proposed percentages they thought necessary to prevent retrogression.
The Democratic Conference plaintiffs reply to Alabama’s evidence in several other ways, all of them unpersuasive. First, they argue that a bizarrely shaped district cannot be narrowly tailored, (Doc. 272 at 28-29), but District 23 is not bizarrely shaped. Second, they argue that Alabama is estopped from arguing for a higher black population percentage than it argued for in 2001. (Doc. 272 at 30). But with time comes new data, including recent election trends of which the two incumbents in this area would be aware. Third, the Democratic Conference plaintiffs offer evidence that a lower black population percentage would have been sufficient. (Id. at 30-31). But this evidence fails to prove that the legislature lacked a strong basis in evidence in the form of Senator Sanders’s testimony, and it instead proves only that the plaintiffs may have had a strong basis in evidence for a different percentage. The plaintiffs assert that we credited Alan Lichtman’s testimony that a bare majority provides a sufficient ability to elect, (Id. at 38; Doc. 287 at 19-20), but we actually discredited Lichtman’s methodology:
[W]e do not credit Lichtman’s opinion that race is the motivating factor for this voting pattern in Alabama. Lichtman did not conduct any statistical analysis to determine whether factors other than race were responsible for the voting pat*1110terns. He did not consider affluence, strength of a political campaign, or party-loyalty. Instead, he asserted repeatedly that the resulting voter patterns were similar, which suggests that race is the motivating factor. Lichtman also did not conduct any analysis of Democratic primaries between black and white candidates, which might have offered further evidence about whether white voters are more likely to support white Democrats and black voters are more likely to support black Democrats.
(Doc. 203 at 78-79 (citations omitted)). The plaintiffs still have not provided evidence of this sort. They rely heavily on the number of uncontested elections, but that fact does not prove anything about contested elections. Fifth, they point to the testimony of other legislators proposing different percentages, (Doc. 272 at 38), but that legislators from other areas wanted lower percentages does not disturb the strong basis in evidence provided by the comments of two incumbents from the area.
The Democratic Conference plaintiffs also mount new attacks in their briefs explaining Plan A. They observe that both Sanders and Jackson, near the end of the special session, voted for plans with lower black population percentages. (Doc. 301 at 19). But legislators may vote for a bill for varied reasons, and those choices alone do not impeach the legislators’ earlier comments to the Committee. The fact remains that the legislature had these public statements from the influential incumbent of the district and from the incumbent of an overlapping house district, and there is evidence that they relied on that testimony. The Democratic Conference plaintiffs also argue that the Committee ignored other public comments, (Doc. 301 at 19), but it would be impossible to heed every comment.
In addition, the dissent argues that Senator Sanders submitted a map that did not comply with his own advice and that our dismissal of his alternative plan is our attempt to bolster his credibility. (Dissent at 1363). But the dissent misunderstands our analysis. The question is not whether we choose to credit Senator Sanders’s comments at the public hearing over his alternative plans, but whether the legislature was justified in making that choice. After the legislature chose to gather input about its redistricting effort through public hearings, the legislature was entitled to rely on testimony from those hearings.
Nor are the district court decisions cited by the dissent dispositive here. In Page v. Va. State Bd. of Elections, No. 3:13-CV-678, 2015 WL 3604029 (E.D. Va. June 5, 2015), the district court struck down a redistricting plan that employed a racial target because the drafters presented no supporting evidence that the target was necessary to comply with the Voting Rights Act. Id. at *18. And in Smith v. Beasley, 946 F.Supp. 1174 (D.S.C. 1996), the court invalidated a similar redistricting plan that employed a racial target without supporting evidence. Id. at 1210. By contrast, the testimony of the longtime incumbent of Senate District 23 provided a strong basis in evidence for the choices of the drafters.
Finally, the Black Caucus plaintiffs argue that Alabama should not be allowed to offer new arguments about strict scrutiny, (Doc. 271 at 27-28) but Sanders’s and Jackson’s testimony is not new. It was in the trial record, and we cited it in our previous final opinion. (Doc. 203 at 27-28). Even if it were a new argument, Alabama should be allowed to make new arguments on remand to defend against a new challenge brought on remand. We will not enjoin the use of Senate District 23.
*1111c. Senate District 24 (West Black Belt)
Although it is another close call, we find that race did not predominate in the design of Senate District 24. The plaintiffs drew an alternative plan with one fewer county, but we cannot say that this map— or any other evidence submitted by the plaintiffs — proves that race predominated in the assignment of a significant number of residents. The plaintiffs have failed to prove that the shape of the district, the black population percentage, the precinct splits, the choices of which counties to include, the choices about how to split counties, or the choice to include Clarke County were the result of race predominating over traditional districting criteria.
District 24 was underpopulated in 2012 and constrained by Mississippi to the west and District 23, also underpopulated, to the east. (Doc. 1 at 10). These constraints explain why the drafters had to move 70,-988 people to repopulate this district. Hi-naman “took it down further into Choctaw [County] and Clark[e] [County]” to the south and into Pickens County to the north. (Doc. 217 at 123). He also added a “little bit more” population in the Tuscaloosa area. (Id.)
We cannot say that the shape of District 24 is so bizarre that it is proof that race predominated. The borders of District 24 are no less regular or compact than they were in 2001:
[[Image here]]
(Ala. Reapportionment Office, supra).
Nor are they any more unusual than the alternatives proposed by the plaintiffs. The southern border of the Democratic Conference plaintiffs’ district is no less irregular than the southern border in Act 603, and the Democratic Conference plaintiffs’ district appears no more compact.
Senate District 24 in Democratic Conference Plan A and Act 602
*1112[[Image here]]
(Doc. 287-24 at 2). The shape of the 1% Plan district is actually more irregular and less compact. The following map shows the alternative district in light blue:
Senate District 24 in Black Caucus 1% Plan and Act 603
[[Image here]]
(APSX 538).
Looking at the district as a whole, the black population percentage is not suspicious. The enacted district had a total black population percentage of 63.22 percent, (Doc. 30-39 at 2), just meeting its purported target of 62.78 percent, (Doc. 30-41 at 1). Plan A and the 1% Plan had slightly lower percentages of 59.3 and 57.31 percent, respectively. (Doc. 287-2 at *11131; APSX 470). These percentages are not different enough to be strong evidence of racial predominance.
The dissent argues that the black population percentage combined with the testimony of Senator Dial and Senator Keahey provides direct evidence that the state “intended” for race to predominate in the drafting of District 24, (Dissent at 1390, 1392-93), but the testimony is at best equivocal. Senator Dial testified at trial that District 24 “had to grow” because of massive underpopulation in the Black Belt districts. (Doc. 215 at 48). The dissent also plucks Dial’s testimony that District 24 “had to have more minorities” from the record, (id.), but this testimony is not direct evidence of racial predominance. All the statement suggests is that Dial considered race in the drafting of the district, which is permitted.
Senator Keahey’s testimony supports this view. At trial, Keahey testified that Dial would consider amendments to the proposed districts if the affected senators were “in support of’ the amendment, and the amendment did not retrogress “minority districts.” (Id. at 192). When asked what he thought Dial meant by retrogression, Keahey responded that it meant that the enacted plan should “not dilute the minority population.” (Id.). Keahey then clarified his answer that Dial thought preventing retrogression meant precluding a reduction in the population of black voters. (Id.). At most this evidence shows that the drafters considered race among other factors— which is permitted — not that race was the predominant factor in the drafting of Senate District 24.
Both alternative plans improve on the enacted district by reducing the number of counties from eight to seven and the number of split counties from six to two, (Doc. 300-1 at 95-96), but this improvement is insufficient on its own. District 24 has the same number of counties (eight) and the same number of split counties (six) in both the 2001 and 2012 plans. (Doc. 30-40 at 8-9; Doc. 30-44 at 12). And a difference of one county is not suspicious by itself, especially in a district that even the plaintiffs concede should have seven counties. Moreover, it was sensible to take population from overpopulated District 22 in Clarke County, as the enacted plan does, instead of severely underpopulated District 23 in Marengo and Hale Counties, as the alternative plans do. (Doc. 1 at 10).
Several counties remain exactly or generally the same. Sumter and Greene Counties stayed whole, and Tuscaloosa County kept its hook. The Democratic Conference plaintiffs’ own map establishes that the extension into Tuscaloosa County brought in — and failed to include — both white and black areas.
Close-Up of Senate District 24 in Act 603 with Black Population Percentages
*1114[[Image here]]
(ADC Supp. Ex. 38E.) And the black population percentage of the Tuscaloosa County portion of the district is almost identical to what it was in the 2001 plan: it changed from 58.9 percent black to 60.9 percent black. (Def. Ex. 408 at 708; Doc. 30-40 at 9). The plaintiffs made much of this “contorted, bizarrely-shaped hook,” (Doc. 258 at 35), until they included a hook in their own plans. Forced to defend that choice, the Democratic Conference plaintiffs’ ex'pert testified that he could “look at [the hook] as a core of [the] district.” (Doc. 296-7 at 136). The plaintiffs also informed us that the hook “has been a fixture since at least ... 1983.” (Doc. 287 at 17). We agree that the hook is not suspicious, and. the portions of District 24 in Tuscaloosa, Greene, and Sumter counties provide no evidence that race predominated.
Several other counties in District 24 saw changes. The drafters added a portion of Pickens County along the western border of the state, extended the district in the northeastern part of Hale County, smoothed out the existing line down the middle of Marengo County, pushed the district slightly into Clarke County, and stretched the district farther south in Choctaw County. We discuss these counties one by one.
We no evidence ⅛ Pickens County that race predominated. The Democratic Conference plaintiffs included all of Pick-ens County, and the Black Caucus plaintiffs failed to articulate a reason why they did not enter the county, so we find that entering Pickens County does not prove that race predominated. The plaintiffs and the dissent also have failed to prove that the decision to split Pickens County instead of keeping it whole was the result of race predominating over traditional criteria. Act 603 put a 74 percent black population in the district from a county that was only 42 percent black, (Def. Ex. 475; Doc. 263 at 85), but the shape sensibly anchors to the western border of the state and has a relatively smooth line to the border with Greene County. And, while we agree with our dissenting colleague that Senator Dial and the drafters considered race when they drafted District 24, (Dissent at 1390-91), the Supreme Court permits the consideration of race.
*1115Act 603 split one precinct in Pickens County, Carrollton 4 Service Center, and we find that race did not predominate. The split put 770 people in District 24, 78 percent of whom were black, and 889 people in District 21, 28 percent of whom were black. (Def. Ex. 475). But the split made the district line more regular, and it placed majority-black blocks along the border in majority-white District 22 and majority-white blocks along the border in District 24.
Carrollton 4 Service Center Precinct in Act 603
[[Image here]]
(APSX 360). As best we can tell, any racial disparity is the result of demographics.
The split of Marengo County also provides no evidence that race predominated. The drafters preserved the core of the district by drawing a line in roughly the same place through the middle of the county. They also placed a lower black population percentage into the district than the county had as a whole, 50 percent to 52 percent. (Def. Ex. 475; Doc. 263). If the legislature had made Marengo County whole, as both plaintiffs did, it would have raised the black population percentage. And the only split precinct in Marengo County, Cornerstone Church, is split with District 23, another majority-black district. The plaintiffs do not explain how this split with a majority-black district along smooth lines proves that race predominated.
*1116Cornerstone Church Precinct in Act 603
[[Image here]]
(APSX 337).
We find no evidence in the split of Clarke County that race predominated. Neither plaintiff included Clarke County in their alternative districts, and District 24 did not include Clarke County in 2001. The split placed a 61 percent black population in the district, higher than the 44 percent black population in the county as a whole. (Def. Ex. 475; Doc. 263 at 85). Traditional districting criteria might dictate taking new territory in counties that the district would have included anyway, but it seems equally consistent with traditional district-ing criteria to take population from overpopulated District 22 in Clarke County instead of underpopulated District 23 in Marengo or Hale Counties, as both plaintiffs did. (Doc. 1 at 10). In addition, the line in the enacted plan is not particularly suspicious and splits only three precincts with majority-white District 22, none of which proves that race predominated. In the Ba-shi Methodist Church and Thomasville National Guard Armory precincts, the drafters dealt with irregularly shaped precincts. They pushed south into both precincts to reach a large cluster of majority-black blocks and stopped before reaching majority-white blocks.
Bashi Methodist Church and Thomasville National Guard Precincts in Act 603
*1117[[Image here]]
(APSX 293). But with this split, the district picked up the core of Thomasville, respecting the traditional districting principle about communities of interest. The following map from the Census Bureau, of which we take judicial notice, shows the boundary of District 24 with a purple line and Thomasville in tan, with streets as gray lines and highways as yellow lines:
City of Thomasville in Act 603
*1118[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State Senate District 24 (Alabama), http://www2.census. gov/geo/maps/delOmap/SLD_RefMap/ upper/st01_al/sldu01024/DC10SLDU 01024_006.pdf). The third split, of Fulton City precinct, sensibly placed populated areas in the northeastern part of the precinct with adjacent populated areas in District 24 instead of zero-population and sparsely populated areas in District 22. The line also smoothed out the irregular shape of the precinct.
Fulton City Precinct in Act 603
*1119[[Image here]]
(APSX 294).
In Choctaw County, we find insufficient evidence that race predominated. The drafters preserved the core of the district in Choctaw County. They added territory along the Mississippi border and the southern county line, presumably to repopulate the district. Both plaintiffs also included Choctaw County in the district, so the choice to include it was not suspicious. The Democratic Conference plaintiffs also split the county. The line that the drafters used, although different from the line in Plan A, is not suspicious. The drafters included only a moderately higher percentage of black population from Choctaw County than the county had overall, 54 percent to 44 percent, (Def. Ex. 475; Doc. 263 at 85), and the Democratic Conference plaintiffs actually put a higher percentage of black population, 63 percent, in the district. (Doc. 297-4 at 7).
The drafters split seven precincts in Choctaw County, but we do not find a racial pattern in five of the splits. First, in Bogueloosa precinct, the most irregular part of the split is between majority-white blocks, and the drafters placed two majority-black blocks in District 22 that they could have placed in District 24:
Bogueloosa Precinct in Act 603
*1120[[Image here]]
(APSX 287). Second, the split of Branch-Bladon Springs-Cullomburg precinct placed a significantly higher percentage of black population in District 24 than District 22, but it does not snake around looking for black population and removed only 53 people from the precinct. (Def. Ex. 475).
Branch-Bladon Springs-Cullomburg Precinct in Act 603
*1121[[Image here]]
(APSX 288). Third, the split of Butler-Lacava-Mt. Sterling Voting District precinct smoothed out an irregularity in the shape of the precinct.
Butler-Lacava-Mt. Sterling Voting District in Act 603
*1122[[Image here]]
(APSX 289). The Democratic Conference plaintiffs also split this precinct and put a similar percentage of black population in the district (39 percent) as the legislature did (37 percent). (Def. Ex. 475; Doc. 296-6 at 1). Fourth, we find no racial pattern in the split of Silas-Souwilpa-Isney-Toomey Voting District because the split left several majority-white blocks along the border:
Silas-Souwilpa-Isney-Toomey Voting District
*1123[[Image here]]
(APSX 291). Fifth, the split of Riderwood-Rock Springs precinct managed to place 43 people in District 24, none of'them black. (Def. Ex. 475). But the Democratic Conference plaintiffs also split this precinct and put a higher percentage of black population in the district (62 percent) than the legislature did (49 percent). (M; Doc. 296-6 at 1). As the following map illustrates, the legislature left several majority-white blocks along the border:
Riderwood-Rock Springs Precinct in Act 603
*1124[[Image here]]
(APSX 290). The record does not establish that race predominated in the split of this precinct.
There are two mildly suspicious split precincts in Choctaw County, but they do not provide meaningful evidence that race predominated. First, the split of Lusk-Pleasant Valley-Ararat Voting District is suspicious because it put only majority-black blocks and zero-population blocks in District 24 using an irregular shape, but the split put only 55 people in the district. (Def. Ex. 475).
Lusk-Pleasant Valley-Ararat Voting District in Act 603
*1125[[Image here]]
(APSX 289). Second, the split of Toxey-Gilbertown-Melvin-Hurricane Voting District extended District 24 into majority-white areas and placed 34 people from three majority-black blocks in the district:
Toxey-Gilbertown-Melvin-Hurricane Voting District in Act 603
*1126[[Image here]]
(APSX 292). We cannot draw any strong inferences from this mixed pattern.
We find at most slight evidence in Hale County that race predominated. The drafters reduced the concavity of the split, and they preserved the bulk of the district in this county. The black population percentage from Hale County in District 24 (67 percent) is only slightly higher than the percentage in the county overall (59 percent). (Def. Ex. 475; Doc. 263 at 85). The line is sensible in overall shape, and it split only three precincts, all of them sparsely populated. First, the drafters split the irregularly shaped Havanna A precinct along a straight line.
Havanna A Precinct in Act 603
*1127[[Image here]]
(APSX 306). The split put 122 people in District 24, 43 percent of them black, and 53 people in District 14,11 percent of them black. (Def. Ex. 475). Second, they split Valley B precinct in a way that may have sorted residents by race.
Valley B Precinct in Act 603
[[Image here]]
(APSX 307). But this precinct is also sparsely populated: the split put 58 people in District 24, 59 percent of them black, and 36 people in District 14, 19 percent of them black. (Def. Ex. 475). Third, the *1128drafters split the sparsely populated Valley C precinct.
Valley C Precinct in Act 603
[[Image here]]
(APSX 308). The split put only 22 people in District 24, 64 percent of them black, and only 57 people in District 14, 23 percent of them black. In all, the precinct splits added 202 people to a district with 137,724 people, or 0.15 percent of the population. (Def. Ex. 475).
There is also one county, Washington, that the Black Caucus plaintiffs added and the legislature did not. But the Black Caucus plaintiffs did not explain why this choice was required by traditional district-ing criteria, and no reason is apparent to us from the record.
The Black Caucus plaintiffs make one other argument on remand, that a desire to maintain racial targets in Senate District 24 led Senator Dial to “drastically” change Senate District 22, Senate District 1, and other districts. (Doc. 256 at 155). But the plaintiffs have not challenged those districts, and a plaintiff has standing to challenge only his own district as a racial gerrymander. See Hays, 515 U.S. *1129737, 115 S.Ct. 2431; Sinkfield, 531 U.S. 28, 121 S.Ct. 446. Senator Dial testified that the need to repopulate Senate Districts 23 and 24 had effects on other districts, (Doc. 215 at 45-46, 48-50), but the Black Caucus plaintiffs do not explain how these changes prove that race predominated over traditional districting criteria in the challenged districts. Further, both districts needed to gain population by growing and that growth would have affected other districts in any plan.
We have weighed the evidence and arguments presented by the plaintiffs. Everything except four precinct splits is clearly consistent with traditional districting criteria, and we cannot say that race predominated even if we give the benefit of the doubt to the plaintiffs on all four ambiguous splits. The splits — Lusk-Pleasant Valley-Ararat Voting District, Toxey-Gilber-town-Melvin-Hurrieane Voting District, Valley B, and Valley C — assign only 433 black people to District 24, which is less than half a percent of the total population of the district. If we unsplit these four according to our method that assigns the entire precinct to the district that took the majority of it, the black population would decrease by only 0.17 percent and still meet the purported target. If we remove all four precincts entirely, the black population percentage increases by only 0.11 percent. Based on the totality of the evidence, we find that race did not predominate in the design of District 24.
d. Senate District 26 (Montgomery)
The Supreme Court expressed particular concern about Senate District 26. See Ala. Legislative Black Caucus, 135 S.Ct. at 1271. The overall shape of the district does not establish that race predominated, but the exhibits submitted on remand make clear that a significant number of people were assigned on the basis of race in five of the split precincts. Based on these new exhibits, we find that race predominated in the design of this district. Because the district also fails strict scrutiny, we must enjoin its use in future elections.
District 26 was underpopulated by 11.64 percent in 2010, (Doc. 1 at 10), so Hinaman removed the large, rural portion of southern Montgomery County from District 26 and added more population from the city of Montgomery. (Doc. 217 at 129-30). The following map shows District 26 in the center in green and District 30 to the north and west in gray:
[[Image here]]
(Ala. Reapportionment Office, supra).
The removal of the large rural portion made sense for two reasons. First, Senate District 25 needed a land bridge to Cren-shaw County to fix Senate District 30. (Doc. 217 at 129-30). Second, District 26 is primarily an urban district, and communi*1130ties of interest were better served by making a more compact, urban district.
The Democratic Conference plaintiffs also drew a district centered on the city of Montgomery and wholly contained within Montgomery County. Fairfax defended this choice at his deposition:
Q: Okay. So you’ve got no problem with the small, compact Senate District 26 focusing around the city limits and having Senate District 25 take the rural areas of Montgomery County and going into other counties to make up whatever — to grab the population it needs?
A: Correct. Whole counties, yes.
Q: That concept makes sense to you.
A: It is. It’s logical. Not necessarily in every case, but in this case, it made sense.
(Doc. 296-7 at 97).
But the Democratic Conference plaintiffs made two different choices. First, they removed outlying precincts along the southern half of the district and filled in the “lagoon” in the eastern part of the district. Second, they split the “crab claw” precinct in the southern part of the district, which has been the subject of much attention in this litigation.
Senate District 26 in Conference Plan A and Act 603
[[Image here]]
(Doc. 287-24 at 3). Fairfax conceded that splitting the crab claw was not required by traditional districting criteria:
Q: So what makes you a better decider of which one you do than the Legislature?
A: I don’t necessarily believe that I’m a better judgment [sic] for that. I made the judgment at that time that I thought was best.
Q: Okay. If someone made a different judgment that says I would rather have a whole precinct even if it’s slightly irregular, that’s not always a wrong choice; correct?
A: That could be done, definitely.
(Doc. 296-7 at 100-01).
The Black Caucus plaintiffs drew a very different district, straddling three counties *1131and pairing part of the city of Montgomery with rural counties. The following map shows the 1% Plan district shaded in light purple:
Senate District 26 in Black Caucus 1% Plan and Act 603
[[Image here]]
(APSX 589). We do not give any weight to this alternative district because Cooper admitted to splitting counties and choosing population on the basis of race. (Doc. 297-1 at 128-29).
The shape of District 26 in Act 603 does not establish that race predominated. The Supreme Court questioned the change “from rectangular to irregular,” Ala. Legislative Black Caucus, 135 S.Ct. at 1271, but the current district is more compact than the previous one and centers more on the city of Montgomery. The district includes 136, 451 people, 96 percent of whom live in the city of Montgomery, (Def. Supp. Ex. 6), up from 86 percent under the previous lines, (Doc. 30-44 at 38). Concentrating the district within city lines was a reasonable decision that furthered a traditional districting criterion. See Bush, 517 U.S. at 963, 116 S.Ct. 1941.
We have no concerns about either the claw or the lagoon. The supplemental exhibits proved that the incumbent lives in the claw, (Def. Supp. Ex. 5), and the claw follows precinct lines, (APSX 539). The lagoon is roughly similar to the same area under the 2001 plan:
[[Image here]]
(Ala. Reapportionment Office, supra). As the supplemental exhibits proved on remand, District 3 of the Montgomery County Commission roughly matches the *1132lagoon. (Def. Supp. Exs. 12-15). The following map shows District 26 shaded in green and the county commission districts enclosed by blue lines:
Senate District 26 and Montgomery County Commission Districts
[[Image here]]
(Def. Supp. Ex. 13). Because more than one session of the Alabama Legislature and more than one state entity have considered this area to be a community of interest, the shape of the lagoon does not help the plaintiffs prove that race predominated in the design of District 26.
We also reject the plaintiffs’ arguments about the land bridge to District 30. The plaintiffs assert that the drafters should have added Crenshaw County to District 26, (Doc. 258 at 37-38), but they provide no race-neutral reason for doing so, never explain how to solve the ensuing underpopulation of District 25, and do not follow their own advice in Plan A or the 1% Plan. (Doc. 287-22; APSX 539). The Democratic Conference plaintiffs further argue that the drafters did not consider District 25 when they drew District 26 because they “redrew the black majority districts first,” (Doc. 272 at 58), but there is no evidence that the drafters drew the majority-black districts while ignoring the rest of the districts. Hinaman was aware of the other districts even as he began with the majority-black districts.
We also do not consider the black population percentage in this district suspicious. The percentage increased by less than 3 points, from 72.69 to 75.13 percent black. (Doc. 263-2). The earlier alternative plans, Sanders and Reed-Buskey, had very high black population percentages, but the plans on remand had significantly lower percentages.
*11332010 Census Total Black Population Percentages in Senate District 26 Across Alternative Plans
[[Image here]]
The fact that District 26 could have a lower black population percentage does not convince us that the actual percentage is suspicious, especially because some of the alternative plans had similarly high percentages of black population.
Nor does the argument made by the Democratic Conference plaintiffs that their alternative district better matches the black population percentage of the city of Montgomery, 56 percent, persuade us. (Doc. 287 at 18). The city of Montgomery is too large for a single district. Because the state needed to split Montgomery, the plaintiffs must explain how the enacted district splits the city in a way that does not respect smaller communities of interest within the city. They have not done so.
Finally, we are not persuaded by the plaintiffs’ argument about racial predominance that the net population added to District 26 included 14,806 black people but only 36 white people. {See, e.g., Doc. 256 at 172). The net population statistic that the plaintiffs cite is misleading. And the Supreme Court may have been misled when it stated, “Alabama’s plan added 15,-785 new individuals, and only 36 of those newly added individuals were white.” Ala. Black Legislative Caucus v. Alabama, — U.S. -, 135 S.Ct. 1257, 1263, 191 L.Ed.2d 314 (2015). The Court later repeated that misleading statistic. Id. at 1271 (explaining that the addition of “just 36” white people to District 26 is “a remarkable feat given the local demographics.”). This representation of the evidence gives the false impression that the Alabama legislature sought out 15,739 black individuals and 36 white individuals to add to the existing population of District 26.
The evidence before us establishes that 11,966 white people and 6,858 black people were removed from Senate District 26; 12,002 white people (not a mere 36) and 21,664 black people were added to the district. Afterward, 14,613 white individuals and 80,856 black individuals remained in the district. In the end, a net total of 52,490 people, black and white, were moved into and out of District 26. 36 white individuals were added to District 26. In other words, post-redistricting, District 26 had 36 more white people than it had before redistricting. To us, that a large number of individuals — white and black— were swept in and out of the district is the significant fact, not that District 26 had a net gain of 36 white people. Although we find that race predominated in the drafting of District 26, and that it fails strict scrutiny, our conclusion does not follow from the supposed “remarkable feat” of the drafters.
But the supplemental exhibits on the precinct splits provide persuasive evidence that race predominated in District 26. Two of the precinct splits are not suspicious. In 3F Goodwyn Community Center Voting District, the drafters split the precinct along U.S. Route 231 and an unpopulated area. This choice respects traditional dis-tricting criteria by following a major road *1134and keeping populated areas together. (See Doc. 217 at 184 (testimony of Hinaman explaining that precinct lines “don’t necessarily follow roads and boundaries”)).
3F Goodwyn Community Center Voting District
[[Image here]]
(APSX 357).
Census Bureau Map in the Vicinity of 3F Goodwyn Community Center Voting District
[[Image here]]
*1135(U.S. Census Bureau, State Legislative District Reference Map: State Senate District 26, (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ upper/st01_al/sldu01026/DC10SLDU 01026_001.pdf). In 5M Bell Road YMCA Voting District, the split follows County Road 43, another choice that respects traditional districting criteria.
5M Bell Road YMCA Voting District Precinct in Act 603
[[Image here]]
*1136(APSX 359).
Census Bureau Map in the Vicinity of County Road 43
[[Image here]]
(U.S. Census Bureau, TigerWEB, http:// tigerweb.geo.census.gov/tigerweb/).
The other five splits are statistically and visually suspicious. None of these five precincts were in District 26 under the 2001 line. (Def. Ex. 409 at 742-43). In Act 603, all five are split with District 25, which is 22.82 percent black. (Doc. 263-2 at 2). The following .map shows District 26 in green, District 25 in purple, and precinct lines in blue:
Senate District 26 in Act 603
[[Image here]]
*1137(Def. Supp. Ex. 59 (precinct labels added by the Court)). The splits tend to put a higher percentage of black people in District 26 than District 25, with the effect of increasing the black population percentage in District 26 and keeping the black population percentage in District 25 below 25 percent. As drawn, District 26 is 75.13 percent black. (Doc. 263-2 at 2). If we unsplit the precincts using our method, the black population percentage in District 26 drops to 72.05 percent. If we put all of the suspicious precincts entirely in District 25, the black population percentage in District 26 drops to 71.19 percent. And if we put all of the suspicious precincts entirely in District 26, the black population percentage in District 26 drops to 69.47 percent.
We discuss each split in turn. First, the legislature pushed District 26 into two corners of 1A Cloverdale Community Center Voting District precinct. In the southwest corner, the district took only majority-black blocks in step-wise fashion and left no accessible majority-black blocks in District 25. In the north, District 26 absorbed several majority-white blocks in an irregular shape, but it left behind no adjacent majority-black blocks.
1A Cloverdale Community Center Voting District Precinct in Act 603
[[Image here]]
(APSX 353). The split placed 1,011 people in District 26, 68 percent of them black, and 6,739 people in District 25, only 16 percent of them black. (Def. Ex. 475).
Second, District 26 absorbed two parts of IB Vaughn Park Church of Christ precinct, both parts comprised of mostly majority-black blocks. In the northwest, the district could have reached more majority-black blocks, but did not do so. In the east, the split left behind no majority-black blocks.
IB Vaughn Park Church of Christ Voting District Precinct in Act 603
*1138[[Image here]]
(APSX 354). The split put 5,976 people m District 26, 56 percent of them black, and 3,895 people in District 25, only 25 percent of them black. (Def. Ex. 475).
Third, the legislature pushed District 26 into 1C Montgomery Museum of Fine Arts Voting District precinct to reach a cluster of majority-black blocks. Along most of the border, it stopped before reaching majority-white blocks. In one area, it left behind accessible majority-black blocks.
1C Montgomery Museum of Fine Arts Voting District Precinct in Act 603
[[Image here]]
*1139(APSX 355). But again, the racial pattern is clear: the split put 3,829 people in District 26, 69 percent of them black, and 3,599 people in District 25, only 37 percent of them black. (Def. Ex. 475).
The same is true of ID Whitfield Memorial United Methodist Church. The drafters reached almost all of the majority-black blocks in the precinct, although they excluded some accessible majority-black blocks and included some majority-white blocks that they could have left behind.
ID Whitfield Memorial United Methodist Church Precinct in Act 603
[[Image here]]
(APSX 356). The split put 4,564 people in District 26, 67 percent of them black, and 1,781 people in District 25, only 18 percent of them black. (Def. Ex. 475).
Fifth, in 3G Alcazar Shrine Temple Voting District precinct, the legislature put populous majority-black blocks and no majority-white blocks in District 26, although they again left behind an accessible majority-black block.
3G Alcazar Shrine Temple Voting District Precinct in Act 603
*1140[[Image here]]
(APSX 358). The split put 2,203 people in District 26, 80 percent of them black, and 1,411 people in District 25, only 43 percent of them black. (Def. Ex. 475).
These five splits put 17,583 people in District 26, 58 percent of them black. (Id.). In contrast, the splits'put 17,425 people in District 25, 25 percent of them black. (Id.). On the basis of the evidence about these five precinct splits, we find that race predominated in the design of Senate District 26.
We further conclude that District 26 does not survive strict scrutiny. Alabama makes no arguments about strict scrutiny that are specific to this district. It cannot rely on the testimony of Sanders and Jackson in this district because District 26 has a black population percentage above 65 percent. Based on the expanded record on remand, we must enjoin the use of Senate District 26 in future elections.
e. Senate District 28 (East Black Belt)
We find that race predominated over traditional districting criteria in the design of Senate District 28, which was underpopulated by 3.8 percent in 2010. (Doc. 1 at 11). The district contains all or part of seven counties (Barbour, Bullock, Henry, Houston, Lee, Macon, and Russell), and the black population percentage rose over eight points in the enacted plan to 59.83 percent. (Doc. 30-41 at 2; Doc. 30-39 at 3). The biggest changes to the district were the addition of a claw that reaches into Houston County to take in population from the city of Dothan and several protrusions into Lee County on the northern border of the district:
*1141[[Image here]]
(Alabama Policymaker’s Dashboard, sn-pra).
Democratic Conference Plan A drew a majority-black district, (Doc. 287-2 at 1), with a slightly less irregular protrusion into Houston County and no protrusion into Lee County.
Senate District 28 in Democratic Conference Plan A and Act 603
[[Image here]]
(Doc. 287-24 at 3). The Black Caucus 1% Plan drew a plurality-black district with no protrusion into Houston County and a slightly less irregular protrusion into Lee County,
*1142Senate District 28 in Black Caucus 1% Plan and Act 603
[[Image here]]
(APSX 540).
The drafters far exceeded any target of preserving a black population of 51.05 percent, (Doc. 263-3 at 2), but the expansion of the borders splits precincts in a race-driven fashion. We find it suspicious that the black population percentage increased by eight points as a result of redistricting. Neither Democratic Conference Plan A (51.7 percent black) nor the Black Caucus 1% Plan (50.98 percent black) had a similarly high percentage, so it is not obvious that this percentage was unavoidable. (Doc. 296-1 at 4).
The addition of population in Dothan had the effect of combining significant urban population with otherwise mostly rural population. In 2001, six percent of the population of the district lived in areas classified by the Census Bureau as urbanized areas. (Doc. 30-44 at 39-40; U.S. Census Bureau, Alphabetically Sorted List of UAs, http://www2.census.gov/geo/docs/ reference/ua/ua2k.txt). Although we would prefer to use data on urban and rural population in 2010 under the old district lines, the parties did not provide us with that information. Under the plans enacted in 2012, 18 percent of the population lived in an urbanized area. (Def. Supp. Ex. 6; U.S. Census Bureau, A National, State-Sorted List of All 2010 Urbanized Areas for the U.S., Puerto Rico, and Island Areas First Sorted by State FIPS Code, Then Sorted by UACE Code, http://www2. census.gov/geo/docs/reference/ua/ua_st_ list_ua.txt). Dothan accounted for less than one percent of the district in 2001, but it accounted for 16 percent of the district under the currently enacted plan, and 70 percent of the population included from Dothan in the enacted plan is black. This split also breaks the community of interest in Dothan. We do not base our finding of racial predominance solely on this change in the character of the district, but we observe that it is suspect.
This suspicion is confirmed by the individual precinct splits. The expansion into Dothan manages to grab almost every majority-black census block in the split precincts, while avoiding almost every majori*1143ty-white block. In Doug Tew Community Center precinct, the drafters split the precinct along irregular lines, putting all but two majority-black blocks in District 28.
Doug Tew Community Center Precinct in Act 603
[[Image here]]
(APSX 309). The resulting split put 3,196 people in District 28, 51 percent of them black, and 4,526 people in District 29, 15 percent of them black. (Def. Ex. 475). In Farm Center precinct, the drafters split the precinct along irregular lines, again putting almost every majority-black block in District 28.
Farm Center Precinct in Act 603
*1144[[Image here]]
(APSX 310). The resulting split put 927 people in District 28, 72 percent of them black, and 5,133 people in District 29, 25 percent of them black. (Def. Ex. 475). The drafters split the Library precinct with irregular lines that put almost all of the majority-black blocks in District 28.
Library Precinct in Act 603
*1145[[Image here]]
(APSX 313). The split put 3,330 people into District 28, 78 percent of them black, and 4,716 people in District 29, 12 percent of them black. (Def. Ex. 475). In Lincoln Community Center, the drafters put all but one of the majority-black blocks in District 28.
Lincoln Community Center Precinct in Act 603
*1146[[Image here]]
(APSX 314). The split put 1,253 people in District 28, 82 percent of whom were black, and 861 people in District 29, 15 percent of whom were black. (Def. Ex. 475).
Other splits in Dothan County are less clearly racial. The split of Johnson Homes put 4,762 people in District 28, 92 percent of whom were black, and 215 people in District 29, 20 percent of whom were black. (M).
Johnson Homes Precinct in Act 603
*1147[[Image here]]
(APSX 311). But the Democratic Conference plaintiffs split the precinct with a population that was 90 percent black in District 28 and a population that was 29 percent black in District 29, suggesting that such a split is unavoidable. (Doc. 296-6 at 1). The split of Kinsey precinct in the enacted plan put all of the majority-black blocks in District 28.
Kinsey Precinct in Act 603
*1148[[Image here]]
(APSX 312). But it also put majority-white blocks in District 28 that were not necessary to reach the majority-black blocks, (id.), and the Democratic Conference plaintiffs likewise split the precinct in a way that put a higher black population percentage in District 28 than in District 29, (Doc. 296-6 at 1). We cannot draw many conclusions from Vaughn Blumberg Center precinct, in which the drafters used straight lines with two irregularities to put all of the majority-black blocks in District 28. The irregularity in the west keeps populated areas together, but the irregularity in the south appears to be race-based.
Vaughn Blumberg Center Precinct in Act 603
[[Image here]]
*1149(APSX 315). And in Wiregrass Park precinct, the drafters used irregular lines that included most of the majority-black blocks, but also some majority-white blocks that were unnecessary to reach majority-black blocks,
Wiregrass Park Precinct in Act 603
[[Image here]]
*1150(APSX 316). On balance, the splits in Houston County tend to establish that race predominated.
Similarly, the split precincts in the northern extension tend to prove that race predominated. In Lee County, the drafters split the Waverly, Loachapoka, Auburn, Beuaregard School, and Marvyn precincts. First, in Waverly precinct, they drew an irregular shape that put nothing but majority-black and unpopulated blocks in District 28.
Waverly Precinct in Act 603
[[Image here]]
(APSX 336). The split placed 212 people in District 28, 85 percent of them black, and 257 people in District 27, 21 percent of them black. (Def. Ex. 475). Second, in Auburn precinct, the drafters gave District 28 two irregularly shaped protrusions that captured mostly majority-black census blocks, although several majority-black blocks were put in District 27 that could have been put in District 28.
Auburn Precinct in Act 603
*1151[[Image here]]
(APSX 332). The split put 3,644 people in District 28, 71 percent of whom were black, and 48,831 people in District 27, 13 percent of whom were black. (Def. Ex. 475). Third, in Beuaregard School precinct, the drafters put majority-white population in District 28, which built a bridge to Auburn precinct. (APSX 333; Def. Ex. 475). Fourth, in Marvyn precinct, the drafters drew an irregular shape that put all of the majority-black blocks in District 28.
Marvyn Precinct in Act 603
*1152[[Image here]]
(APSX 335). The split placed 419 people in District 28, of whom 57 percent were black, and 244 people in District 27, of whom 20 percent were black. (Def. Ex. 475). Fifth, the split of Loachapoka precinct is irregular and put almost all of the majority-black blocks in District 28, while also forming a bridge to the split in Wa-verly precinct.
Loachapoka Precinct in Act 603
*1153[[Image here]]
(APSX 334). It assigned 1,809 people to District 28, 81 percent of them black, and 1,517 people to District 27, 16 percent of them black. (Def. Ex. 475). On balance, the precincts split in Lee County are evidence that race predominated.
Finally, the drafters split five precincts in Russell County in a way that provides some evidence that race predominated. First, in Roy Martin Center precinct, District 28 picked up significant majority-black population and then continued to reach even more significant majority-white population. (Def. Ex. 475). This split is not evidence that race predominated.
Roy Martin Center Precinct in Act 603
*1154[[Image here]]
(APSX 364). The split put 2,900 people in District 28, only 38 percent of them black. (Def. Ex. 475). Second, in Austin Sumbry Park Voting District, the drafters put all of the majority-black blocks in District 28, but added two unnecessary majority-white blocks that make the overall addition majority-white. (Def. Ex. 475).
Austin Sumbry Park Voting District in Act 603
*1155[[Image here]]
*1156(APSX 361). This split is not evidence that race predominated. Third, in Ladonia Fire Dept precinct, District 28 forms an odd protrusion containing no majority-white blocks.
Ladonia Fire Dept Precinct in Act 603
[[Image here]]
(APSX 362). But the split places only 71 people in District 28, so we give this split less weight. (Def. Ex. 475). In National Guard Armory Voting District precinct, the drafters followed smooth lines that take in more majority-white blocks than are necessary to reach the majority-black blocks.
National Guard Armory Voting District Precinct in Act 603
*1157[[Image here]]
*1158(APSX 363). And in Seale Courthouse Voting District precinct, the drafters put 1,299 people in District 28, 43 percent of them black, and 939 people in District 27, 16 percent of them black. (Def. Ex. 475). We cannot identify any racial pattern.
Seale Courthouse Voting District Precinct in Act 603
[[Image here]]
(APSX 365). On the basis of the suspicious precinct splits and the eight-point increase in black population in this underpopulated district, we find that race predominated in the design of Senate District 28.
We further conclude that Senate District 28 does not satisfy strict scrutiny. Alabama does not explain why section 2 or section 5 required raising the voting-age black population percentage over eight points, from 49.82 in the benchmark to 58.03 in Act 603. It also does not .explain why section 2 or section 5 required converting a plurality-black district into a majority-black district. The comments by Sanders and Jackson are not relevant in this district because it falls short of a 62 percent black population. And we find it persuasive that Plan A has a majority-black District 28 without the problematic protrusion in Lee County and with a less troubling protrusion in Houston County. We must enjoin the use of Senate District 28 in future elections.
f. Senate District 33 (Mobile)
We find that race did not predominate over traditional districting criteria in the design.of Senate District 33. The plaintiffs argue that race predominated because the drafters drew the boundaries to meet a target black population percentage, but this district exceeded the previous percentage by 6.82 percentage points, increasing from 64.89 to 71.64 percent. (Doc. 30-41 at 2; Doc. 30-39 at 3). Even if exceeding the target constitutes meeting the target, the plaintiffs have not demonstrated how the *1159drafters subordinated traditional district-ing principles to race.
Senate District 33 was underpopulated by 18.05 percent and needed to gain significant population, but it was limited in how it could grow. The district could not cross Mobile Bay into Baldwin County, as neither the Mobile County delegation nor the Baldwin County delegation wanted Mobile Bay to be split. (Doc. 215 at 41-44). And there was a complicated shift among the southern districts as the drafters attempted to satisfy incumbents and avoid crossing Mobile Bay. District 35 also needed to gain population, so that district took population from District 34. This decision pushed District 34 north and east, which forced District 22 south and east, to take population from overpopulated District 32. And the incumbent in District 35 lived near the southwest edge of District 33, limiting growth in that direction. (Def. Supp. Ex. 5). The Democratic Conference plaintiffs’ own expert on remand, Fairfax, described the situation as “somewhat of a landlock.” (Doc. 296-7 at 108). Although changing little in shape and maintaining a core around the city of Prichard and the north part of the city of Mobile, the district grew to the south while remaining wholly in Mobile County. The district now has cleaner lines than it had under the 2001 plans, and it maintained the same core.
[[Image here]]
*1160(Ala. Reapportionment Office, supra).
Instead of moving south, the Democratic Conference plaintiffs moved north.
Senate District 33 in Democratic Conference Plan A and Act 603
[[Image here]]
(Doc. 287-23 at 2). Fairfax conceded that, on an eyeball test, the enacted District 33 was probably more compact than his own. (Doc. 296-7 at 110). The Black Caucus 1% Plan also expanded District 33 to the north.
Senate District 33 in Black Caucus 1% Plan and Act 603
[[Image here]]
*1161(APSX 541). We cannot say that this district is significantly more regular or compact.
The plaintiffs argue that race is the only explanation for the shape of District 33, but their argument fails. Hinaman testified about the constraints on District 33, and the decision to move the district south was not dictated by race. (Doc. 217 at 131). The incumbent in District 34 limited expansion to the northwest, the incumbent in District 35 limited expansion to the west, and moving District 33 north could interfere with Senate District 22. (Id.) Based on these race-neutral constraints, Hinaman explained that “going south was essentially the easiest course.” (Id.) Moreover, the shape of the district is contiguous and compact. It also preserves an urban core— 82 percent of the population of the enacted district lives in the urbanized area of Mobile. (Def. Supp. Ex. 6).
The Democratic Conference plaintiffs initially argued that the district expanded southward into large concentrations of black population, instead of westward where more of the population was white, (Doc. 258 at 28), but they supply no race-neutral reason to move westward. Moreover, their own plan moved north, not west, leaving us unable to say that moving west was required by traditional district-ing criteria.
The plaintiffs criticize the splitting of precincts, but three of the split precincts are not suspicious. The drafters split Satsuma City Hall precinct in a way that smoothed out the line and put no people in District 33.
Satsuma City Hall Precinct in Act 603
*1162[[Image here]]
(APSX 341). In Mormngside Elementary School precinct, the drafters divided the precinct along Interstates 65 and 10.
Street Map and Population Map for Morningside Elementary School Precinct
[[Image here]]
(U.S. Census Bureau, TigerWEB, https:// tigerweb.geo.census.gov/tigerweb/; APSX 339). The split leaves behind majority-black blocks in the west and takes in a *1163majority-white block of 315 people in the east. And in Chickasaw Auditorium precinct, the drafters used logical straight lines that improve the shape of an oddly shaped precinct. The line follows zero-population blocks, it does not weave around looking for majority-black blocks or avoiding majority-white blocks, and it keeps areas of dense population together on each side of the line.
Chickasaw Auditorium Precinct in Act 603
[[Image here]]
(APSX 338). The split put 3,689 people in District 33, 43 percent of whom were black, and 2,417 people in District 34, 19 percent of whom were black. (Def. Ex. 475). We find that this split is not the result of racial predominance.
Two other split precincts are slightly suspicious, but they do not assign a significant number of people on the basis of race. In Riverside Church of the Nazarene precinct, the drafters put all of the majority-black blocks in District 33. The split assigned 1,238 people to District 33, 57 percent of whom were black, and 466 people to majority-white District 35, 8 percent of whom were black. (Id.). But the drafters used smooth lines, kept the banks of the river together, and included several majority-white blocks that were unnecessary to reach majority-black blocks.
Riverside Church of the Nazarene Precinct in Act 603
*1164[[Image here]]
(APSX 340). In St. Andrews Episcopal Church precinct, the drafters drew a more irregular line that put 3,061 people in District 33, 49 percent of whom were black, and 427 people in District 35, 5 percent of whom were black. (Def. Ex. 475).
St. Andrews Episcopal Church Precinct in Act 603
[[Image here]]
*1165(APSX 342).
Overall, we find that race did not predominate in the design of District 33. The district maintains a coherent urban core in Mobile. The shape of the district is more compact and regular. And the splits of Riverside Nazarene Church and St. Andrews Episcopal Church precincts, although suspicious, do not affect a significant part of the district.
g. House Districts 19 and 53 (Madison County)
We find that race did not predominate over traditional districting criteria in the design of House District 19, but we find on the basis of precinct splits that race predominated in the design of District 53. In the 2001 plan, District 19 was in Madison County, and District 53 was in Jefferson County. Both were underpopulated — District 19 by 6.9 percent and District 53 by 22.28 percent. (Def. Ex. 406). On Hinaman’s recommendation, the legislature moved District 53 from Jefferson County to Madison County so that its former population could repopulate the other majority-black districts in Birmingham, all of which were severely underpopulated. (Doc. 217 at 132-33). Hinaman chose to move District 53 in particular because he was told that the incumbent was retiring, (Doc. 134-4 at 132), and the incumbent has since died, (Doc. 203 at 52). House District 53 moved to the Huntsville area, where the black population had grown enough to justify drawing another majority-black district. It took part of the former District 19, and both districts, though centered on Huntsville, had to expand outside of the city to satisfy the ±1% population deviation.
The dissent and the Democratic Conference plaintiffs argue that moving District 53 was race-based because Hinaman moved the district to avoid retrogressing the Birmingham districts, (Doc. 258 at 47), but the Democratic Conference’s own expert disagreed when it came time to draw an alternative district. In Plan A, the Democratic Conference moved District 53 to Madison County, and Fairfax defended the move as race-neutral:
Q. What did you do with District 53?
A. That district was removed to the northern end, Madison County, and the reason for that, I know that was done in the 2012 State Plan. And from my vantage point, it seemed logical because of the lack of population there. Once I add it up, it was something like 70,000, and lack of population amongst all the districts [in Jefferson County] and that’s sort of where the city of Birmingham area—
Q. So moving a district from Jefferson County [that] lacked population to Madison County that had lots of population, lots of growth, you could go along with that.
A. That seemed logical to me.
(Doc. 296-7 at 152). Fairfax’s testimony about “logical” choices establishes that traditional districting criteria were not subordinated to race.
Nor, contrary to the dissent, was the relocation of District 53 the result of a “self-inflicted challenge” to achieve equal population. (Dissent at 1380). The Democratic Conference plaintiffs’ expert did not defend the relocation of District 53 on the ground of equal population. Instead, he defended it as a race-neutral choice in response to- population changes in Jefferson and Madison Counties. At bottom, this argument is merely another version of the plaintiffs’ and dissent’s same, failed argu*1166ment that because the drafters considered race, it must have predominated. This argument is further undermined by the plaintiffs’ decision to do the same thing in their alternative plans. We find that the relocation of District 53 in itself provides no evidence that race predominated.
The districts are not so unusually shaped as to suggest that race predominated. District 19 is less compact than it was under the 2001 plan, but it is not facially bizarre and has kept its core. District 58 is relatively compact and its borders are not facially bizarre:
[[Image here]]
(Ala. Reapportionment Office, supra). The district crosses no county lines, and it is coherently urban — about two-thirds of District 19 and over 99 percent of District 53 live in the city of Huntsville. (Def. Supp. Ex. 3).
Nor are the plaintiffs’ alternative plans more regular and compact. The Democratic Conference plaintiffs traded the western half of District 19 for a southern protrusion:
House District 19 in Conference Plan A and Act 602
[[Image here]]
*1167(Doc. 287-15 at 4). In District 53, they swapped precincts in the south and east for precincts in the north and west:
House District 53 in Conference Plan A and Act 602
[[Image here]]
(Id.). They failed to explain these choices except on the ground of precinct splits, a theory that we reject later.
The 1% Plan is not good evidence about whether race predominated in Act 602 because it does not respect the principle of incumbent protection. The Black Caucus plaintiffs kept District 53 in Jefferson County but drew a new majority-black House District 6 in Madison County. It is located roughly where District 19 is, and the incumbent in that district represented District 19. In the following map, House District 6 in the 1% Plan is yellow, House District 19 in Act 602 is marked by the number 19 in a purple circle and enclosed in purple lines, and the House District 19 incumbent is marked by a blue star and the label “Hall”:
House District 6 in Black Caucus 1% Plan and House District 19 in Act 602
*1168[[Image here]]
(APSX 506). The 1% Plan m turn put a Republican incumbent in District 19, located roughly where District 53 is in the enacted plan and Plan A, but we can expect that this district will lean Democratic based on the plaintiffs’ evidence at trial about voting patterns. (See Doc. 203 at 77-81). Avoiding partisan mismatch is a legitimate explanation for the choice made by the legislature. As the Democratic Conference plaintiffs mapmaker explained, incumbent protection” is a legitimate dis-tricting principle that includes not simply avoiding conflicts, but also avoiding “putfting] the incumbents in a different area.” (Doc. 296-7 at 46-48).
House District 19 in Black Caucus 1% Plan and House District 53 in Act 602
[[Image here]]
(APSX 505).
The plaintiffs also have failed to produce evidence that the drafters pursued a target in either district. The black population percentage in House District 19 decreased from 69.82 to 61.25 percent, which means the drafters came up 8.57 points short of *1169any target. The plaintiffs argue that the reason for the drop in percentage was that the drafters could not find any more black population, but they present no evidence that the drafters tried to increase the black population percentage of House District 19. District 19 does not cross county lines, and the precinct splits made little difference in a district that badly missed its target.
Moreover, except for their final plans on remand, the plaintiffs produced several versions of District 19 with comparable black population percentages. Three of them, the McClammy, Reed-Buskey, and Knight Plans, were closer to the supposed target than the enacted plan was:
2010 Census Total Black Population Percentages in District 19 or Equivalent Under Various Plans
[[Image here]]
(*District 6 percentage because of Black Caucus renumbering)
Because some of the plaintiffs’ plans have similar — and even higher — black population percentages, and because of the degree to which the drafters missed any target, this argument provides no support for a claim of racial gerrymandering.
The black population percentage in House District 53 remained roughly the same, increasing from 55.70 to 55.83 percent. Throughout this litigation, the plaintiffs have treated almost meeting, meeting, or exceeding a supposed target as evidence of racial predominance. The 1% Plan has a District 53 that exceeds the supposed target, and both the 1% Plan and Plan A have a second district in Jefferson County with a black population percentage within three points of the supposed target:
2010 Census Total Black Population Percentages in District 53 or Equivalent Under Various Plans
[[Image here]]
(*Located in Madison County) (** Located in Jefferson County)
In the light of these alternative plans, we find that the black population percentage in District 53 is not suspicious.
Nearly all of the precinct splits in District 19 are unsuspicious. In Chase Valley United Meth precinct, the drafters actually put a higher percentage of black population in majority-white District 21 (86 percent) than in District 19 (36 percent). (Def. Ex. 405 at 488, 491). In Harvest Bapt Church precinct, the drafters used an irregular line, (APSX 152), but they put roughly equal black population percentages in District 19 (36 percent) and District 6 (31 percent). (Def. Ex. 405 at 458, 488). In Pineview Baptist Church precinct, the drafters again drew an irregular line between Districts 6 and 19, (APSX 158), but they put roughly similar black population percentages in both districts. The split put 6,041 people in District 19, 33 percent *1170of them black, and 3,715 people in District 6, 22 percent of them black. (Def. Ex. 405 at 457, 487). In Meridianville 1st Baptist Church precinct, the drafters split the precinct along the limits of Meridianville.
Meridianville 1st Baptist Church Precinct in Act 602
[[Image here]]
(APSX 156).
Close-Up of Census Bureau Map of House District 19 in Act 602
[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 19 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01019/DC10SLDL01019_ 001.pdf). In Chapman Middle School precinct, the drafters drew a smooth line along unpopulated blocks that improved the shape of District 19.
Chapman Middle School Precinct in Act 602
*1171[[Image here]]
*1172(APSX 144). Ed White Middle School precinct consists of two noncontiguous areas, one of which the drafters put in District 19 and one of which the drafters put in District 53, improving the shape of both districts.
Ed White Middle School Precinct in Act 602
[[Image here]]
*1173(APSX 149). Similarly, in Mad Co Teacher Resource Center precinct, the drafters put one part of a noncontiguous precinct in District 19 and another in District 21.
Mad Co Teacher Resource Center Precinct in Act 602
[[Image here]]
*1174(APSX 155). In Grace United Meth Church precinct, the drafters drew a line between Districts 6 and 19 along a line of zero-population blocks.
Close-Up of Grace United Meth Church Precinct in Act 602
[[Image here]]
(APSX 151). In Sherwood Baptist Church precinct, the drafters split District 19 from District 6 along zero-population blocks for most of the boundary and then several sparsely populated blocks for the rest, a decision that we do not find suspicious.
Sherwood Baptist Church Precinct in Act 602
*1175[[Image here]]
(APSX 161). In Blackburn Chapel CP Church precinct, the drafters drew an irregularly shaped three-way split, (APSX 143), but the result was to place nearly identical percentages of black population in Districts 19 (39 percent) and Districts 6 and 53 (40 percent). (Def. Ex. 405 at 458, 488). Finally, the drafters split Highlands School precinct, (APSX 153), Lewis Chapel CP Church precinct, (APSX 154), St. Luke Missionary Baptist Church precinct, (APSX 162), with another majority-black block with no racial pattern.
The only slightly suspicious split does not prove that race predominated. In Church of Christ Meridianville precinct, the drafters used an irregular finger to reach population between Districts 6 and 21.
Church of Christ Meridianville Precinct in Act 602
*1176[[Image here]]
(APSX 147). This split put 110 people in District 19, 65 percent of them black, and 4,121 people in Districts 6 and 21, 77 percent of them black. (Def. Ex. 405 at 458, 491). But the black people placed in District 19 from this precinct account for less than a quarter of a percent of the total population of the district. If we remove the precinct from District 19, the black population percentage remains essentially unchanged at 61.2 percent. We find that race did not predominate in the design of District 19.
Most of the precinct splits in District 53 also are not suspicious. As discussed already, the split of Ed White Middle School precinct put one part of this noncontiguous precinct in District 53 and the other in District 19. (APSX 149). In Phillips CME Church precinct, the drafters improved the ■ shape of the district without splitting any population.
Phillips CME Church Precinct in Act 602
*1177[[Image here]]
(APSX 157). In Charles Stone Agricultural Center precinct, the drafters again improved the shape of the precinct by putting a block of two people in majority-white District 21.
Charles Stone Agricultural Center Precinct in Act 602
*1178[[Image here]]
(APSX 145). In Fire and Rescue Academy precinct, the drafters once again improved the shape of the district by extending the northern border of Charles Stone Agricultural Center precinct into Fire and Rescue Academy precinct until it intersected straight line of unpopulated blocks,
Fire and Rescue Academy Precinct in Act 602
*1179[[Image here]]
(APSX 150). The drafters divided West-lawn Middle School precinct along a straight line composed of unpopulated and sparsely populated blocks.
Westlawn Middle School Precinct in Act 602
[[Image here]]
*1180(APSX 164). In Blackburn Chapel CP Church precinct, the drafters used a smooth line of zero population blocks to improve the shape of District 53.
Blackburn Chapel CP Church Precinct in Act 602
[[Image here]]
(APSX 143). And in Highlands School precinct, Lewis Chapel CP Church, and St. Luke Missionary Baptist Church precincts, the drafters drew splits with another majority-black district. (APSX 153, 154, 162). The drafters fail to explain how these splits with another majority-black district prove that race predominated.
But five precinct splits are suspicious, and they convince us that race predominated in the design of District 53. In Eastside Community Center precinct, the drafters put one populated block of 154 people in District 53, 55 percent of whom were black, and 501 people in District 21, only 13 percent of whom were black. (Def. Ex. 405 at 490, 550).
Eastside Community Center Precinct in Act 602
*1181[[Image here]]
*1182(APSX 148). In Airport Road Fire Station #6 precinct, the drafters followed smooth lines of zero population blocks except for an irregular incursion of 21 people in the northwest corner of the precinct.
Airport Road Fire Station # 6 Precinct in Act 602
[[Image here]]
(APSX 142). In University Place School precinct, the drafters used an irregular line to put 4,201 people in District 53, 42 percent of them black, and 1,731 people in majority-white District 6, 13 percent of them black. (Def. Ex. 405 at 457, 550).
University Place School Precinct in Act 602
*1183[[Image here]]
(APSX 163). In Senior Center precinct, the drafters put an irregular area of 316 peo-pie in majority-white District 10, 5 percent of whom were black, and 3,111 people in District 53, 33 percent of whom were black. (Def. Ex. 405 at 467, 551).
Senior Center Precinct in Act 602
*1184[[Image here]]
(APSX 160). In Ridgecrest School precinct, the drafters appear to have snaked around looking for black people. They put 2,853 people in District 53, 38 percent of them black, and 2,323 people in District 10, 12 percent of them black. (Def. Ex. 405 at 467, 551).
Close-Up of Ridgecrest School Precinct in Act 602
*1185[[Image here]]
(APSX 159). On the basis of these precinct splits, we find that race predominated in the design of House District 53.
We further conclude that District 53 does not survive strict scrutiny because Alabama makes no arguments that apply to District 53. The comments of Senator Sanders and Representative Jackson do not provide a strong basis in evidence in this district. District 53 is located in a different part of the state, with which we cannot presume Sanders and Jackson would be familiar. And District 53 is more urban than their districts — over 99 percent of the district lives in Huntsville. (Def. Supp. Ex. 3). We must enjoin the use of House District 53 in future elections.
h. House District 32
We find that race predominated over traditional districting criteria in the design of House District 32. The district was underpopulated by nearly 15 percent, (Def. Ex. 406), but it maintained its general orientation and location in Talladega and Calhoun Counties. The drafters added a southern protrusion, an eastern arm, and modifications along the rest of the borders.
*1186[[Image here]]
(Ala. Reapportionment Office, supra).
Some of the irregularity is explained by avoiding incumbent conflicts, a legitimate districting principle. The incumbent in District 32 lived in the “head” of the northeastern part of the district as drawn in 2001. The incumbents in District 35 and 36 also lived close to the boundary with District 32, constraining its growth to the east or the north. (Def. Supp. Ex. 2). Overall, we cannot say that the shape of this district is so facially bizarre that it leads us to infer that race predominated.
The Democratic Conference plaintiffs’ map looks slightly better:
House District 32 in Democratic Conference Plan A and Act 602
[[Image here]]
(Doc. 287-18 at 1). It avoids some of the precinct splits, removes the eastern arm and southern protrusion, and keeps the same orientation and location in Calhoun and Talladega Counties.
By contrast, District 32 is noticeably worse in the Black Caucus 1% Plan. Like the enacted district, it is located in Tal-ladega and Calhoun Counties and contains its own southern protrusion and eastern arm. Unlike the enacted district, it is shaped somewhat like a tilted number 3. The following map marks it with a tan color and marks the enacted district with a pimple line and the number 32 in a purple circle:
House District 32 in Black Caucus 1% Plan and Act 602
*1187[[Image here]]
(APSX 507). This alternative district does not convince us that race predominated in the design of District 32.
Instead, the black population percentage and the precinct splits persuade us that race predominated. The black population percentage of the district barely increased from 59.34 to 60.0 percent, (Doc. 263-2), but most of the alternative plans had lower black population percentages.
2010 Census Black Population Percentage in District 32 Under Various Plans
[[Image here]]
When the state drew a hypothetical district for this litigation with no split precincts, the black population percentage was 6.6 points lower, taking the district from slightly above the supposed target to well below it.
The lines of the precinct splits are also suspicious. In the eastern arm, the drafters extended District 32 to the eastern edge of Eastaboga Comm Center-Old Lincoln High Gym precinct, picking up all of the majority-black blocks along the way and forming a bridge to Old Mumford High, where the drafters drew an irregular shape that picked up a cluster of majority-black blocks.
Eastaboga Comm Center-Old Lincoln High Gym and Old Mumford High Precincts in Act 603
*1188[[Image here]]
(APSX 260). The drafters also formed two indentations in District 33 to pick up clusters of majority-black blocks. The racial pattern gives rise to an inference that race predominated. The 2001 district split the same precinct, but it did not push as far east. Moreover, the 2001 district did not include any part of Old Mumford High precinct. (Doc. 287-30 at 1). And the statistics about these splits are suspicious. The drafters put 3,023 people from Eastaboga Comm Center-Old Lincoln High Gym precinct in District 32, 34 percent of whom were black. (Def. Ex. 405 at 512). To the west they put 3,678 people in District 33, 18 percent of whom were black, (id. at 515), and to the east they put 456 people in District 35, 5 percent of whom were black, (id. at 520). In Old Mumford High precinct, the drafters put 552 people in District 32, 74 percent of whom were black, and 2,873 people in District 35, 11 percent of whom were black. (Id. at 512, 520). Alabama offers no race-neutral explanation for these precinct splits.
In the southern protrusion of the district, the drafters split Limbaugh Community Center-Bon Air-Oak Grove precinct in a racial fashion. The drafters formed an irregular line through District 33, capturing a large cluster of majority-black blocks and then stopping.
Limbaugh Comm Center-Bon Air-Oak Grove Precinct in Act 602
*1189[[Image here]]
(APSX 261). The split put 1,908 people in District 32, of whom 62 percent were black, and 9,268 people in District 33, of whom 24 percent were black. (Def. Ex. 405 at 512, 516). There is no race-neutral explanation for this precinct split.
Several other precinct splits are also suspicious, and one of them, Anniston, accounts'for over a third of the population of District 32. In the north of that precinct, majority-white District 36 pushed across the line to capture large majority-white blocks, although District 32 absorbed several majority-white blocks adjacent to District 36. In the west, the legislature split the precinct along a line that zigs to put majority-white blocks in District 36 and zags to put majority-black blocks in District 32. Further south, District 36 crosses the precinct line to capture mostly majority-white'blocks and leaves mostly majority-black blocks in District 32, although District 36 takes several majority-black blocks and District 32 takes some majority-white blocks.
Anniston Precinct in Act 602
*1190[[Image here]]
(APSX 61). This split put 17,705 people in District 32, 66 percent of whom were black; 1,973 people in District 36, 11 percent of whom were black; and 426 people in District 40, 5 percent of whom were black. (Def. Ex. 405 at 511, 520, 529). The precinct was also split in 2001 along different irregular lines, (Def. Ex. 413 at 866), but the split changed in 2012, so we cannot say that the drafters preserved existing lines.
In 1st Presby/Mental Health/Golden Springs/Donoho precinct, the drafters drew a winding line that captured all of the populous majority-black blocks and very few populous majority-white blocks.
1st Presby/Mental Health/Golden Springs/Donoho Precinct in Act 602
*1191[[Image here]]
(APSX 60). The split put 3,238 people in District 32, 44 percent of whom were black, and 7,587 people in District 36, 14 percent of whom were black. (Ex. DX 405 at 511, 522). The percentages and the notable, although not perfect, racial sorting are evidence that race predominated.
Renfroe Fire Hall-Stemley Fire Hall, which is divided between Districts 32 and 33, provides more evidence that race predominated. District 33 takes the riverbank, which contains almost exclusively majority-white blocks, and District 32 takes all of the populous majority-black blocks. The line between the districts zigs and zags to cause that result. ■
Renfroe Fire Hall-Stemley Fire Hall Precinct in Act 602
*1192[[Image here]]
(APSX 264). District 32 took 1,966 people from the precinct, of whom 49 percent were black, and District 38 took 2,721 people from the precinct, of whom 8 percent were black. (Def. Ex. 405 at 512, 516). A “Renfroe Fire Hall” precinct was split in 2001, but we cannot tell whether this is the same precinct as “Renfroe Fire Hall-Stemley Fire Hall,” (see DX 413 at 866), so we draw no conclusions from this fact. The split of Renfroe Fire Hall-Stemley Fire Hall precinct is evidence that race predominated in the design of the district.
We note that some of the precinct splits might be explained by traditional factors. According to Fairfax, the following precincts were all split in the 2001 plan: Eula-ton/Bynum/West Park Heights Bapt., 1st Presby/Mental Health/Golden Springs/Do-noho, Eastaboga Comm Center-Old Lincoln High Gym, Talladega National Guard Armory-Spring St Comm Center-Berni-ston, Mabra-Kingston Bapt-Talla Co Central High, Winterboro Vol Fire-Gable’s Corner Yol Fire, Limbaugh Comm Center-Bon Air-Oak Grove, and Renfroe Fire Hall-Stemley Fire Hall. (Doc. 296-7 at 181). Alabama argues on remand that the Democratic Conference plan splits most of the same precincts in similar ways, (Doc. 295 at 32), but the Act 602 splits are not along exactly the same lines as in 2001, and the Plan A splits are not along exactly the same lines as in Act 602. We find that race predominated in the design of House District 32.
We further conclude that District 32 does not survive strict scrutiny. Alabama makes no district-specific arguments. The comments of Sanders and Jackson do not provide a strong basis in evidence in this district because the district has a black population percentage of 57.78 percent, shy of the 62 percent minimum advocated by Senator Sanders and Representative Jackson. We have no credible evidence about what percentage is necessary under the Voting Rights Act, and Alabama has the burden of proof on strict scrutiny. We must enjoin the use of House District 32 in future elections.
i. House Districts 52, 54, 55, 56, 57, 58, 59, and 60 (Jefferson County)
We find that race did not predominate over traditional districting criteria in the *1193design of House Districts 52, 55, 56, 57, 58, 59, or 60, but we find that race predominated in the design of District 54. Most of these districts were severely underpopulated, as was District 53, which the legislature moved from Jefferson County to Mad-json County in 2012-
[[Image here]]
Hinaman was concerned that there was no way to draw nine majority-black districts in the Birmingham area while maintaining the ±1% deviation and avoiding dramatic retrogression in all of the districts. (Doc. 134-4 at 60). To repopulate the districts, Hinaman suggested moving District 53 from Jefferson County to Madison County so that its former residents could repopulate the majority-black districts in Birmingham, all of which were severely underpopulated. (Doc. 217 at 132-38). Hi-naman chose to move District 53 because he was told that the incumbent was retiring, (id. at 132), and the incumbent has since died, (Doc. 203 at 52).
The plaintiffs argue that each of these districts is a racial gerrymander because the drafters “cannibalized” District 53 to repopulate the remainder of the majority-black districts in Birmingham. (Doc. 258 at 48). The dissent also argues that the decision of the drafters to move District 53 “should be given significant weight in the predominance analysis of its surrounding districts.” (Dissent at 1379). We disagree, for the reasons articulated by the Democratic Conference plaintiffs’ own expert in defense of Plan A, which also moved District 53 to Madison County:
Q. What did you do with District 53?
A. That district was removed to the northern end, Madison County, and the reason for that, I know that was done in the 2012 State Plan. And from my vantage point, it seemed logical because of the lack of population there. Once I add it up, it was something like 70,000, and lack of population amongst all the districts there and that’s sort of where the city of Birmingham area—
Q. So moving a district from Jefferson County [that] lacked population to Madison County that had lots of population, lots of growth, you could go along with that.
A. That seemed logical to me.
Q. Okay. Okay. So—
A. In this particular case.
(Doc. 296-7 at 152-53). We also observe that the Democratic Conference plaintiffs made the same choice in Plan A, and they assert that race did not predominate in their plan.
The plaintiffs fail to prove that the legislature could have maintained nine majori*1194ty-black districts in Birmingham while complying with the Voting Rights Act. Plan A has three majority-black House districts in Jefferson County with black voting-age populations under 55 percent and one under 51 percent. The 1% Plan has three majority-black House districts in Jefferson County under 55 percent and one under 52 percent. (Doc. 296-2). Neither group of plaintiffs have established that these districts provide an equal opportunity to participate in the political process (for purposes of section 2) or avoid retrogression (for purposes of section 5), and the earlier plans do not comply with the ± 1% population deviation. In sum, none of the alternative plans prove that the legislature could have kept nine majority-black House districts in Jefferson County.
Finally, the evidence about the decision to move District 53 to Madison County establishes that the drafters were concerned about traditional districting principles. They moved District 53 because the incumbent in District 53 was retiring. (Doc. 134-4 at 132). This choice minimized the effect of the move on the incumbent legislators. Even if the choice was political — the incumbent was a Democrat — it is neither surprising nor racially motivated that a Republican-controlled legislature would choose to inconvenience a Democratic legislator. Moreover, that the “core” of District 53 could not be preserved points to the difficulty of districting, not racial gerrymandering. Almost no traditional criterion can be maintained without exception in a statewide plan. That the drafters on one occasion had to move a single district to respond to massive underpopulation is not evidence of an overriding racial purpose. The only other district that the drafters moved was House District 73. House District 73 was not a majority-black district, but it was a Democratic district, and it was also moved to respond to severe underpopulation of surrounding districts. If the drafters had not moved District 53, the resulting Birmingham districts might have had to change dramatically. It is consistent with traditional districting criteria to sacrifice the core of one district to maintain the least amount of change among the remaining districts. At the very least, the plaintiffs fail to prove otherwise.
The plaintiffs next argue that race predominated over traditional criteria in each of the Birmingham districts because six of the eight districts maintained black population percentages similar to their previous percentages. Only if race predominated over ordinary criteria, they argue, could the drafters achieve such precision in black population percentages. The plaintiffs point to numerous split precincts as evidence of the subordination of traditional districting principles. A close analysis of each district reveals that the plaintiffs have failed to prove their case with respect to all but one of the districts.
House District 52, which kept a nearly unchanged black population percentage of 60.13 percent, ceded its southwestern corner to District 56, also a majority-black district.
*1195[[Image here]]
(Ala. Reapportionment Office, supra). The enacted district is not noticeably odder than the old district.
The black population percentage is not suspicious either. All of the alternative plans proposed black population percentages within 4.5 points of the enacted plan, and two plans met and exceeded the supposed target.
2010 Census Black Total Population Percentages in House District 52 Under Various Plans
[[Image here]]
The Democratic Conference plaintiffs insist that race did not predominate in the design of their alternative district, and we find it to be evidence that a black population percentage in the low 60s is not evidence that race predominated.
Our dissenting colleague would disregard the alternative plans because the enacted district was closer to the alleged target than the alternative plans. (Dissent at 1383). But our colleague premises this argument on his test for racial predominance, which infers that race predominated if the legislature came close to its alleged target. The test of the Supreme Court, by contrast, demands that we presume that the legislature acted in good faith until that presumption is overcome. Miller, 515 U.S. at 915, 115 S.Ct. 2475. That the enacted plan and the alternative plans drafted a district with similar black percentages suggests to us that the racial percentage is not enough to prove racial predominance. As the dissent acknowledges “demographic realities in Jefferson County” required drafting a district with a similar black percentage to the 2001 plan. (Dissent at 1383). These demographic realities also justify the decision of the drafters to move 19,284 people in and out of the district.
The plaintiffs drew the districts differently than the legislature did, but we find that they are no more regular or compact than the enacted district. As the following maps illustrate, both groups of plaintiffs drew District 52 with more of a north-south orientation:
House District 52 in Democratic Conference Plan A and Act 602
*1196[[Image here]]
(Doc. 287-14 at 9).
House District 52 in Black Caucus 1% Plan and Act 602
[[Image here]]
(Ex. APSX 508 (Black Caucus district shaded in blue)). The plaintiffs fail to explain why these particular boundaries were required by traditional districting criteria other than precinct splits, and we find that they are no more regular or compact than the enacted districts.
The split precincts prove nothing with respect to race. District 52 split only two precincts with a majority-white district. From Shades Cahaba Elementary School, the legislature placed 1,198 people in District 52 and 2,741 people in District 46. (Def. Ex. 405 at 91,101). Six percent of the people placed in District 52 were black, *1197and so were 4 percent of the people placed in District 46. (Id. at 91, 100). From Birmingham Botanical Garden, the legislature placed 389 people in District 52, 1 percent of them black, and 609 people in District 46, 1 percent of them black. (Def. Ex. 405 at 91, 101). We .find no evidence of racial predominance in these splits.
District 52 also splits five precincts with other majority-black districts, and the Black Caucus plaintiffs argue that “the lack of any clear pattern” in these splits “only shows how black and white populations were being shaved between majority-black districts to hit their arbitrary target percentages.” (Doc. 256 at 47-48). They do not provide evidence that these splits were the result of the subordination of traditional districting criteria, and it is not apparent how the lack of any clear pattern could be proof of racial gerrymandering.
Our dissenting colleague disagrees. The precinct splits, according to the dissent, “are notable not because the State used them to pack black people into existing majority-black districts ... but because, without them, the State would have been too far afield of its racial target.” (Dissent at 1383). But race is not the primary explanation for the black population percentage, meaning the precinct splits cannot “corro-borat[e]” this rebuttable presumption of racial predominance. (Id. at 1370). Our response is becoming a refrain. The existence of an alleged racial target is not per se evidence of racial predominance. Because the alternative plans proposed districts with similar black populations and the precinct splits are explainable on grounds other than race, we find that race did not predominate in the design of House District 52.
We find that race predominated in the design of House District 54. The district is irregularly shaped, hits its alleged target, and contains a suspicious precinct split that appears to make the difference between hitting and missing the target. At the same time, the district was drawn in part by the incumbent of this and neighboring districts, the plaintiffs also drew irregularly shaped districts, and all of the alternative plans with a majority-black District 54 also hit the target. We find that the plaintiffs satisfied their burden of proof.
District 54, which maintained a nearly unchanged black population percentage of 56.83 percent, is possibly the most irregular of the Birmingham districts. (Def. Ex. 403, 406). The district has a thin bottleneck that connects two otherwise normally shaped halves:
[[Image here]]
(Ala. Reapportionment Office, supra).
Representative Patricia Todd, the white incumbent Democrat, helped design the district with the two black incumbent Democrats from Districts 58 and 59. (Doc. 217 at 136-37, 231-32). The incumbent in *1198District 58 proposed a “solution” that would “make everybody happy.” (Id. at 231-32). After testifying about other areas in which incumbents proposed district lines, Representative McClendon testified about the role played by the Democratic incumbents in Districts 54, 58, and 59:
And Representative [Oliver] Robinson came to me and he said that he had a solution; that he could make everybody happy. And I called [Representative Mary] Moore.... [A]nd I told her that Representative Robinson was there and that he had made this proposal and I wanted to make sure she was happy. And we all met up in the reapportionment office, and we got one of the ladies that worked up there to help us work on the lines. And Ms. Todd, Representative Todd, joined us and got involved in the process.
But once again, we had contiguous districts, and they were essentially trading folks, keeping the deviation in line and making changes that didn’t affect anybody on the outside, and I thought that was fine.
(Doc. 217 at 231-32). Hinaman testified similarly at trial:
Q. ... What involvement did you have with the plan after it was produced and in the hands of Mr. McClendon and Senator Dial?
A. I was here during the session, and when there were changes that were made to the map, I would — I didn’t do all of them, but I sat down with some of the legislators who had changes that they wanted made and remade them.
Q. Like who did you sit down with?
A. I sat down with Oliver Robinson; Patricia Todd; Mary Moore was on the phone with Oliver Robinson when I did that.
(Id. at 136). That the legislature adopted the compromise of three incumbents and no one alleges that the incumbents drew the districts on the basis of race is strong, but not dispositive, evidence that race did not predominate.
The plaintiffs drew more regular-looking districts in their alternative plans, although they moved the district significantly and rejected the incumbents’ plan. The Democratic Conference plaintiffs kept only the western part of the district and expanded it farther north.
House District 54 in Democratic Conference Plan A and Act 602
[[Image here]]
*1199(Doc. 287-14 at 10), The Black Caucus plaintiffs shifted the district to the north but kept a similar orientation. The alternative district is shaded in lavender, and the enacted district is marked by a purple line and the number 54 in a purple circle:
House District 54 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 510).
Notably, the alternative plans either draw District 54 with a black population percentage that exceeds the target by a greater margin or with no black majority at all. The 1% Plan increased the total black population percentage to 60.64 percent, and Democratic Conference Plan A increased it to 60.6 percent. (Doc. 296-1). Two of the plans that made no effort to comply with the Commission guidelines also had black population percentages that exceeded the supposed target — 58.72 percent in the Knight Plan and 61.06 percent in the New Black Caucus Plan. (Common Ex. 46; APSX 36). The other two alternative plans, McClammy and Reed-Buskey, gutted the black population to repopulate the surrounding majority-black districts. (Common Exs. 45, 42).
2010 Census Black Total Population Percentages in House District 52 Under Various Plans
*1200[[Image here]]
These alternative plans might have threatened Representative Todd’s ability to win reelection because drastically decreasing the black population percentage would probably reduce the number of reliable Democratic voters, based on the plaintiffs’ own evidence, (Doc. 203 at 78-79). It is likely that Representative Todd-helped to draw her district in a partisan fashion to avoid that outcome, and the plaintiffs fail to prove otherwise.
We nevertheless find strong evidence that race predominated in the split of Clearview Baptist Church precinct. The following map shows majority-white District 44 (in blue), majority-black District 54 (in purple), and majority-black District 58 . (in green):
Clearview Baptist Church Precinct in Act 602
*1201[[Image here]]
(APSX 104). The legislature put all of the easily accessible majority-black blocks in District 54 and no unnecessary majority-white blocks. The split put 561 people in District 54, 71 percent of them black; 4,437 people in District 44, 18 percent of them black; and 738 people in District 58, 31 percent of them black. (Def. Ex. 405). Finally, if we remove this precinct from District 54, the black population percentage drops below the alleged target.
That said, none of the other precinct splits prove that race predominated. The following map shows the split of Mountain View Baptist Church precinct, which has no discernible racial pattern:
Mountain View Baptist Church Precinct in Act 602
*1202[[Image here]]
(APSX 119). Second, the legislature split Irondale Senior Citizens Building precinct in three parts. The division between Districts 44 and 54 followed a smooth line of zero-population blocks that may be a major road, and the division between Districts 45 and 54 has no discernible racial pattern.
Irondale Senior Citizens Building Precinct in Act 602
*1203[[Image here]]
(APSX 115). There is also no discernible pattern in the split of Birmingham Fire Station # 31 precinct.
Birmingham Fire Station # 31 Precinct in Act 602
[[Image here]]
(APSX 98). The other precinct splits are between or among majority-black districts, and the plaintiffs fail to offer a theory of how those splits prove that race predominated. But on the basis of the shape and the split of Clearview Baptist Church precinct, we find that race predominated in the design of House District 54.
We conclude that District 54 does not survive strict scrutiny. As in several other *1204districts, Alabama makes no district-specific arguments. The comments of Sanders and Jackson do not provide a strong basis in evidence in District 54 because the district has a black population percentage of 56.88 percent, shy of their minimum of 62 percent minimum. Alabama has not proved that District 54 was narrowly tailored to comply with the Voting Rights Act.
We find that race did not predominate in the design of House District 55. District 55, which maintained a black population percentage of 73.55 percent, (Doc. 268-2 at 2), is an elongated district stretching from the center of Birmingham to the northwest.
[[Image here]]
(Ala. Reapportionment Office, supra). It splits no precincts with majority-white districts. (Doc. 256 at 55; Doc. 258 at 59). District 55 borders majority-black District 57 to the west, majority-black District 52 to the south, and majority-black District 60 to the east. The oddest portion of the district is the northwest border with majority-white District 16, but those lines follow precinct lines without interruption. Our dissenting colleague’s characterization of District 55 to the contrary, (Dissent at 1384), this design is no odder than the old lines.
The plaintiffs again drew the district differently than the legislature did, but they fail to explain why their choices were required by traditional districting criteria. The Democratic Conference district has little overlap with the enacted district because it moves out to rural Jefferson County:
House District 55 in Democratic Conference Plan A and Act 602
*1205[[Image here]]
(Doc. 287-16 at 11). The Black Caucus plaintiffs shifted the district, shaded green in the following map, to the west:
House District 55 in Black Caucus \% Plan and Act 602
[[Image here]]
(APSX 511). Without further explanation, these maps do not prove that race predominated in the design of District 55.
The plaintiffs argue that the drafters split precincts between House District 55 and other majority-black districts to maintain their racial targets, but all of the splits are with other majority-black districts, and the plaintiffs fail to explain how these splits are evidence that race predominated. Moreover, in its unsplitting exercise, Alabama could not find a combination *1206of whole precincts that satisfied the ±1% deviation in this district. (Doc. 263-3). A visual examination of the splits between District 55 and other majority-black districts reveals no outrageous lines or bizarre contortions; if anything, the district lines smooth out oddities in the precinct lines and connect communities of interest. (APSX 91, 92, 103, 110, 111, 116, 125, 127, 129).
The dissent makes several mathematical observations, but none of them prove that race predominated in this district. The drafters hit their alleged target nearly exactly, (Dissent at 1383-84), but the plaintiffs and the dissent fail to point to a single choice from which we can infer that the drafters subordinated traditional district-ing criteria to race. The drafters also moved 28,143 people in and out of the district to remedy an underpopulation of 9,949, (id.), but as in other districts in the Birmingham area, this kind of turnover is unsurprising in an area that was redrawn to cure severe underpopulation. We find that race did not predominate in the design of House District 55.
We also find that race did not predominate in the design of House District 56. District 56, which has almost the same black population percentage (62.14 percent) as in 2010 (62.13 percent), (Doc. 263-2 at 2), is a compact district reaching west from central Birmingham. It is noticeably more compact than it was in the previous plan:
[[Image here]]
*1207(Ala. Reapportionment Office, supra).
In their court-ordered alternative plans, both plaintiffs drew a District 56 that was similar to the enacted district:
House District 56 in Democratic Conference Plan A and Act 602
[[Image here]]
(Doc. 287-14 at 12).
District 56 m Black Caucus 1% Plan and Act 602 House
[[Image here]]
(APSX 512 (Black Caucus district shaded in yellow)).
The black population percentages in this district are similar across all of the alter*1208native plans. Those plans — with the exception of the radically different Knight Plan — proposed black population percentages within four points of the enacted plan:
2010 Census Total Black Population Percentages in House District 56 Under Various Plans
[[Image here]]
Most notably, the 1% Plan — which the Black Caucus plaintiffs aver is race-neutral — hits the target and exceeds it by only 0.91 percentage points. The alternative districts defeat any inference of racial predominance in the enacted district based solely on its black population percentage.
District 56 has four precinct splits, none of which are persuasive evidence that race predominated. The plaintiffs fail to explain how the two precincts split with other majority-black districts prove that race predominated, and no patterns are obvious to us. In the third split, the drafters divided Canaan Baptist Church precinct with District 15. (APSX 101,114).
Canaan Baptist Church in Act 602
[[Image here]]
(APSX 101). The split put 1,426 people in District 56, 12 percent of them black, and 3,652 people in District 15, 21 percent of them black. (Def. Exs. 479, 556). That is, the drafters put a lower black population percentage in the majority-black district.
*1209In the fourth split, the drafters divided Hunter Street Baptist Church precinct with majority-white Districts 15 and 46.
Hunter Street Baptist Church Precinct in Act 602
[[Image here]]
(APSX 114). This split follows 1-459 and the Louisville and Nashville Railroad:
Census Bureau Map in Vicinity of Hunter Street Baptist Church Precinct
[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 56 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01056/DC10SLDL01056_ 001.pdf). The split put 1,685 people in District 56, 20 percent of them black; 7,733 people in District 15, 10 percent of them black; and 1,705 people in District 46, 5 percent of them black. (Def. Ex. 405 at 479, 538, 556). These similar percentages in a split that followed manmade boundaries do not support a finding of racial predominance.
Although our dissenting colleague agrees “that the shape and compactness of [House District] 56 are not suspicious” and that the “boundaries apparently used to split the Hunter Street Baptist Church precinct ... demonstrate ... valid raee-*1210neutral criterion,” he would find that race predominated because the drafters came within .01 percent of their alleged target and they moved a large number of people in and out of the district to do so. (Dissent at 1385-86). Our colleague’s analysis of House District 56 — and racial predominance, generally — could have been accomplished with the answer to one question: did the drafters come close to meeting their alleged target? If the answer to that question was “yes,” then he would find that race predominated so long as some other evidence corroborates that view. This approach is not the approach of the Supreme Court. Ala. Legislative Black Caucus, 135 S.Ct. at 1267 (explaining that the use of racial targets only “provides evidence” that race predominated). The Court did not say that evidence of the application of an alleged target was conclusive evidence or even presumptive evidence of racial predominance.
The dissent also argues that race predominated because the drafters moved House District 53, which affected the black population of House District 52, which in turn allowed the drafters to hit their alleged target in House District 56. (Dissent at 1385). But this connection is attenuated at best. The drafters moved House District 53 because the Birmingham districts were underpopulated. And it is unclear how that move proves that race predominated in House District 56, which as the dissent acknowledges was not contiguous to former House District 53. (Id. at 1385). We find that race did not predominate in the design of House District 56.
We also find that race did not predominate in the design of House District 57. District 57, with a total black population percentage that remained almost unchanged at 68.49 percent, (Doc. 263-2 at 2), is a compact district that has the same northwestern orientation as District 55:
[[Image here]]
(Ala. Reapportionment Office, supra). The dissent asserts that the new district is not compact because it resembles the “landmass of the United Kingdom,” but offers no reason why this resemblance makes the district non-compact. (Dissent at 1387). As with the overall shape of District 55, the overall shape of District 57 is not suspicious.
The plaintiffs again drew districts that are not necessarily better than the district drawn by the legislature, and again they do not persuade us that race predominated in District 57. The Democratic Conference plaintiffs drew a district running from east to west instead of north to south:
House District 57 in Democratic Conference Plan A and Act 602
*1211[[Image here]]
(Doc. 287-14 at 18). The Black Caucus plaintiffs drew a District 57, shaded in pink in the following map, that made only minor changes:
House District 57 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 513). These alternative plans do not prove that race predominated.
The black population percentages of the alternative plans convince us that the black population percentage in the enacted district is not suspicious. Both the 1% Plan and Plan A have total black population percentages that are similar to the 68.47 percent in Act 602-72.51 percent in the 1% Plan and 66.1 percent in Plan A. (Doc. *1212296-1 at 2). Moreover, they deviate in opposite directions, making the enacted plan the median of the three options.
Our dissenting colleague makes several arguments, but none of them are persuasive. Our dissenting colleague argues that the percentages of the alternative plans support the opposite conclusion because Alabama’s plan came closer to the alleged target than the alternative plans did. (Dissent at 1387). But our colleague’s misbegotten rule that race predominated if the legislature met its alleged racial target, compels his conclusion. This approach is contrary to the approach of the Supreme Court. In addition, the dissent is correct that the drafters moved a large number of people in and out of the district, (Dissent at 1386), but, as repeatedly noted, this kind of turnover is unsurprising in an area that was redrawn to cure underpopulation.
The precinct splits are also unpersuasive. The enacted district split five precincts, including Pleasant Grove First Baptist Church. That split placed a higher percentage of black population into District 57, but a visual examination of the map shows that the split is produced by a relatively smooth J-shaped line:
Pleasant Grove First Baptist Church - Precinct in Act 602
[[Image here]]
*1213(APSX 124). The legislature put 6,679 people in District 57, 51 percent of them black, and 2,927 people in District 15, 22 percent of them black. (Def. Ex. 405 at 479, 558). We observe majority-black and majority-white blocks along both sides of the district line, and we are not persuaded that this precinct is evidence that race predominated. Our dissenting colleague contends that the reapportionment guidelines do not include exceptions for precinct splits based upon “smooth lines.” (Dissent at 1387). But the dissent’s analysis of House District 67 faults Alabama because it “could have made boundary choices that would have resulted in smoother lines.” (Id. at 1396). In addition, the Supreme Court commands us to consider a “district’s shape,” Miller, 515 U.S. at 916, 115 S.Ct. 2475, which includes consideration of a district’s “boundary lines.” Shaw, 509 U.S. at 646, 113 S.Ct. 2816. We find that race did not predominate in the design of House District 57.
We also find that race did not predominate in House Districts 58 and 59. District 58, which decreased in black population from 77.86 to 72.76 percent, is sandwiched between Districts 59 and 54, northeast of downtown Birmingham. (Doc. 263-2 at 2). District 59, which increased in black population percentage from 67.03 to 76.72 percent, (id.), borders District 54 to the south and District 58 to the' east. The incumbents in these districts worked together with Representative Todd to draw the districts. Both Districts 58 and 59 maintain their cores, and they extend northeast to pick up population from District 44.
[[Image here]]
(Ala. Reapportionment Office, supra).
The plaintiffs’ alternative plans do not persuade us that race predominated in the design of Districts 58 and 59. The Democratic Conference plaintiffs drew a similar District 58 and fail to convince us that their minor changes were compelled by traditional districting criteria:
House District 58 in Democratic Conference Plan A and Act 602
*1214[[Image here]]
(Doc. 287-14 at 14). The same is true of Districts 58 and 59 in the 1% Plan. Those districts are shaded in olive green and light blue, respectively, in the following maps:
House District 58 m Black Caucus 1% Plan and Act 602
[[Image here]]
*1215(APSX 514).
House District 59 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 515). In contrast, the Democratic Conference plaintiffs drew a different District 59, but they have not explained how that their changes prove that race predominated.
House District 59 in Democratic Conference Plan A and Act 602
[[Image here]]
*1216(Doc. 287-14 at 15).
The plaintiffs argue that the legislature split Pinson United Methodist Church precinct with majority-white District 44 along racial lines, (Doc. 256 at 61-62), but their argument fails. As an initial matter, the Democratic Conference plaintiffs split the same precinct in Plan A. They did so along different lines, but they placed a higher black population percentage in District 58 than in District 44. (Doc. 296-6 at 2). To split the precinct without placing a higher black population percentage in District 58 than District 44 would have required bizarre lines. The majority of the white population lives on the' east side of the precinct, contiguous with District 44, and the majority of the black population lives on the west side, contiguous with Districts 58 and 59. For Districts 58 and 59 to reach the white population in the precinct, they would have had to reach around the contiguous black population in a bizarre fashion, and the plaintiffs offer no legitimate reason why the drafters should have done so.
Pinson United Methodist Church Precinct in Act 602
[[Image here]]
(APSX 123). As Fairfax said about a different racially unbalanced precinct split, it was “just natural, unfortunately, population patterns, demographical population patterns.” (Doc. 296-7 at 143-44).
Nor do any of the other precinct splits in Districts 58 or 59 strike us as visually or statistically suspicious. If any two districts dispel the notion that the drafters pursued a mechanical goal of racial targets in every district, it is these two districts. District 58 fell short of its purported target by about 5 percentage points, and District 59 overshot its target by about nine percentage points. These two districts are directly beside each other. If the drafters had intended to meet a racial target, all they needed to do was transfer black population from District 59 to District 58. The drafters could have reached at least one of their *1217targets and come substantially closer to hitting their other target. But the drafters declined to do so because they did not have an overriding policy of mechanical racial targets in the design of either district. We find that race did not predominate in the design of House Districts 58 or 59.
Finally, we find that race did not predominate in the design of House District 60. The district, which barely increased in black population from 67.41 percent to 67.68 percent, is similar to its former iteration.
[[Image here]]
(Ala. Reapportionment Office, supra). The new district includes 80 percent of the same people as the old district. (ADC Supp. Ex 4). It reaches slightly farther southeast into central Birmingham, and it reaches slightly farther northeast to take population from District 51.
We find no evidence of racial predominance in the black population percentage or the number of people moved relative to the underpopulation of the district. Although most of the alternative plans drew District 60 with a lower black population percentage, the Reed-Buskey Plan came within 2.08 percentage points of the previous percentage, (Def. Ex. 406; CE 42), making it less suspicious that the enacted district is close to the previous percentage. And in a district that was underpopulated by 8,817 people, we do not find it suspicious that the drafters moved 9,170 people. Even our dissenting colleague acknowledges that this change is about as minimal as it gets in the thorny process of redistricting. (Dissent at 1388).
District 60 split only two precincts with a majority-white district, and neither split proves that race predominated. Fultondale Senior Citizen’s Center precinct is overwhelmingly white, and the drafters did not place an abnormal number of whites or blacks into either district or draw a suspicious line between them.
Fultondale Senior Citizen’s Center Precinct in Act 602
*1218[[Image here]]
(APSX 107). The drafters placed 858 people into District 60, 16 percent of whom are black, and 3,933 people into District 51, 8 percent of whom are black. (Def. Ex. 405 at 547, 565). In Gardendale Civic Center precinct, the drafters placed only three populated census blocks in District 60.
Gardendale Civic Center Precinct in Act 602
*1219[[Image here]]
(APSX 108). They placed 624 people in District 60, 47 percent of them black, and 13,739 people in District 51, 6 percent of them black. (Def. Ex. 405 at 546, 564). But the black population in this incursion accounts for less than two-thirds of a percent of the total population of this district, and removing the small incursions into this precinct by District 60 would cause the black population in District 60 to change negligibly from 67.68 to 67.97 percent. The plaintiffs and the dissent have not proven that the drafters split these precincts for predominantly racial purpose's,
The Democratic Conference plaintiffs drew a district that was different from the 2012 plan (and by extension the 2001 plan).
[[Image here]]
*1220(Doc. 287-14 at 16; Doc. 287-26 at 15). The Black Caucus plaintiffs drew the district differently than both the legislature and the Democratic Conference plaintiffs. The following map shows House District 60 in the 1% Plan district in orange:
House District 60 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 516). Neither group of plaintiffs explained why their choices were required by traditional districting criteria other than precinct splits, and any improvements are not obvious to us.
Our dissenting colleague makes several arguments, but none of them persuade us that race predominated. First, our dissenting colleague considers the black percentage in the enacted district to be “impressive” evidence of racial predominance. (Dissent at 1388). But, as repeatedly noted, this evidence is not conclusive of racial predominance and it is hardly suspicious in the light of the similar percentage under the Reed Buskey Plan. Second, our colleague reiterates his argument that the move of House District 53 was motivated by racial considerations and that move affected the drafting of House District 60. (Id. at 1388). We disagree, because the decision of the drafters to move House District 53 was motivated by massive under population in the Birmingham area. Third, our colleague argues that the enacted district is “irregularly shaped.” (Id. at 1388). But the shape of the enacted district is no more irregular than the district under the Black Caucus Plan or Plan A.
Fourth, our dissenting colleague argues that we have ignored evidence of precinct splits between House District 60 and other majority black districts. (Id. at 1389). He argues “that had [the defendants] not used any split precincts, [House District] 60 would have been 72.5% black — more than [five percent] above its racial target.” (Id. at 1389). Beyond the false premise that race predominated because the drafters hit their alleged racial target, this argument highlights the predicament of the drafters. On one hand, if they drafted the district without precinct splits, which would have increased the black population, the plan *1221would have been susceptible to attack as an attempt at packing. On the other hand, if the drafters split precincts to avoid retrogression, their plan was susceptible to attack as a racial gerrymander. The dissent’s mechanical view of racial predominance only exacerbates this dilemma. We find that race did not predominate in the design of House District 60.
j. House Districts 67, 68, 69, 70, 71, and 72 (West Black Belt)
We find that race did not predominate over traditional districting criteria in the design of House Districts 67, 69, and 72, but we find that race predominated in the design of House Districts 68, 70, and 71. Each of these districts is located in the rural Black Belt of south-central Alabama, and each was severely underpopulated:
[[Image here]]
These districts are contained in and surrounded by rural counties with high black population percentages. Pickens, Choctaw, Clarke, Monroe, Conecuh, and Butler counties are over 40 percent black; Hale and Marengo counties are over 50 percent black; Dallas and Perry counties are over 60 percent black; Wilcox, Sumter, and Lowndes counties are over 70 percent black; and Greene County is over 80 percent black. (U.S. Census Bureau, American Factfinder, supra). Unsurprisingly, as the districts expanded to gain more population, their racial composition did not change much.
2010 Census Total Black Population Percentage in Districts 67-72
[[Image here]]
Nevertheless, even in a rural area with majority-black population, only three of the districts matched the previous black population percentage with any precision.
District 67, which barely changed in racial composition, is a compact, sensible district. In the 2001 plan, it fell entirely within Dallas County. Because Dallas County does not have enough population to comply with the ± 1% deviation, the drafters filled out all of Dallas County and stretched the district north into Perry County to gain the remaining necessary population.
*1222[[Image here]]
(Ala. Reapportionment Office, supra). This district is compact and largely the same as the previous plan.
Both the Democratic Conference and Black Caucus plaintiffs traded Perry County for Chilton County, a choice that they do not adequately explain,
[[Image here]]
(Doc. 287-18 at 2; APSX 517). The alternative districts are no more compact, and the plaintiffs do not explain why their choice was required by traditional district-ing criteria other than precinct splits, which we address below. Chilton County, unlike Perry County, is not considered part of the Black Belt, which all parties have recognized as a community of interest. (Doc. 295 at 40). Alabama argues that the plaintiffs had race in mind, (id.), to which the plaintiffs reply that they split Chilton County because of population considerations and because Chilton County was already split, (Doc. 301 at 10). This skirmish is irrelevant because Alabama has identified a legitimate districting criterion that it followed when it drew District 67 differently than the plaintiffs did. And the design in Perry County is not suspicious on its own merits — the District 67 portion of Perry County is 60 percent black and the rest of Perry County is 70 percent black. (Def. Ex. 405 at 584, 601). The alternative plans do nothing to persuade us that race predominated in the ■design of District 67.
The plaintiffs have made two unpersuasive arguments about the overall design of this district. First, both sets of plaintiffs complained about the ± 1% population deviation, but these attacks are futile for the reasons we discussed earlier. Second, both *1223sets of plaintiffs argue that the drafters should not have extended the district outside of Dallas County, (Doc. 256 at 67-68; Doc. 258 at 63-64), but they now admit that Dallas County does not have enough population for a district within the deviation and their court-ordered alternative plans both extend the district outside of Dallas County. An equally good choice is not proof that race predominated.
In the six alternative plans, District 67 consistently had a black population percentage close to the percentage in the enacted plan. The New Black Caucus Plan, the McClammy Plan, and the Knight Plan even came within half a percentage point.
2010 Census Total Black Population Percentages for District 67 Under Various Plans
[[Image here]]
When all of the alternative plans are so similar to the old district and the enacted district, it is unreasonable to infer that race predominated based on the bare fact that the enacted district matches the percentage of the old district.
The plaintiffs also argue that the drafters unnecessarily split precincts, but their argument fails. The district splits four precincts, all of them with majority-black District 72, (Def. Ex. 405 at 583-84, 600-01), and the plaintiffs fail to explain how these splits prove that race predominated. Overall the splits placed roughly equal percentages of black population in the two districts — 656 people in District 67, 68 percent of them black, and 6,319 people in District 72, 70 percent of them black. (Id,). Not every precinct split is evidence of racial gerrymandering, and the plaintiffs fail to explain how the splits prove that race predominated.
The dissent agrees with us that race-neutral districting criteria explain the addition of portions of Perry County to the district, but nevertheless finds that race predominated because the drafters came within “three black people” of their alleged target. (Dissent at 1395). To find, as the dissent would have us, that race predominated because the alleged target was met defies the instructions of Supreme Court to presume that the legislature acted in good faith, Miller, 515 U.S. at 915, 115 S.Ct. 2475, and that Alabama’s policy only “provides evidence,” of racial predominance. Ala. Legislative Black Caucus, 135 S.Ct. at 1267. In addition, that every other factor is indicative of race-neutral drafting choices suggests that the Dissent’s approach fails to consider the totality of the evidence. Because we must follow the approach of the Supreme Court, we find that race did not predominate in the design of District 67.
Based on the way that District 68 splits counties and precincts, we find that race predominated in its design. The district, which increased in black population percentage from 62.55 percent to 64.56 percent, (Doc. 263-2), was difficult to repopulate because it was surrounded by severely underpopulated districts. To the north were Districts 69, 71, and 72; to the west were District 65 and Mississippi; to the south was District 64; and to the east was District 90. (Def. Ex. 406). Moreover, Districts 65 and 71 could not grow west into Mississippi (visible in black on these maps).
*1224[[Image here]]
(Ala. Reapportionment Office, supra). The incumbents in these districts, marked by green dots, further limited the directions that District 68 could grow:
2012 District Lines with Incumbent Locations
[[Image here]]
(Def. Supp. Ex. 3). In response, the drafters took an irregular district from 2001 and made it irregular in different ways.
Both Plan A and the 1% Plan include fewer counties in District 68, suggesting that the drafters might have subordinated this part of the Committee guidelines to race. The Democratic Conference plaintiffs drew their own irregular District 68, but it entered two fewer counties.
House District 68 in Democratic Conference Plan A and Act 602
*1225[[Image here]]
(Doc. 287-18 at 3). District 68 in the 1% Plan also enters two fewer counties.
House District 68 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 518).
The county splits raise further concerns. In all six splits, District 68 took a population with a higher percentage of black residents than the adjacent district or districts in that county. The part of the population of Baldwin County in District 68 is 78 percent black, but the part in District 64 is 15 percent black. (Def. Ex. 405 at 573, 584). The part of Clarke County in *1226District 68 is 67 percent black, but the part in District 65 is 24 percent black. (Id. at 577, 586). The part of Conecuh County in District 68’is 61 percent black, but the part in District 90 is 22 percent black. (Id. at 587, 629). The part of Marengo County in District 68 is 72 percent black and the part in District 72 is 70 percent black, but the part in District 65 is 32.50 percent black and the part in District 71 is 31 percent black. (Id. at 578, 588, 596, 600). The part of Monroe County in District 68 is 59 percent black, but the part in District 64 is 8 percent black. (Id. at 574, 590). And the part of Washington County in District 68 is 82 percent black, but the part in District 65 is 17 percent black. (Id. at 580, 591).
The inference that race predominated over the guideline on the number of counties is strengthened by the difference in the overall black population percentage in the alternative districts. District 68 in the 1% Plan is 53 percent black, and District 68 in Plan A is 57 percent black. (Doc. 296-1). Moreover, none of the earlier alternative plans had a total black population percentage above 55 percent. (Common Ex. 46). In short, the black population percentage in District 68 is suspiciously high.
The district also split 30 precincts, many along racial lines. Some of the precinct splits have legitimate explanations. For instance, the split of Bashi Methodist Church precinct, which creates one of the knobs along the northwestern border of the district, keeps the residence of the incumbent in the district. (APSX 66; Def. Supp. Ex. 2). In addition, the Black Caucus plaintiffs also split this district in their 1% Plan. (APSX 633 at 5).
But it is apparent that the drafters split other precincts to find black population. (See, e.g., APSX 71; APSX 71A). Alabama admits that splitting the precincts raised the black population percentage in the district by almost seven percentage points, meeting the alleged target almost exactly. We discuss the suspicious splits individually in the following paragraphs.
In Baldwin County, the drafters placed two clusters of majority-white blocks from Tensaw Volunteer Fire Dept precinct in District 64 and placed the intervening majority-black blocks in District 68.
Tensaw Volunteer Fire Dept Precinct in Act 602
*1227[[Image here]]
(APSX 52). But this split placed only 12 people in District 64. (Def. Ex. 405 at 572). In Vaughn Community Center precinct, District 64 follows a body of water until doing so would require putting large majority-black blocks into majority-white District 64, at which point it deviates and traces a line between the majority-white and majority-black blocks.
Vaughn Community Center Precinct in Act 602
*1228[[Image here]]
(APSX 53). It put 501 people in District 58, 79 percent of them black, and 304 people in District 64, 14 percent of them black. (Def. Ex. 405 at 584, 572).
In Clarke County, majority-white District 65 takes a small part of the majority-white blocks in Jackson City Hall precinct and leaves all of the adjoining majority-black blocks in District 68.
Zoom of Jackson City Hall and Skipper Fire Station-Jackson National Guard-Jackson Fire Dept. Precincts in Act 602
*1229[[Image here]]
(APSX 71A). On the other side of the precinct border in Skipper Fire Station-Jackson National Guard-Jaekson Fire Dept., the drafters put a tiny portion of the precinct in District 68, but out of the 295 people included, 61 percent were black. (APSX 71; Def. Ex. 405 at 585). The drafters put 3,943 people from this precinct in District 65, of whom only 20 percent were black. (Def. Ex. 475 at 576). Although including a small population is sometimes not strong evidence that race predominated, in this case we find that the split is part of a pattern of racial fine-tuning. In Overstreet Grocery precinct, the split between Districts 65 and 68 forms a somewhat regular line, but all of the majority-black blocks were put in District 68.
Overstreet Grocery Precinct in Act 602
*1230[[Image here]]
(APSX 70). The split put 368 people in District 68, 80 percent of them black, and 286 people in District 65, 17 percent of them black. (Def. Ex. 405 at 584, 576). District 68 also crosses into Fulton City Hall precinct at three points to pick up majority-black blocks.
Fulton City Hall Precinct in Act 602
*1231[[Image here]]
(APSX 67). Although it left some majority-black blocks in majority-white District 65 and moved only 73 people, it put a 62-percent black population in District 68, compared with a 5-percent black population in District 65. (Def. Ex. 405 at 586). Taken together, these precinct splits are evidence that race predominated in District 68.
The two precinct splits in Conecuh County do not provide strong evidence that race predominated. First, the drafters split Brownville Fire Dept by putting a 55 percent black block of 40 people in District 68. (APSX 73; Def. Ex. 405 at 586). This split is only slight evidence of racial predominance because it forms a straight line with part of the precinct border and leaves majority-black- blocks in majority-white District 90.
Brownville Fire Dept. Precinct in Act 602
*1232[[Image here]]
(APSX 73). Second, the drafters split Nazarene Baptist Church precinct by putting most of the majority-black blocks in District 68, but they also put majority-white blocks in District 68 and majority-black blocks in majority-white District 90.
Nazarene Baptist Church Precinct in Act 602
*1233[[Image here]]
(APSX 76). This split is not evidence that race predominated.
The precinct splits in Marengo County provide mixed evidence. First, in Cornerstone Church precinct, the drafters drew an irregular arm that put mostly majority-black blocks in District 68.
Cornerstone Church Precinct in Act 602
[[Image here]]
*1234(APSX 165). The split put 699 people in District 68, 87 percent of whom were black, and 1,121 people into majority-white District 65, 27 percent of whom were black. (Def. Ex. 405 at 577, 587). Although the drafters left several accessible majority-black blocks in District 65, this split provides some evidence that race predominated. Second, in Thomaston precinct, the drafters split the precinct along a relatively smooth line.
Thomaston Precinct in Act 602
[[Image here]]
(APSX 172). Because the drafters did not snake around looking for black population, we find that this split is not evidence of racial predominance. Third, the drafters split Dixon’s Mill precinct into three parts, putting the mostly black section in District '68 and the mostly white sections in Dis-triet 65.
Dixon’s Mill Precinct in Act 602
*1235[[Image here]]
(APSX 167). The resulting split placed 1,377 people in District 68, of whom 89 percent were black, and 234 people in District 65, of whom 7 percent were black. (Def. Ex. 405 at 588, 577). This split brought in majority-black blocks with an irregular shape while avoiding majority-white blocks, and it is evidence that race predominated. Finally, in Octagon precinct, the drafters put two majority-black blocks from opposite corners of the precinct in District 68.
Octagon Precinct in Act 602
*1236[[Image here]]
(APSX 169). Of the 33 people put in District 68, 91 percent were black, compared with the 218 people put in District 65, 22 percent of whom were black. (Def. Ex. 405 at 577, 588). This split is evidence of racial sorting.
In Monroe County, three split precincts contribute to our finding that race predominated. First, in Shiloh/Grimes precinct, the drafters placed all four of the majority-black blocks in District 68 using a suspicious indentation.
Shiloh/Grimes Precinct in Act 602
*1237[[Image here]]
(APSX 219). The split placed 87 people m District 68, 76 percent of them black, and 26 people in District 64, 4 percent of them black. (Def. Ex. 405 at 590, 574). Second, in Frisco City FD precinct, the drafters created a narrow protrusion into District 64 with two irregular bumps to reach majority-black blocks.
Frisco City FD Precinct in Act 602
*1238[[Image here]]
(APSX 211). The drafters put 1,319 people from the precinct in District 68, 49 percent of whom were black, and 1Í0 people in District 64, none of whom were black. (Def. Ex. 405 at 590, 574). Third, in Excel-Coleman precinct, the drafters put a mostly black row of census blocks in District 68. The only majority-white blocks in the district were necessary to reach majority-black blocks, and the drafters missed no easily accessible majority-black blocks.
*1239Excel-Coleman Precinct in Act 602
[[Image here]]
(APSX 210). In the split, District 68 took 173 people, 48 percent of whom were black, and District 64 took 3,379 people, 8 percent of whom were black. (Def. Ex. 405 at 590, 574).
Finally, we find that two splits in Washington County are evidence that race predominated. In McIntosh Voting House Voting District precinct, District 68 invaded in a winding fashion to reach almost every reasonably accessible majority-black block.
McIntosh Voting House Voting District Precinct in Act 602
*1240[[Image here]]
(APSX 286). The drafters put 467 people into District 68, 77 percent of whom were black, and 1,491 people in District 65, 5 percent of whom were black. (Def. Ex. 405 at 590, 579). In Carson/Preswick precinct, the drafters zigged and zagged to put almost every majority-black block in District 68 and almost every majority-white block in District 65.
Carson/Preswick Precinct in Act 602
*1241[[Image here]]
(APSX 283). They assigned 241 people to District 68, of whom 86 percent were black, and 329 people to District 65, of whom 17 percent were black. (Def. Ex. 405 at 590, 580). Overall, the irregularity of the district and the precinct splits convince us that race predominated in the design of District 68.
We further conclude that District 68 satisfies strict scrutiny. The legislature had a strong basis in evidence that a black population percentage of 62 to 65 percent was necessary in District 68 to comply with the Voting Rights Act, and the drafters drew a district within this range (64.56 percent). District 68 is Representative Jackson’s district, and he made clear at a public hearing of the Committee that the district should be “sixty-two or sixty-five percent.” (Doc. 30-23 at 8). House District 68 is contained mostly within Senate District 23, and the longtime incumbent in Senate District '23, Hank Sanders, testified at even greater length about why majority-black districts needed a black population percentage of at least 62 percent:
One of many concerns is we are not to have any less African-American — the majority African-American districts than you have, and that those districts ought not be less than 62 percent. And I just want to say why 62 percent, ought not to be less than 62 percent. Many times a population of a district is not *1242reflective of the voters at all in that district. Sometimes a lot of people don’t vote. Sometimes a lot of people can’t vote. They might be in prison or other kinds of institutions. Sometimes a lot of folks are discouraged for one reason or another. So I would hope that 62 percent is a minimal for the majority African-American district.
(Doc. 80-28 at 6). The specific statements of the incumbents in the area, given to the Committee, provided the drafters with a strong basis in evidence to believe that a black population percentage of 62 to 65 percent was necessary in this area.
The plaintiffs’ evidence at trial confirmed what Sanders and Jackson told the Committee. Dr. Reed testified that a majority-black district needs to be at least 60 percent black to allow minority voters to elect the candidate of their choice. (Doc. 216 at 159-60; Doc. 203 at 69-70). In addition, Dr. Theodore Arrington, an expert witness for the plaintiffs in this litigation, testified in 2001 that a 61 percent black population percentage in nearby Dallas County did not guarantee black voters the ability to elect a county-commission candidate of their choice. (Doc. 217 at 80-81). See also Wilson v. Jones, 130 F.Supp.2d 1315, 1326 (S.D. Ala.) (“Dr. Ar-rington’s position [is] that at least a 62% black voting age population was needed to assure blacks an opportunity to elect their choices in a district....”), aff'd sub nom. Wilson v. Minor, 220 F.3d 1297 (11th Cir. 2000). The legislature may not have relied on this testimony, but it confirms that even the plaintiffs’ experts would agree with the evidence on which Alabama did rely.
The plaintiffs argue that the legislature also would have had a strong basis in evidence for a lower percentage, but even if it did, it would not prove that Alabama lacked a strong basis in evidence for the percentage it chose. The standard of a strong basis in evidence “does not demand that a State’s actions actually be necessary to achieve a compelling state interest in order to be constitutionally valid.” Ala. Legislative Black Caucus, 135 S.Ct. at 1274 (quoting Brief for the United States as Amicus Curiae 29). “[L]egislators ‘may have a strong basis in evidence to use racial classifications to comply with a statute when they have good reasons to believe such use is required, even if a court does not find that the actions were necessary for statutory compliance.’” Id.(quoting Brief for the United States as Amicus Curiae 29). Legislators might have a strong basis in evidence for several different percentages.
The dissent makes several arguments of its own, but none of them defeat Alabama’s strong basis in evidence. First, the dissent asserts that there is no evidence that the legislature “ever applied,” “contemplated,” or “actually relied” on the testimony of Senator Sanders or Representative Jackson. (Dissent at 1363-64). But the evidence before us proves that they did. At the public hearing in Thomasville, Representative Jackson — a member of the Black Caucus — explained that majority-black districts in his area should be “sixty-two percent or sixty-five percent.” (Doc. 30-23 at 8). At the public hearing in Selma, Senator Sanders — also a member of the Black Caucus — told Senator Dial that none of the majority-black districts should be less than 62 percent black. (Doc. 30-28 at 6). Senator Sanders explained why he thought this minimum was necessary — low turnout, incarceration rates, and voter apathy. (Id.) And Senator Dial testified that he considered Senator Sanders’s opinion credible. (Doc. 215 at 37). Reliance on Representative Jackson’s and Senator Sanders’s testimony is not a mere litigating position.
Second, the dissent argues that “the comments were not ‘sufficiently measura*1243ble to permit judicial scrutiny of the policies adopted to reach them.’ ” (Dissent at 1364 (quoting Fisher v. Univ. of Tex., — U.S. -, 136 S.Ct. 2198, 2211, 195 L.Ed.2d 511 (2016))). But the Supreme Court has never required that a state rely on studies to justify the drafting of a voting district in which race predominated. Cf. Shaw, 517 U.S. at 915, 116 S.Ct. 1894 (“[W]e have not always provided precise guidance on how closely the means (the racial classification) must serve the end (the justification or compelling interest.”)). Indeed, the Supreme Court admonished us that narrow tailoring does not “insist that a legislature guess precisely what percentage reduction a court ... might eventually find to be retrogressive.” Ala. Legislative Black Caucus, 135 S.Ct. at 1273. More fundamentally, the comments of Senator Sanders and Representative Jackson are capable of being scrutinized. For instance, we know that Representative Jackson formulated his opinion as the incumbent of District 68, and he spoke as the representative of that specific geographic area. If detailed comments of an incumbent legislator do not provide a strong basis in evidence the burden of proving good reasons has been transformed into a burden of proving actual necessity.
Third, the dissent asserts that Senator Sanders and Representative Jackson “proposed imprecise remedies” that do not address the “precise question,” whether “the remedy proposed ... was tailored to ... achieving [section] 5 compliance.” (Dissent at 1362-63). But Senator Sanders and Representative Jackson proposed precise remedies — indeed, exact percentages- — -to ensure compliance with the Voting Rights Act. Representative Jackson said a majority-minority district should be 62 percent to 65 percent African American, (Doc. 30-23 at 8), and Senator Sanders said a district should not fall below 62 percent. (Doc. 30-28 at 6).
Fourth, the dissent argues that other “three-judge court[s]” have found it “easy ... to conclude” that the use of a racial target cannot satisfy strict scrutiny, (Dissent at 1367), but, as noted above, the decisions the dissent cites are distinguishable. In Page v. Va. State Bd. of Elections, No. 3:13-CV-678, 2015 WL 3604029 (E.D. Va. June 5, 2015), the district court struck down a redistricting plan that relied on a racial target because the drafters presented no evidence that it was necessary to comply with the Voting Rights Act. Id. at *18. And in Smith v. Beasley, 946 F.Supp. 1174 (D.S.C. 1996), the court struck down a similar redistricting plan that employed a racial target without supporting evidence. Id. at 1210. Here, by contrast, the testimony of incumbent legislators provided the basis for the choices of the drafters.
Fifth, the dissent suggests that the comments of Representative Jackson and Senator Sanders are “exactly the type of stereotyping about black voting behavior that strict scrutiny is intended to prohibit,” (Dissent at 1368), and he repeatedly recites this charge. (Id. at 1356-57, 1358-59, 1362, 1364-65, 1368). As already explained, it is hard to conceive how the suggestions of two longtime incumbent legislators based on, in the words of Representative Jackson, their “f[i]ght,” (Doc. 30-23 at 7), against problematic redistricting plans in the past is not a strong basis in evidence, but instead pernicious racial stereotyping. This evidence is just the kind of evidence — rooted in knowledge of the redistricting process and representative of the interests of black voters — that provides a district-specific, strong basis in evidence. Representative Jackson pleaded that the voters of House District 68 “be heard,” (id.), and the drafters listened.
The defendants have established good reasons for their design of District 68 *1244based on Senator Sanders’s and Representative Jackson’s statements and the plaintiffs fail to rebut this strong basis in evidence. We will not enjoin the use of House District 68.
Next, we find that race did not predominate in the design of House District 69. The district, which increased negligibly in black population from 64.11 to 64.21 percent, changed very little. It contains all of Wilcox and Lowndes Counties and parts of
Autauga and Montgomery Counties. In the 2001 plan, the district contained part of Dallas County instead of part of Montgomery County. The legislature put all of Dallas County in District 67, which forced it to move District 69 east to find population. The district could not easily move south or west, because all of the adjacent districts were already underpopulated, so District 69 instead reached into Montgomery County:
[[Image here]]
(Ala. Reapportionment Office, supra). The design of the district is not bizarre. The shape of the new district is similar to the old district and no odder.
Neither of the court-ordered alternative plans establishes that race predominated in the design of District 69. The enacted district is at least as regular and compact as the alternative in Plan A, and it is more regular and compact than the alternative in the 1% Plan.
Overlay of Democratic Conference Plan A District 69
*1245[[Image here]]
(Doc. 287-18 at 4).
Overlay of Black Caucus 1% Plan District 69
[[Image here]]
(APSX 519). In fact, both alternative districts increased the number of counties in the district. Neither plaintiff explains why *1246the extra county is required by traditional districting criteria.
The only slightly odd feature of the enacted district, the hook that wraps around into Montgomery County, avoids incumbent conflicts. To reach that population, District 69 (in light blue) had to wrap around District 78 to avoid the residence of the incumbent there:
Incumbent Locations in and Near District 69
[[Image here]]
(Def. Supp. Ex. 2). Because neither plaintiff extended this district into Montgomery County, they did not show that it was possible to draw this portion of the district in a better way.
The plaintiffs also argue that the drafters split precincts to increase the black population, but their argument is unconvincing. In Montgomery County, they cite the split in Ramer Library precinct, where the drafters placed a 97 percent black population in District 69. (Doc. 256 at 87; Doc. 258 at 66). This percentage is high, but the lines follow a railroad and a state road. The following map puts the precinct side-by-side with a Census Bureau map of District 69 (orange line):
*1247[[Image here]]
(APSX 240; U.S. Census Bureau, State Legislative District Reference Map: State House District 69 (Alabama), http://www 2.census.gov/geo/maps/dcl0map/SLD_Ref MapAower/st01_al/sldl01069/DC10SLDL 01069_001.pdf). Moreover, the drafters placed only 61 people from this precinct in District 69, and the Democratic Conference plaintiffs’ map establishes that the drafters could have reached for even more black population in this area. (ADC Supp. Ex. 19C).
Almost all of the other precinct splits in District 69 fail to establish that race predominated. Several of them — IF Al. Industrial Development Training, 2F Fire Station No. 14 Voting District, 2G Hayneville Road Community Center Voting District, 5E Fitzpatrick Elementary School Voting District, and 5N Peter Crump School Voting District precincts — are split between or among majority-black districts, and the plaintiffs fail to identify any racial pattern. (APSX 223, 226, 227, 289, 241, 244). The legislature split 2D Montgomery Boys Club Voting District precinct along unpopulated blocks, which is a race-neutral choice that keeps the southwestern corner of the precinct with contiguous populated blocks.
2D Montgomery Boys Club Voting District Precinct in Act 602
*1248[[Image here]]
(APSX 225). We cannot find a racial pattern in the split of 21 Southlawn Elementary School Voting District precinct, which improved on the shape of the precinct.
21 Southlawn Elementary School Voting District Precinct in Act 602
*1249[[Image here]]
(APSX 229). The split of 5A Seth Johnson Elementary School Voting District placed no population in District 69. (Def. Ex. 405 at 593). And the split of Booth Volunteer Fire Station precinct in Autauga County mostly divided majority-white blocks between Districts 42 and 69, placing 808 people in District 69, 28 percent of them black, and 781 people in majority-white District 42, 17 percent of them black. (Id. at 591, 532).
Booth Volunteer Fire Department Precinct in Act 602
*1250[[Image here]]
(APSX 50).
According to our dissenting colleague, the drafters split 5B Snowdoun Womens Club Voting District to create a “suspicious land bridge” that allowed them to meet their alleged target. (Dissent at 1397-98 n.23).
5B Snowdoun Womens Club Voting District Precinct in Act 602
*1251[[Image here]]
(APSX 239). While we agree with our colleague that this split allowed the drafters to draw population from 5N Peter Crump School Voting District and 5E Fitzpatrick Elementary School Voting District into District 69, we disagree that the only explanation of the split is race. (Dissent at 1397 and n. 23). One of the drafters’ stated goals was incumbent protection and this split allowed the drafters to avoid an incumbent’s residence and add additional majority-white population from Snowdoun and majority-black population from Peter Crump and Fitzpatrick to the underpopulated District 69. (Def. Ex. 405 at 145). And even if the drafters split the precinct to create a land bridge, they did not add a number of majority-black census blocks in 5N Peter Crump that would have allowed them to hit their alleged target. At best,, this split is suspicious, but it is not conclusive that race predominated.
There is one suspicious precinct split in Autauga County, but the population placed within or without the district on the basis of race is insignificant. In Safe Harbor Ministries precinct, the drafters appear to have followed Alabama SR-14 and a horizontal line for most of the split, except for a stepwise deviation in the eastern corner that put exclusively majority-black blocks in District 69 and exclusively majority-white blocks in District 42.
*1252[[Image here]]
(APSX 51; U.S. Census Bureau, State Legislative District Reference Map: State House District 69, http://www2.census.gov/ geo/maps/delOmap/SLD_RefMapAower/st 01_a]/sldl01069/DC10SLDL01069_001.pdf). But this deviation placed only 68 people in District 69, and we cannot say that race predominated on the basis of that number. If we remove this precinct from the district entirely, the black population percentage stands almost unchanged at 64.39 percent.
The dissent is correct about several statistics in this district, but it misunderstands their significance. We agree that the legislature drew a district with a 64.21 percent black population, which is above the supposed target of 64.16 percent. (Dissent at 1396). But this percentage, without evidence about what choices reflect the subordination of traditional districting criteria to race, is insufficient to satisfy the plaintiffs’ burden of proof. We also agree that the drafters moved 24,373 people in a district that was underpopulated by 7,949, (id. at 149), but it is unsurprising that the drafters had to move a lot of people to draw a sensible district with enough population in an area with several underpopulated districts. As with the black population percentage, we cannot find that race predominated without evidence of how the drafters subordinated traditional district-ing criteria to race. For the reasons discussed above, the plaintiffs have failed to prove that race predominated in the design of House District 69.
We find that race predominated in the design of House District 70, which is an urban district centered on the city of Tuscaloosa. The black population percentage in this district increased slightly from 61.83 to 62.03 percent. (Def. Ex. 406 at 657; Def. Ex. 405 at 594). The plaintiffs argue that the drafters split precincts and violated principles of compactness to create the northeastern portion, shaped like the mouth of a “roaring tiger,” (Doc. 272 at 83). Their argument is persuasive.
*1253In 2012, the legislature moved the district to the east and added the tiger’s mouth in Tuscaloosa:
[[Image here]]
(Ala. Reapportionment Office, supra). The design shifted black population from District 70 to District 71, a majority-black district constrained by Mississippi to its west and by underpopulated districts on most of its other borders. The legislature then reached into the city of Tuscaloosa with the tiger’s mouth to pick up black population.
The only explanation for the bizarre northeastern portion of the district is race. Maps of Holt Armory and Peterson Methodist Church precincts make this pattern clear:
Holt Armory Precinct in Act 602
[[Image here]]
*1254(APSX 274). The split of Holt Armory put 3,809 people in District 70, of whom 65 percent were black, and 2,032 people in District 62, of whom 17 percent were black. (Def. Ex. 405 at 594, 569). The mouth of the tiger also reaches into neighboring Peterson Methodist Church precinct for accessible majority-black blocks.
Peterson Methodist Church Precinct in Act 602
[[Image here]]
(APSX 279). This split put 328 people in District 70, 92 percent of whom were black, and 2,274 people in District 62, 7 percent of whom were black. (Def. Ex. 405 at 594, 569).
We find that several other precinct splits are also suspicious. In McFaland Mall precinct, the drafters used irregular lines to place 13,374 people into District 70, 53 percent of them black; 2,925 people into District 62, 21 percent of them black; and 134 people into District 63, 10 percent of them black. (Id. at 570, 571, 594).
McFaland Mall Precinct in Act 602
*1255[[Image here]]
(APSX 277). In Bama Mall precinct, the drafters used straight lines along zero-population areas to divide majority-black District 71 from District 70, but they also created irregular incursions in the north by majority-white District 63 that placed 497 people in District 63, only 5 percent of them black, compared with 5,907 people in District 70, 59 percent of them black.
Bama Mall Precinct in Act 602
*1256[[Image here]]
(APSX 271).
Alabama argues that it could draw a District 70 — in isolation — with no split precincts and a higher black population percentage. (Doc. 263-3 at 2). But Alabama placed all of Stillman College precinct in the hypothetical District 70, adding 6,002 people, 94.07 percent of them black. (Def. Ex. 405 at 593, 598). It also placed all of Stillman College precinct in its hypothetical District 71. (Doc. 263-3). This choice both undermines the probative value of the exercise and masks racial gerrymandering in the other splits.
Both plaintiffs drew a majority-black District 70 without the tiger’s mouth. The Democratic Conference plaintiffs managed to draw a district with a black population percentage of 61.9 percent, only 0.1 point lower than the enacted plan.
House District 70 in Democratic Conference Plan A and Act 602
*1257[[Image here]]
(Doc. 287-18 at 5; Doc. 296-1 at 2). This alternative district split only one precinct. (Doc. 296-6 at 2). The Black Caucus plaintiffs drew a district that eliminated the mouth of the tiger and had a significantly lower black population percentage of 57.52 percent. (Doc. 296-1 at 2).
House District 70 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 520). This alternative district split no precincts. (Doc. 300-1 at 49). The alternative plans in this district are evidence that race predominated. We find that race predominated in the design of District 70.
We also conclude that Alabama failed to prove that District 70 satisfies strict scrutiny. Alabama makes no district-specific arguments that this district survives strict scrutiny. The comments of Senator Sanders and Representative Jackson do not provide a strong basis in evidence for this district because of differences between District 70 and their districts. District 70 is *1258in a different part of the state than the districts that Sanders and Jackson represent, and District 70 is more urban than those districts — 87 percent of the population lived in the urbanized area of Tuscaloosa in 2010. (Def. Supp. Ex. 3 at 53). The state also fails to prove that Plan A does not comply with federal law or the Committee guidelines, and District 70 in Plan A is just shy of 62 percent black and involves less consideration of race. We recognize the difficulties facing the drafters in this region, but the defendants failed to satisfy their burden under strict scrutiny. We must enjoin the use of District 70 in future elections.
We also find that race predominated in the design of House District 71. The black population increased from 64.3 to 66.9 percent, (Def. Ex. 406 at 657; Def. Ex. 403 at 422), despite severe underpopulation problems and the challenge of keeping underpopulated District 70 as a majority-black district. The legislature drew a district that split six counties and contained none of them wholly within it, which is two more counties than the Democratic Conference plaintiffs included in Plan A. Moreover, one of the county splits placed a higher black population percentage in District 71 than in the adjacent district and did so using an irregular shape with suspicious precinct splits.
District 71 was severely underpopulated, and it bordered Mississippi to the west, underpopulated districts to the south and east, and mostly underpopulated districts to the north. The only adjacent district with excess population was District 62 to the northeast, which explains why District 71 grew to the northeast. As discussed already, District 71 also grew into District 70, which then shifted east to take population from District 62:
[[Image here]]
*1259(Ala. Reapportionment Office, supra). The growth of District 71 to the northeast was limited by two other incumbents in the Tuscaloosa area, which partially explains the wraparound at the convergence of Districts 62, 70, and 71.
Location of Incumbents in Districts 61, 62, and 70
[[Image here]]
(See Def. Supp. Ex. 2).
Both plaintiffs split fewer counties in their alternative plans. The Democratic Conference plaintiffs did not enter Pickens or Choctaw Counties, instead making Sumter and Marengo Counties whole and filling out more of other counties:
House District 71 in Conference Plan A and Act 602
[[Image here]]
(Doc. 287-18 at 6). The Black Caucus plaintiffs drew a different district with all of Marengo County and parts of Pickens, *1260Sumter, and Wilcox Counties. The following map shades District 71 in yellow:
House District 71 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 521). These alternative plans are evidence that the drafters used race to the detriment of the Committee guideline on counties.
Several county splits are not suspicious. The percentages in Greene, Marengo, and Sumter Counties are relatively close to the percentages in the adjoining districts.
Black Population Percentages in Greene, Marengo, and Sumter Counties by House District
[[Image here]]
(Def. Ex. 405 at 595-98, 578, 566). Only in Greene County does the split strike us as odd, but Hinaman testified at trial that Representative Harper requested that part of Greene County because he “had a house or a cabin on that property in Greene County and was thinking of potentially moving there.” (Doc. 217 at 151). This explanation is a race-neutral reason for the county split.
Only one precinct split in these counties is suspicious, but it is not significant. In Marengo County, the legislature split Jefferson precinct along racial lines:
Jefferson Precinct in Act 602
*1261[[Image here]]
(APSX 168). The legislature put 79 people in District 65, 4 percent of them black, and 637 people in District 71, 85 percent of them black. But this split does not affect a significant number of people by itself.
We also find no evidence in Tuscaloosa County that race predominated. The splits with Districts 61, 62, and 63 are suspicious at first glance, but the Democratic Conference plaintiffs drew an alternative district with a similar shape in Tuscaloosa County. The following map marks their District 71 with a black line and shades the enacted plan in orange and red:
Democratic Conference District 71 in Tuscaloosa County
*1262[[Image here]]
(Doc. 287-18 at 6). Fairfax also testified that county splits are not always the result of racial considerations:
Q. Okay. So is this true, that if you look at a county that’s split three ways and one section is 60-65 percent black, the other section’s 20 percent, the other section’s 15 percent black, that’s not enough to tell you whether there was something hinky going on, is there? There might be other reasons—
A. Right, absolutely. That’s just one piece.
(Doc. 296-7 at 137). And the southern border is explained by a state highway, an interstate highway, and a body of water, while the odd mouth around District 62 has to do with the residences of incumbents.
Census Bureau Map of House District 71
*1263[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House Dis-triet 71 s (Alabama), https://www.census. gov/geo/maps-data/maps/sldl/st01_al.html).
Location of Incumbents in Districts 62 and 70
[[Image here]]
[[Image here]]
Most of the precinct splits in Tuscaloosa County are not suspicious. Northport Community Center precinct appears suspi-mostly zero-population blocks.
Northport Community Center Precinct in Act 602
*1264[[Image here]]
(APSX 278). This precinct split provides no evidence that race predominated in District 71.
The split of Pickens County follows traditional districting criteria.
Census Bureau Map of House District 71 in Pickens County
*1265[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 71 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01071/DC10SLDL01071_ 001.pdf; see also U.S. Census Bureau, Tig-erWEB, http://tigerweb.geo.census.gov/ tigerweb/). The district begins in the northwest at the Mississippi border along a precinct line just north of Piekensville. It follows precinct lines to the northeast along Coal Fire Creek until it reaches County Road 26. It then splits Carrollton 4 Service Center precinct along County Road 26 to the intersection with State Route 86, at which point it splits the precinct along County Road 26, then follows the precinct boundary, which is also State Route 17.
Carrollton 4 Service Center Precinct in Act 602
*1266[[Image here]]
(APSX 250). From the city of Carrollton, the drafters assigned a 45 percent black population to District 71 and a 41 percent black population to District 61. (Def. Supp. Ex. 3). The district then splits Aliceville 2 Nat’l Guard Armory precinct with no apparent racial pattern, putting a 78 percent black population in majority-white District 61 and an 82 percent black population in District 71. (Def. Ex. 405 at 596, 567). Finally, the district reaches precinct lines and exits the county. The split of the city of Aliceville provides no evidence that race predominated; the legislature put an 86 percent black population in District 71 and a 71 percent black population in District 61, both of which are higher percentages of black population than District 71 has as a whole. (Def. Supp. Ex. 3).
Aliceville 2 Nat’l Guard Armory Precinct in Act 602
*1267[[Image here]]
(APSX 249). We find that the split of Pickens County does not prove that race predominated.
But the design of the district in Choctaw County is evidence that race predominated. The following map of the portion of District 71 in Choctaw County shows District 71 in gray, District 65 in yellow, and precinct lines in blue:
Close-Up of District 71 in Choctaw County in Act 602
[[Image here]]
(Def. Supp. Ex. 38 (precinct and county labels added by the Court)). Along the border with Mississippi, the district includes both Yantley-Cromwell precinct and part of Lisman-Pushmataha precinct. The split of Lisman-Pushmataha precinct put a 56 percent black population into a majority-white district and an 89 percent black population into District 71. The line does not wind around suspiciously, but it does increase the black population percentage for the district as a whole.
*1268Lisman-Pushmataha Precinct in Act 602
[[Image here]]
(APSX 64). The district line took an 87 percent black portion of Riderwood-Rock Springs precinct, leaving a 35 percent black portion in District 65. (Def. Supp. Ex. 405 at 595, 576).
Riderwood-Rock Springs Precinct in Act 602
[[Image here]]
*1269(APSX 65). Again, the split does not wind around, but it did increase the black population percentage of District 71. The district then cuts east across Butler-Lacava-Mt. Sterling Voting District, avoiding the majority-white town of Butler, (Def. Supp. Ex. 3), and several majority-white blocks along the border in District 65.
Butler-Lacava-Mt. Sterling Voting District Precinct in Act 602
[[Image here]]
(APSX 62). The split put a 92 percent black population in District 71 and a 33 percent black population in District 65. (Def. Supp. Ex. 405 at 595, 575). From there, the district heads north to split Crossroads-Intersection-Halsell Voting District precinct.
Crossroads-Intersection-Halsell Voting District Precinct in Act 602
*1270[[Image here]]
(APSX 63); This split put 581 people in District 71, 81 percent of them black, and 15 people in District 65, none of them black. (Def. Supp. Ex. 405). From there, the district splits no more precincts in the county. But the design in Choctaw County exhibits a pattern of racial sorting. It placed 3,461 people in District 71, 81 percent of whom were black, (Def. Ex. 405 at 595), and excluded 3,719 in precinct splits, only 34 percent of whom were black.
The percentage of black population in District 71 provides further evidence that race predominated. All of the alternative plans had a lower percentage.
2010 Census Total Black Population Percentages in District 71 Under Various Plans
[[Image here]]
Although not dispositive, this disparity is significant.
We are convinced by the exhibits about Choctaw County and the court-ordered alternative plans that race predominated in the design of District 71. In the enacted plan, District 71 is 67 percent black, but in the court-ordered alternative plans it would be 60 percent black. District 71 did not previously contain Choctaw County, and the Democratic Conference plaintiffs drew District 71 without it. District 71 includes 3,461 people from Choctaw County, 81 percent of them black, (Def. Ex. 405 at 595), and District 65 includes 10,398 people from Choctaw County, only 31 percent of them black. (Id. at 576). In Choctaw County, District 71 follows irregular *1271lines with four split precincts out of six overall. In those four split precincts, the drafters included 1,791 people, 86 percent of them black, (id. at 594-95), and excluded 1,254 people, 34 percent of them black, (id. at 575-76). Based on this evidence about Choctaw County and the alternative plans, we find that race predominated in the design of District 71.
We next conclude that District 71 does not satisfy strict scrutiny. Alabama cannot rely on the comments of Senator Sanders and Representative Jackson because District 71 exceeded the 62 to 65 percent range by nearly two points. The state provided no evidence that a higher percentage was necessary to comply with either section 2 or section 5 in District 71. We must enjoin the use of House District 71 in future elections.
We do not find that race predominated in House District 72. The black population increased from 60.12 to 64.60 percent, (Def. Ex. 406 at 657; Def. Ex. 403 at 422), and the district suffered from the same pressures as the other districts in this area. It was severely underpopulated and surrounded almost entirely by other severely underpopulated districts. The drafters expanded west into Greene and Sumter Counties, but gave most of the population of Marengo County to Districts 65 and 68. District 72 also gave part of its Perry County population to District 67, which needed a small number of people to reach the ±1% population deviation.
District 72 is about as compact and regular as it was under the 2001 plan, and it maintains the same core.
[[Image here]]
(Ala. Reapportionment Office, supra). The edges of the district reach out just far enough to take in parts of several small cities: Eutaw in the northwest, Livingston in the southwest, Marion in the southeast, Brent in the northeast, and Centerville in the northeast. (U.S. Census Bureau, State Legislative District Reference Map: State House District 72 (Alabama), http://www 2.census.gov/geo/maps/dcl0map/SLD_Ref Map/lower/st01_al/sldl01072/DC10SLDL 01072_001.pdf).
Both plaintiffs eliminated at least one county in their alternative plans, but this difference is not conclusive evidence that race predominated. The Democratic Conference plaintiffs eliminated two partial counties, Greene and Sumter.
House District 72 in Democratic Conference Plan A and Act 602
*1272[[Image here]]
(Doc. 287-18 at 7). District 72 in Plan A has a black population percentage of 54.0 percent (Doc. 296-1 at 2), 6.2 points lower than the 2010 Census population under the 2001 lines and 10.6 points lower than the 2010 Census population under the 2012 lines. (Dei. Ex. 406 at 657; Def. Ex. 403 at 422). The Black Caucus plaintiffs eliminated only Sumter County,
House District 72 in Black Caucus 1% Plan and Act 602
[[Image here]]
*1273(APSX 522). The black population percentage in District 72 in the 1% Plan is 60.88 percent (Doc. 296-1 at 2), which more closely met the previous percentage of 60.12 percent.
The additional counties in the enacted plan are not evidence that race predominated because the black population in both of those counties was lower in District 72 than in the county as a whole:
[[Image here]]
(Def. Ex. 405 at 599-601; U.S. Census Bureau, American FactFinder, supra). Whatever the drafters might have been doing by adding those counties, they were not attempting to increase the black population by adding portions of Greene and Sumter Counties, which is all that the plaintiffs have argued. And unlike in some other districts, the overall picture of county splits does not suggest that race predominated: the black population percentages in the Bibb County and Marengo County portions of District 72 are significantly higher than in those counties as a whole, but the portions in Greene and Sumter Counties are lower and the portion in Perry County is less than one percentage point away from the county as a whole. The Democratic Conference plaintiffs initially argued on remand that each new county split brought a majority-black population into the district, (Doc. 258 at 73), but when they drew a plan attempting to comply with the Committee guidelines, they also put a majority-black population from Marengo County in the district while averring that race did not predominate in their plan.
• The Democratic Conference plaintiffs also argue for the first time in their reply brief that the splits of the cities of Brent and Centreville prove that race predominated. This argument is not persuasive because of the history of District 72. We recognize that these cities border each other, the legislature split both of them, and the black population from each city is higher in District 72 than in District 49. But the Democratic Conference plaintiffs neglect to mention — and Alabama did not have a chance to point out — that the 2001 plan also split both cities. The plaintiffs do not argue that the old district is a racial gerrymander, so the mere fact that these cities were split is not evidence that race predominated.
The shape of the splits — which the Democratic Conference plaintiffs neglect to discuss — also is not evidence that race predominated. As an initial matter, the drafters preserved the basic shape from 2001 by taking population from the south in both cities. It is not unusual that the shape of the split would change in a district that needed to grow, and we do not find the current shape to be bizarre. The following maps compare the old and new lines, with the district line in red, the city limits in gray, and highways in yellow and blue:
*1274[[Image here]]
(Ala. Reapportionment Office, supra). The small square protrusion to the north in Centreville was present in the 2001 lines. In Brent (the western city), the only new irregularity is a peak that follows one highway and a valley that follows another highway. In Centreville (the eastern city), the district takes a coherent portion of this L-shaped city. The following map from the Census Bureau makes both of these choices clear:
Census Bureau Map of Centreville and Brent
[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 72 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01072/DC10SLDL01072_ 001.pdf).
The Democratic Conference plaintiffs rely on the fact that District 72 takes a higher black population percentage from both cities than District 49 does, but we do not find any evidence that race predominated. In the enacted plan, District 72 has 3,903 people from Brent, 64 percent of them black, and District 49 has 1,044 people from Brent, 16 percent of them black. (Def. Supp. Ex. 3). But in 2001 — the parties have not provided us with data on the splits of cities in 2010 — District 72 had 3,631 people from Brent, 53 percent of *1275them black, and District 49 had 663 people from Brent, 13 percent of them black. (Def. Ex. 411 at 810, 822). The difference'is similar (48 percent in 2012 and 40 percent in 2001), and the black population in the city increased between the 2000 Census and 2010 Census from 47 percent to 53 percent. The numbers in Centreville may seem more suspicious at first: 128 people in District 72 in 2001 (0 percent black), 2,338 people, in District 49 in 2001 (19 percent black), 749 people in District 72 in Act 602 (47 percent black), and 2,029 people in District 49 in Act 602 (15 percent black). (Def. Supp. Ex. 3; Def. Ex. 411 at 811, 822). But Centreville changed between 2000 and 2010: it grew from 2,446 people in 2000 to 2,778 people in 2010, and it went from 18 percent black in 2001 to 24 percent black in 2010. (Def. Supp. Ex. 3; Def. Ex. 411 at 811, 822). The drafters put a black population percentage from Centre-ville in District 49 that is almost the same as the overall percentage in the city. On this record, the plaintiffs’ last-minute argument about Brent and Centreville fails to convince us that race predominated.
Of course, the plaintiffs could still prove that race predominated because the drafters split precincts on the basis of race, but they fail to do so. In Bibb County, the legislature split Brent City Hall-13 precinct.
Brent City Hall-13 Precinct in Act 602
[[Image here]]
*1276(APSX 54). The split put 67 people into District 49, 6 percent of them black, and 3,569 people in District 72, 68 percent of them black. (Def. Ex. 405 at 543, 599). If we include those 67 people in the district, the black population percentage remains almost unchanged at 64.51 percent, instead of 64.60. We find that this split does not prove that race predominated.
The remaining precinct splits in Bibb County are not at all suspicious. First, in Eoline Fire Dept-3 precinct, the legislature put 408 people in District 72, 2 percent of them black, and 90 people in District 49, none of them black. (Id. at 542, 599). Second, in Rock Building-5, the legislature put 436 people in District 72, 19 percent of them black, and 1,445 people in District 49, 8 percent of them black. (Id.). There is nothing visually suspicious about the split, which smooths out the district line.
Rock Building-5 Precinct in Act 602
[[Image here]]
(APSX 59). Third, in Brent National Guard Armory precinct, the legislature put 697 people in District 72, 41 percent of them black, and 2,436 people in District 49, 17 percent of them black. (Def. Ex. 405 at 542, 599). But again, the split is not visually suspicious.
Brent National Guard Armory Precinct in Act 602
*1277[[Image here]]
*1278(APSX 59). Fourth, in Eoline Fire Dept.-12 precinct, the legislature put 189 people in District 72, 46 percent of them black, and 754 people in District 49, 7 percent of them black. But there is nothing visually suspicious about this split either.
Eoline Fire Dept.-12 Precinct in Act 602
[[Image here]]
(APSX 56).
Finally, the black population percentage in District 72 does not prove that- race predominated. District 72 exceeds its target by 4.4 points, but three plans, including the 1% Plan, also exceed the target:
2010 Census Total Black Population Percentages in District 72 Various Plans
[[Image here]]
*1279Whether the theory is that the legislature sought to meet the previous percentage as closely as possible or that the legislature sought to meet or exceed the previous percentage, the alternative plans are yet again proof that demographics and legitimate districting criteria dictated the design of the district. We find that race did not predominate in the design of House District 72.
k. House Districts 76, 77, and 78 (Montgomery)
We find that race did not predominate over traditional districting criteria in the design of House Districts 76 and 78, but that race predominated in the design of House District 77. All three districts, located in the city of Montgomery, were underpopulated in 2010, two of them severely:
[[Image here]]
(Def. Ex. 406 at 657). To solve the underpopulation problem, Representative McClammy, a black Democrat and the incumbent in House District 76, proposed that the drafters move District 73 — a district in the Montgomery area that had been majority-white but had become plurality-black, and was represented by a white Democrat — to Shelby County. (Doc. 217 at 134-35, 229-30). This move solved both the underpopulation in Montgomery and the overpopulation in Shelby County. (Id. at 133-34).
The plaintiffs argue that the decision to move District 73 proves that race predominated over traditional districting criteria, but they made the same decision in the New Black Caucus Plan and Democratic Conference Plan A. Even if the plaintiffs had not made the same decision, their challenge would still fail. We previously found that “Hinaman moved House. District 73 ... to avoid retrogression of the majority-black House districts in Montgomery County.” (Doc. 203 at 33). He also moved the district to solve the underpopulation of the districts in Montgomery County and the overpopulation of the districts in Shelby County. Hinaman testified that Shelby County “was the fastest growing county in the state,” that “every district whole or part that was in Shelby County was overpopulated,” and that House District 41 “was dramatically overpopulated.” (Doc. 217 at 134). “[I]t made sense to move [House District 73] to a much faster growing area.” (Id.) The Shelby County House districts were all overpopulated by as much as 60 percent. (Def. Ex. 411 (Districts 41^13, 48-50)). Moving District 73 into Shelby County was a change from the previous plan, but it was reasonably calculated to be less disruptive than evening out every district in Shelby County. The move also solved the problem of the underpopulation of the majority-black districts in Montgomery, and the Republican majority made a legitimate partisan choice to move a Democratic district instead of a Republican one.
The 2012 districts bear little resemblance to the size or shape of the 2001 districts, although the drafters kept District 77 to the north of District 76 and District 78 to the west of both:
*1280[[Image here]]
(Ala. Reapportionment Office, supra). The districts are not noticeably less regular or compact than in the 2001 plan.
The Democratic Conference plaintiffs drew a different configuration, but we cannot say that it proves that race predominated. In Plan A District 76, the only overlap with the enacted district is the incumbent’s home precinct.
House District 76 in Democratic Conference Plan A and Act 602
[[Image here]]
*1281(Doc. 287-16 at 4). In District 77, the Democratic Conference plaintiffs drew a district farther to the south.
House District 77 in Conference Plan A and Act 602
[[Image here]]
(Id. at 5). And in District 78, they drew a district farther to the west and south.
House District 78 in Democratic Conference Plan A and Act
[[Image here]]
*1282(Id. at 6).
The Black Caucus 1% Plan suffers from the same shortcomings. The Black Caucus plaintiffs moved District 76 southward.
House District 76 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 523). In, District 77, they made choices similar to the Democratic Conference plaintiffs.
House District 77 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 524 (Black Caucus District 77 in pink)). And in District 78, they again made choices similar to the Democratic Conference plaintiffs.
House District 78 in Black Caucus 1% Plan and Act 602
*1283[[Image here]]
(APSX 525 (Black Caucus District 78 in gray)). The plaintiffs never explain why their choices were required by traditional districting criteria other than avoiding precinct splits, an argument we reject later.
None of the three enacted districts came particularly close to matching the prior black population percentage, which undermines the plaintiffs’ arguments about mechanical targets in these districts.
2010 Census Total Black Population Percentages Under Various Plans
[[Image here]]
Moreover, as with House Districts 58 and 59 in Birmingham, the drafters put “too much” black population in one district and “too little” in adjoining ones, instead of smoothing out the racial percentages to meet the supposed targets. It is implausible that they were mechanically pursuing a target here.
There is also nothing abnormal about the high percentage of black population in these districts. In fact, several of the alternative plans surpass the black population percentages in the enacted districts:
2010 Census Total Black Population Percentages of Majority-Black Districts in Montgomery County Under Various Plans
*1284[[Image here]]
* The Knight plan moves House District 76 out of Montgomery County and leaves House District 73 in the county, so this table lists the black population percentage for Knight Plan House District 73.
As the table shows, the earlier alternative plans had black populations as high as 83.6 percent. When the plaintiffs drew court-ordered plans attempting to comply with the Committee guidelines, they again drew districts with strong black majorities. In the 1% Plan, District 76 is 9.8 points lower, but District 77 is only 1.61 points lower and District 78 is 3.23 points lower. In Plan A, District 76 is 14.49 points lower, but District 77 is 3.54 points lower and District 78 is 7.51 points higher. No plaintiff drew a map with less than a black supermajority in any of the three districts, at least without creating a 77 percent black population percentage elsewhere in the county, as Plan A would do. And to the extent that the plaintiffs’ alternative districts had a lower black population percentage, they failed to explain why their choices were required by traditional dis-tricting criteria.
We find no suspicious precinct splits in District 76. In 1A Cloverdale Community Center Voting District, IE Aldersgate United Methodist Church Voting District, 2B Beulah Baptist Church Voting District, and 2H Harrison Elementary School Voting District precincts, the drafters split the precinct with one or more other majority-black districts. (APSX 220, 222, 224, 228; Def. Ex. 405 at 611, 613). There is no apparent pattern in the splits, and the plaintiffs identify none. We also observe that the Democratic Conference plaintiffs split 1A Cloverdale Community Center precinct in their Plan A. (Doc. 296-6 at 3). In IF Al. Industrial Development Training precinct, the drafters extended a straight line from the two adjacent precincts and moved only four people into District 69.
IF AL Industrial Development Training Precinct in Act 602
*1285[[Image here]]
(APSX 223). In 5A Seth Johnson Elementary School Voting District precinct, the drafters split off unpopulated areas. (APSX 238). In 5B Snowdoun Womens Club Voting District precinct, the drafters drew a straight line that put a majority-white block with three people and several zero-population blocks in District 76, keeping those areas connected to populated blocks in adjacent precincts.
5B Snowdoun Womens Club Voting District Precinct in Act 602
*1286[[Image here]]
(APSX 239). In 5E Fitzpatrick Elementary School Voting District precinct, the drafters placed 2,784 people in neighboring District 69, 82 percent of them black, and 6,568 people in District 76, 65 percent of them black. (Def. Ex. 405 at 593, 608). The Democratic Conference plaintiffs also split 5E Fitzpatrick Elementary School and put a 75 percent black population in District 77 and a 54 percent black population in District 90. (Doc. 296-6 at 3).
5E Fitzpatrick Elementary School Voting District Precinct in Act 602
*1287[[Image here]]
*1288(APSX 241). In 5M Bell Road YMCA Voting District precinct, the drafters drew smooth lines that deviated only to pick up majority-white blocks.
5M Bell Road YMCA Voting District Precinct in Act 602
[[Image here]]
*1289(APSX 243). And in 5N Peter Cramp School Voting District precinct, the drafters drew a straight line across the middle of the district, splitting majority-black blocks on both sides.
5N Peter Crump School Voting District Precinct in Act 602
[[Image here]]
(APSX 244). We find that race did not predominate in the design of District 76.
We find no suspicious precinct splits in District 78 either. The plaintiffs again offer no explanation about racial predominance in the splits between and among majority-black districts, and we see no evidence in the maps and statistics. As for the other splits, 2D Montgomery Boys Club Voting District precinct the drafters followed a smooth line of zero-population blocks.
2D Montgomery Boys Club Voting District Precinct in Act 602
*1290[[Image here]]
(APSX 225). In 21 Southlawn Elementary School Voting District, 2F Fire Station No. 14 Voting District, and 2G Hayneville Road Community Center Voting District precincts, the drafters split predominantly majority-black blocks with no apparent pattern.
21 Southlawn Elementary School Voting District Precinct
*1291[[Image here]]
*1292(APSX 229).
2F Fire Station No. 14 Voting District Precinct in Act 602
[[Image here]]
*1293(APSX 226).
2G Hayneville Road Community Center Voting District Precinct in Act 602
[[Image here]]
(APSX 227). In 5K Lakeview Baptist Church Voting Center precinct, they split predominantly majority-white blocks with no apparent pattern.
5K Lakeview Baptist Church Voting Center Precinct in Act 602
*1294[[Image here]]
(APSX 242). In 3F Goodwyn Community Center Voting District, they split the precinct along a railroad near unpopulated areas, both legitimate explanations for the split.
3F Goodwyn Community Center Voting District in Act 602
*1295[[Image here]]
(APSX 231).
Census Bureau Map in the Vicinity of 3F Goodwyn Community Center Voting District Precinct
[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 7Jf (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01078/DC10SLDL01078_ 001.pdf). And in 4K Chisholm Community Center Voting District precinct, the drafters put ten people in District 74. If we add those people to the district, the black population changes by about one-hundredth of a point, decreasing to 69.98. (APSX 235). *1296We find that race did not predominate in the design of District 78.
In District 77, most of the precinct splits provide no evidence that race predominated. The plaintiffs offer no argument about — and we can find no sign of — racial predominance in precinct splits between District 77 and majority-black districts. (APSX 220, 222, 224, 232, 233, 234, 236; Def. Ex. 405 at 607, 609-13). The split of 4N Highland Avenue Baptist Church precinct extended a straight line from the precinct to the north, smoothing out the district shape and putting a majority-white block of six people in District 74.
4N Highland Avenue Baptist Church Precinct in Act 602
[[Image here]]
(APSX 237). There is no evidence of racial predominance in these precincts.
But In IB Vaughn Park Church of Christ Voting District precinct, the drafters drew an irregular line that put almost all of the majority-black blocks in District 77.
IB Vaughn Park Church of Christ Voting District Precinct in Act 602
*1297[[Image here]]
(APSX 221). The split placed 6,719 people in District 77, 57 percent of them black, and 3,152 people in District 74, 16 percent of them black. (Def. Ex. 405 at 603, 609). This precinct is split along almost identical lines in Senate District 26, (see APSX 354), and we find that a similar phenomenon occurred in both districts: the drafters split this precinct along racial lines to place a higher percentage of black population in District 77 and keep the overall black population percentage in majority-white District 74 under 25 percent. When Alabama drew a hypothetical district with no split precincts for purposes of litigation, the black population percentage in District 77 increased by only 0.9 points. (Doc. 263-2 at 2). But when we unsplit the precinct according to our method that tracks the plaintiffs’ arguments, the black population percentage in District 77 drops significantly to 63.8 percent instead of 67.04 percent. (Def. Ex. 405 at 609-11). On the basis of all of the evidence about this split precinct, we find that race predominated in the design of District 77.
We further conclude that District 77 fails strict scrutiny. Alabama makes no arguments about why District 77 in particular survives strict scrutiny, and the black population percentage of 67.04 percent exceeds the range recommended by Senator Sanders and Representative Jackson. We must enjoin the use of District 77 in future elections.
1. House Districts 82, 83, 84, and 85 (East Black Belt)
We find that race did not predominate over traditional districting criteria in the design of House Districts 83 and 84, but that race did predominate in the design of House Districts 82 and 85. Each of these districts are located in the eastern portion of the Black Belt, and each was underpopulated:
*1298[[Image here]]
(Def. Ex. 406 at 658).
House District 82, which saw its black population increase from 57.18 to 62.14 percent, (Doc. 263-2 at 3), includes all of Macon County in the old and new plans. Macon County is 83 percent black. (U.S. Census Bureau, American FactFinder, supra ). In the 2001 plan, the district extended south into Bullock County, which was over 70 percent black in the 2010 Census, (id.), and northeast into Lee County. The district now extends into Tallapoosa County instead of Bullock County; it gave up its Bullock County population to District 84, which was also underpopulated. (Def. Ex. 406 at 658). District 82 extends further into Lee County in the east, and it extends north of U.S. Route 280 into Tallapoosa County to take in population from Camp Hill and Dadeville. (U.S. Census Bureau, Alabama State Legislative District Reference Map: State House District 82 (Alabama ), http://www2.census.gov/geo/maps/ dcl0map/SLD_RefMap/lower/st01_al/sldl 01082/DC10SLDL01082_001.pdf).
[[Image here]]
(Ala. Reapportionment OfGce, supra).
The black population percentage of District 82 does not provide much evidence one way or another. The drafters exceeded their supposed target by nearly five percentage points. (Def. Ex. 403 at 423; Def. Ex. 406 at 658). Both Plan A and the 1% Plan come closer to matching the previous percentage than Act 602 did. (Doc. 296-1 at 3). Even assuming that exceeding the target is evidence that race predominated, the enacted district is still similar to the alternative plans. The enacted district has a total black population of 62 percent, (Def. Ex. 403 at 423); the Black Caucus plans on remand had similar populations that were 66 percent black, (APSX 36 at 2), and 61 percent black, (Doc. 296-1 at 3); the McClammy Plan had a black population percentage of 61.14 percent (Common Ex. 45 at 4); and the Democratic Conference Plan A and Reed-Buskey Plan had somewhat lower populations that were 57.9 percent black and 57.22 percent black, respectively, (Doc. 296-1 at 3; Common Ex. 42 at 4).
The Democratic Conference plaintiffs eliminated one of the antennae in Tallapoo-*1299sa County and smoothed out the border in Lee County:
House District 82 in Democratic Conference Plan A and Act 602
[[Image here]]
(Doc. 287-18 at 8).
The 1% Plan is less probative because it is in a different location and because the plaintiffs do not explain why their decision to remove Lee and Tallapoosa Counties was required by traditional districting criteria. The following map shows District 82 in the 1% Plan in gray and District 82 in Act 602 with a purple line marked by “82” in a purple circle:
House District 82 in Black Caucus 1% Plan and Act 602
*1300[[Image here]]
(APSX 526). Because it crosses the same number of county boundaries, has a roughly similar percentage of black population, and does not prove that it is possible to draw a better shape in Lee County, the 1% Plan does not provide much evidence that race predominated in the design of District 82.
The racial pattern of the county splits provides strong evidence that race predominated. The drafters used irregular lines to put 19,043 people from Lee County in District 82, 37 percent of whom were black; 28,644 people in neighboring District 38, 13 percent of whom were black; and 45,972 in neighboring District 79, 12 percent of whom were black. (Def. Ex. 405 at 619, 527, 614). The drafters also used antennae to put 5,363 people from Talla-poosa County in District 82, 68 percent of whom were black, and 36,253 people in neighboring District 81, only 21 percent of whom were black. (Id. at 620, 618).
The plaintiffs also are correct that the precinct splits in Tallapoosa County and the city of Auburn establish that race predominated. The drafters placed a concentrated black population in District 82 while avoiding areas with white population. (APSX 136, 268). The remainder of the district is smooth and compact, but Alabama has offered no reason why the drafters extended two thin antennae that pick up black population and split precincts to do so.
We begin with Auburn precinct because of its size. The drafters pushed District 82 eastward to reach the major areas of black population.
Auburn Precinct in Act 602
*1301[[Image here]]
(APSX 136). They put 9,149 people in District 82, 42 percent of whom were black, and 44,728 people in majority-white District 79, 12 percent of whom were black. (Def. Ex. 405 at 619, 614). The Democratic Conference plaintiffs also split Auburn precinct — which they had to do because this precinct alone would have exceeded the ±1% deviation — but they smoothed out the shape. (Doc. 297-9 at 119, 123).
In Dadeville National Guard Armory precinct, the drafters drew a seesaw that put most of the majority-black blocks in District 82:
Dadeville National Guard Armory Precinct in Act 602
*1302[[Image here]]
(APSX 268). The split put a population that was 61 percent black in District 82 and a population that was 15 percent black in District 81. (Def. Ex. 405 at 619, 616).
Finally, in Mary’s Cross Road Voting House precinct, the drafters drew a reasonable line that put a higher black population percentage in District 82 (65 percent) than in District 81 (45 percent). (Id. at 619, 618). But even without this precinct split, we find that race predominated in the design of District 82.
Alabama argues that its unsplitting exercise increased the black population percentage, (Doc. 296-6 at 24), but this exercise is too flawed to be conclusive. The fact that Alabama could have drawn a district with a slightly higher black population percentage and no precinct splits does not mean that a significant number of people were not assigned to the enacted district on the basis of race. Alabama has failed to explain these suspicious precinct splits, and we find that race predominated in the design of District 82.
We conclude that Alabama failed to establish that District 82 satisfies strict scrutiny. The black population percentage of the district increased by almost five points, and Alabama has not offered any district-specific evidence that such an increase was required by section 2 or section 5. Alabama makes no district-specific arguments, and we cannot assume that Senator Sanders and Representative Jackson have the same expertise about the eastern Black Belt that they have about their own districts in the western Black Belt. District 82 is also more urban (over 40 percent) than either Senate District 23 or House District 68, and it contains significant population from the large city of Auburn. (Def. Supp. Ex. 3;' U.S. Census Bureau, Urbanized Areas and Urban Clusters: 2010, http://www2.census.gov/geo/pdfs/maps-data/maps/reference/2010UAUC_List.pdf).
*1303It is not even obvious that the black populations of Auburn, Camp Hill, and Dadeville, along with the core of Macon County, form a “geographically compact” minority population for purposes of section 2. The first Gingles factor is that the “racial group is sufficiently large and geographically compact to constitute a majority in a single-member district.” Bossier Parish Sch. Bd., 520 U.S. at 479, 117 S.Ct. 1491. But as this map shows, they are some distance apart:
Zoom of Dadeville, Camp Hill, and Auburn in District 82
[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 82 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01082/DC10SLDL01082_ 001.pdf). The state has failed to establish that House District 82 survives strict scrutiny, and we must enjoin the use of this district in future elections.
We find that race did not predominate in the design of House District 83. The district, which increased in black population percentage from 56.92 to 57.52 percent, (Doc. 296-2 at 3), is irregularly shaped. But as the Democratic Conference plaintiffs admit, it was irregular under the 2001 district lines as well, (Doc. 258 at 81):
*1304[[Image here]]
(Ala. Reapportionment Office, supra). It takes up parts of Lee and Russell Counties under both the 2001 and 2012 lines, but it shifted north in the new plan so that District 84, also underpopulated in 2010, could take more of Russell County.
The dissent argues that drafters extended the “north-central limb [of the district].... further north into Lee County” to include majority black census blocks, (Dissent at 1401). But the Democratic Conference ⅛ A makes a similar incursion:
House District 83 in Democratic Conference Plan A and Act 602
[[Image here]]
(Doc. 287-18 at 9). Because the Democratic Conference plaintiffs argue that their district is race-neutral and because of the similarity of the enacted district to their alternative (and the old district), we do not find that the shape is evidence of racial predominance.
This similarity also alleviates any concern we otherwise might have about the number of people moved (18,646) relative to the underpopulation (4,482). The Democratic Conference plaintiffs do not tell us how many people they moved in the creation of their alternative district, but the similarity of the two districts leads us to *1305infer that they moved a similar number of people in their assertedly race-neutral district.
Nor is the black population percentage in District 83 suspicious. All three of the alternative plans that draw a majority-black District 83, including Plan A, have black population percentages that are similar to the enacted district. The McClammy Plan meets and exceeds the target, and the Reed-Buskey Plan and Plan A come within 2.03 points of the plan as passed.
2010 Census Total Black Population Percentages in District 83 Under Various Plans
[[Image here]]
These similarities are once again evidence that the black population percentage in District 83 was the result of legitimate districting choices, not a racial target.
The plaintiffs argue that precincts were split along racial lines, but they again fail to prove that race was the predominant reason. In Beuaregard School precinct, Smiths Station Sr. Center Voting District precinct, Crawford Fire Dept, precinct, and Austin Sumbry Park Voting District precinct, the drafters put roughly equal black population percentages within and without the district. (Def. Supp. Ex. 405 at 526, 614-15, 619-20). The legislature split Seale Courthouse Voting District precinct between two majority-black districts, (APSX 256), and the plaintiffs fail to explain how this split is evidence of racial predominance.
In Opelika B Voting District precinct, the drafters used an irregular line to put 18,201 people in District 83, 59 percent of them black, and 11,738 people in other districts, 12 percent of them black. (Doc. 405 at 78,165,171-72).
Opelika B Voting District Precinct in Act 602
*1306[[Image here]]
(APSX 140). We agree with our dissenting colleague that the precinct split creates an “odd shape,” (Dissent at 1399), but the Democratic Conference plaintiffs, who insist that race did not predominate in Plan A, drew essentially the same line that put a 58 percent black population in District 83. (Doc. 297-9 at 124).
Close-Up of Opelika B Voting District Precinct in ADC Plan A and Act 602
*1307[[Image here]]
(Doc. 287-18 at 9). Because of the similarity between Plan A and Act 602 in this precinct, we find that this precinct split is not evidence that race predominated.
In Lee County Snacks Voting District precinct, the legislature and the Democratic Conference plaintiffs made similar choices. The first map shows the split in the enacted district, and the second map compares the enacted district (orange and red) with the Plan A district (orange and yellow).
Lee County Snacks Voting District Precinct in Act 602
*1308[[Image here]]
(APSX 138).
Lee County Snacks Voting District Precinct in Plan A and Act 602
[[Image here]]
*1309(Doc. 287-18 at 9). The legislature put 1,998 people in District 83, 32 percent of them black, and 107 people in District 38, 3 percent of them black. (Def. Ex. 405 at 526, 620). The Democratic Conference plaintiffs put 1,706 people in District 83, an even higher 37 percent of them black. (Doc. 297-9 at 124). Because the Democratic Conference plaintiffs insist that race did not predominate in Plan A, we find that this split is not evidence that race predominated in the enacted district.
The legislature and the Democratic Conference plaintiffs also split National Guard Armory Voting District along similar lines. The legislature put 4,151 people in District 83, 66 percent of them black. (Def. Ex. 405 at 621). The Democratic Conference plaintiffs put 4,367 people in District 83 in Plan A, 65 percent of them black. (Doc. 297-9 at 125).
National Guard Armory Voting District Precinct in Act 602
[[Image here]]
*1310(APSX 255).
National Guard Armory Voting District Precinct in Plan A and Act 602
[[Image here]]
(Doc. 287-18 at 9). The Black Caucus plaintiffs split this precinct along different lines and put it in District 84 instead of District 83, but they put an even higher percentage of black population — 74.4 percent — in that majority-black district. (APSX 633 at 7). Because the plaintiffs maintain that race did not predominate in their alternative plans, we find that this precinct split does not provide evidence that race predominated in the enacted plan.
In CVCC Voting District precinct, the legislature made an odd choice to put only 27 people in another district, but we cannot say that this split proves that race predominated. The drafters put 2,665 people in District 83, 64 percent of them black, and a mere 27 people in District 80, 33 percent of them black. (Def. Ex. 405 at 615, 621). If we put the 27 people in District 83, the black population percentage in District 83 from this precinct would still be 64 percent. We cannot say that this negligible number of people constitutes proof that race predominated in the placement of a significant number of people.
Two precincts are more suspicious, but they do not account for the assignment of a significant number of people. In Old Salem School precinct, the drafters drew a line with no obvious explanation that put 338 people in District 83, 42 percent of them black, and 1,602 people in majority-white District 38, 9 percent of them black. (Def. Ex. 405 at 526, 620).
Old Salem School Precinct in Act 602
*1311[[Image here]]
*1312(APSX 139).. In Ladonia Fire Dept precinct, the drafters drew a line that put no majority-white blocks in District 83.
Ladonia Fire Dept Precinct in Act 602
[[Image here]]
(APSX 254). But the black people assigned to District 83. in this split account for less than half of a percent of the total population of the district, and if we remove both precincts from the district entirely, the black population percentage remains almost unchanged at 57.61 percent. Our dissenting colleague argues that the precinct splits demonstrate racial predominance because the splits allowed the drafters to place “a significant white population ‘without’ [House District] 83,” allowing the drafters to hit their alleged target. (Dissent at 1399 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475)). But, as noted, had the drafters not split the precincts the black population would have barely decreased and the white population would have still been “without” the district. We find that race did not predominate in the design of District 83.
We also find that race did not predominate in the design of House District 84. District 84, which increased from 50.67 percent to 52.34 percent black, is a compact district that includes all of Bullock and Barbour counties, as well as part of Russell County. In their zeal to attack as many districts as possible, the plaintiffs appear to have challenged House District 84 by mistake. Indeed, the Black Caucus plaintiffs abandoned their challenge to the district in their final reply brief. (Doc. 300-1 at 107). This district meets all of the plaintiffs’ criteria for a district, and the 2012 design made it more compact and split fewer counties:
*1313[[Image here]]
(Alabama Policymaker’s Dashboard, supra).
The Democratic Conference plaintiffs drew an alternative District 84 that was identical to the enacted District 84, except that it eliminated the sole precinct split in the enacted District 84.
House District 84 in Democratic Conference Plan A and Act 602
[[Image here]]
(Doc. 287-18 at 10). That split put a total of 28 people into District 84. (Def. Ex. 405 at 622). If we added all of that precinct to District 84 or removed it all, it would have an effect of less than 0.2 points on the black population percentage of District 84. This miniscule effect is all the more negligible because District 84 already exceeds the previous percentage by 1.7 points. (Def. Ex. 406 at 658; Def. Ex. 403 at 423). The Democratic Conference plaintiffs’ challenge to District 84 is frivolous.
We find that race predominated in the design of House District 85, which increased in total black population percentage from 47.94 to 50.08 percent. District 85 is a new majority-black district by total population, but it is not majority-black by voting-age population. (Def. Ex. 406 at 658; Def. Ex. 405 at 623; Doc. 35-2 at 3).
⅛ ⅛ basic concept, District 85 has changed very little. It includes all of Henry County, and it extends south into the Houston County part of Dothan. But the extension into Houston County has changed significantly. Under the 2001 district lines, the extension was a rounded, sensible protrusion. Now, the extension is a bizarre bootspur that manages to pick up black population while avoiding white population at almost every turn:
*1314[[Image here]]
(Ala. Reapportionment Office, supra). From Dothan, the drafters put a population that was 63 percent black in District 85 and one that was 12 percent black in Districts 86, 87, and 93. (Def. Supp. Ex. 3 at 63-65, 69). The district also split nine precincts in Houston County. Alabama concedes that in its own exercise, unsplit-ting the split precincts decreased the black population percentage by over 14 points, taking it well below 50 percent black (Doc. 263-3 at 2).
We also find evidence in Kinsey, Wire-grass Park, and Westgate Recreational Center precincts that race predominated in the design of the district, although we observe that the Democratic Conference plaintiffs also split these same precincts. (Doc. 296-6 at 3). First, in Kinsey precinct, the drafters drew three separate incursions that reached all of the majority-black blocks in the precinct and then stopped.
Kinsey Precinct in Act 602
[[Image here]]
(APSX 85). The split put 1,352 people in District 85, 68 percent of whom were black, and 886 people in majority-white District 86, 10 percent of whom were *1315black. (Def. Ex. 405 at 622-23). Second, in Wiregrass Park precinct, the drafters drew three fingers from the north that captured all but one majority-black block, which was isolated much farther south in the precinct,
Wiregrass Park Precinct in Act 602
[[Image here]]
(APSX 90). The split put 7,311 people in District 85, 57 percent of whom were black, and 3,293 people in District 86, only 15 percent of whom were black. (Def. Ex. *1316405 at 628-24). Third, in Westgate Recreational Center precinct, the drafters put a majority-black block of 33 people in District 85.
Westgate Recreational Center Precinct in Act 602
[[Image here]]
(APSX 89). These three precincts together are strong evidence that race predominated in the design of the district.
We also find weaker evidence of racial predominance in the splits of Johnson Homes and Library precincts, even though they appear to have taken majority-white blocks that were not necessary to reach majority-black blocks. In Johnson Homes precinct, the drafters drew two fingers that reached all of the majority-black blocks in the precinct.
Johnson Homes Precinct in Act 602
*1317[[Image here]]
(APSX 84). The split put 4,888 people in District 85, of whom 91 percent were black, and 139 people in District 86, of whom 6 percent were black. (Def. Ex. 405 at 622, 623). In Library precinct, the drafters zigged and zagged to place every xna-jority-black block in the precinct into District 85.
Library Precinct in Act 602
*1318[[Image here]]
(APSX 86). The split put 4,154 people in District 85, 68 percent of whom were black, and 3,666 people in District 86, only 7 percent of whom were black. (Def. Ex. 405 at 623, 624). Based on the maps and statistics, we find that the only sensible understanding of District 85 is that it was designed based on race, and Alabama has not offered any other explanation.
Both plaintiffs also drew a similar plurality-black District 85, but with a smoother protrusion into Houston County. The following map shows the version in Plan A:
House District 85 in Democratic Conference Plan A and Act 602
*1319[[Image here]]
(Docs. 294-2 at 2). The population of this district is 48.3 percent black by total population and 45.6 percent black by voting-age population. (Doc. 296-2 at 3). The Black Caucus plaintiffs also made the extension into Houston County more regular:
House District 85 in Black Caucus 1% Plan and Act 602
[[Image here]]
*1320(APSX 529). Their district is 49.0 percent black by total population and 46.3 percent black by voting-age population. (Doc. 296-2 at 3). Both sets of plaintiffs split fewer precincts and drew a district with a lower black population percentage, confirming our finding that race predominated in the design of District 85.
District 85 does not satisfy strict scrutiny. The enacted district cannot be narrowly tailored to comply with the Voting Rights Act because the district is not majority-black by voting-age population. Section 2 “provides a cause of action for protected minority groups that can establish, based on the totality of the circumstances, ‘that [their] members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.’ ” Dillard v. Baldwin Cty. Comm’rs, 376 F.3d 1260, 1265 (11th Cir. 2004) (alteration in original) (quoting 52 U.S.C. § 10301(b)). Under section 2, Alabama must avoid diluting the voting strength of a racial minority where “(i) ‘[the racial group] is sufficiently large and geographically compact to constitute a majority in a single-member district’; (ii) the group is ‘politically cohesive’; and (iii) ‘the white majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate.’ ” Bossier Parish Sch. Bd., 520 U.S. at 479, 117 S.Ct. 1491 (alteration in original) (quoting Gingles, 478 U.S. at 50-51, 106 S.Ct. 2752). If the Gingles factors are present in a district, the court looks to whether “the totality of the circumstances supports a finding that the voting scheme is dilutive.” Id. at 479, 117 S.Ct. 1491.
The first Gingles factor was not satisfied here because the minority group in District 85 does not constitute a majority of the voting-age population. In Bartlett v. Strickland, three Justices explained that “[o]nly when a geographically compact group of minority voters could form a majority in a single-member district has the first Gingles requirement been met.” 556 U.S. at 26, 129 S.Ct. 1231 (opinion of Kennedy, J.). The relevant population, they explained, is the “voting-age population.” Id. at 18, 129 S.Ct. 1231. That opinion binds us under Marks v. United States, 430 U.S. 188, 193, 97 S.Ct. 990, 51 L.Ed.2d 260 (1977). Because the black voting-age population percentage in District 85 is 47.23 percent, (Doc. 35-2 at 3), Alabama cannot claim that its district is narrowly tailored to achieve compliance with section 2.
Nor can Alabama claim that District 85 is narrowly tailored to comply with section 5 of the Voting Rights Act. The parties do not dispute that section 5 required Alabama to maintain at least the same number of majority-black districts under the 2012 plan as under the 2001 plan. (See Doc. 263 at 57). The 2001 plan had 27 majority-black House districts, and without District 85, the 2012 plan has 27 majority-black House districts: Districts 19, 32, 52, 53, 54, 55, 56, 57, 58, 59, 60, 67, 68, 69, 70, 71, 72, 76, 77, 78, 82, 83, 84, 97, 98, 99, and 103. Alabama has not explained why section 5 required it to draw an additional majority-black district by total population or an influence district by voting-age population. Nor has it identified any other compelling interest for drawing District 85 the way that it did. We must enjoin the use of District 85 in future elections.
m. House Districts 97, 98, 99, and 103 (Mobile County)
We find that race did not predominate over traditional districting criteria in the design of House Districts 97, 98, or 103, but we find that race did predominate in the design of District 99. Each of these districts is located in or near the city of *1321Mobile, and each was severely underpopulated:
[[Image here]]
(Def. Ex. 406 at 659). In three of the four districts, the drafters missed the previous black population percentage by a significant margin.
2010 Census Total Black Population Percentages Under 2001 Lines and Enacted Plan
[[Image here]]
And the percentages of black population in the alternative plans are mostly similar to the enacted districts, so we infer that demographics and legitimate districting criteria explain the percentages.
2010 Census Total Black Population Percentages Under Various Plans
[[Image here]]
The Democratic Conference plaintiffs asserted on remand that Hinaman began with House District 97 and reached the racial target for that district. (Doc. 258 at 84-85). After he achieved his target in District 97, they reason, he could no longer achieve his racial targets in the other districts. (Id. at 85) But the Democratic Conference plaintiffs fail to prove this version of events. Hinaman’s only testimony at trial about these districts was that the black population percentage vas reduced, and that “[sjometimes there’s no way to avoid it.” (Doc. 217 at 163). This testimony does not support the Democratic Confer*1322ence plaintiffs’ speculation, and the plaintiffs offer no further proof.
We find that race did not predominate in House District 97, which is a compact district that runs along the western border of Mobile Bay. Its shape changed only a little, mainly through the addition of a-“bishop’s mitre” (which the plaintiffs inaccurately call a bishop’s head) in the northwest
[[Image here]]
(Ala. Reapportionment Office, supra). The Democratic Conference plaintiffs argue that the mitre reaches up for majority-black population — and the dissent agrees, (Dissent at 1402-03) — but their own map shows that the mitre takes in mostly majority-white areas (marked by lighter colors in the map below):
District 97 Bishop’s Mitre
[[Image here]]
(ADC Supp. Ex. 30B). The Black Caucus plaintiffs concede that “the split between HD 96 and HD 97 may have been intended to fair the boundary between these two districts.” (Doc. 256 at 131). We agree. We also note that some of the western border is explained by the residence of the incumbent ⅛ District 99:
Mobile County House Incumbents in 2012
*1323[[Image here]]
(Def. Supp. Ex. 2). The overall shape of the district does not lead us to find that race predominated.
Nor do the alternative plans produced by the plaintiffs. The Democratic Conference plaintiffs drew a district farther to the north and east, but they did not explain why their choice was required by traditional districting criteria other than precinct splits, which we address later.
House District 97 in Democratic Conference Plan A and Act 602
[[Image here]]
*1324(Doc. 287-17 at 2). The Black Caucus plaintiffs made similar changes to the district, again unexplained by anything but precinct splits:
House District 97 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 530).
The black population percentage in this district provides no evidence that race predominated. The enacted plan matches the previous plan exactly, but the McClammy Plan and Reed-Buskey Plans are slightly higher; the Knight Plan, New Black Caucus Plan, and 1% Plan are within four percentage points of the previous plan; and Plan A is within five percentage points.
2010 Census Total Black Population Percentages in District 83 Under Various Plans
[[Image here]]
The enacted district is in the heart of the range of alternative districts.
The plaintiffs argue that precincts were split in District 97 along racial lines, but they cannot identify any racially suspicious precincts. The plaintiffs first argue that there were racially motivated splits between District 97 and majority-white districts, but the plaintiffs disagree about which precinct splits are suspicious. The Democratic Conference plaintiffs argue that the bishop’s mitre was extended to reach black population. (Doc. 258 at 86-87). But as already discussed, the Democratic Conference plaintiffs’ own maps — as well as those of their coplaintiffs — contradict this argument. The Democratic Conference plaintiffs’ map, (ADC Supp. Ex. 30B), proves that the drafters pushed the boundaries past the majority-black blocks *1325to take in majority-white population as well. And the Black Caucus plaintiffs admit that the split of Chickasaw Auditorium precinct in this area “may have been intended to fair the boundary” between House Districts 97 and 96. (Doc. 266 at 131). They are correct: the bishop’s mitre is not obviously drawn with race in mind. (APSX 179).
The Black Caucus plaintiffs and the dissent assert that Saraland Civic Center precinct was split along racial lines, but the map and statistics prove otherwise. The split placed very few people into District 97-306 people, only 33 percent of them black. (Def. Ex. 405 at 638). The vast majority of the precinct was split between District 98 and District 96. District 98 received 118 black people out of 1,430 total, and District 96 received 237 black people out of 2,332 total. (Id. at 640, 637). Moreover, the borders of the split are smooth.
Saraland Civic Center Precinct in Act 6
[[Image here]]
(APSX 197). This precinct has barely any black population, and the black population is roughly equal in the three districts.
We have also examined the other precinct splits in this district and find no evidence that race predominated. In Vigor High School, 100 Black Men of Greater Mobile, Figures Recreation Center, Murphy High School Library, Augusta Evans School, and Rock of Faith Baptist Church precincts, District 97 split the precincts with another majority-black district according to no apparent pattern. (APSX 174, 175, 182, 194, 196, 202, 207). The plaintiffs make no argument to the contrary. In Bishop St. Community College, the drafters split the precinct along the Mobile River and put the zero-population blocks on the other side of the river in District 96. (APSX 178). Likewise, the drafters put only unpopulated blocks on *1326one side of the split in Whitley School precinct. (APSX 208). And in St. Andrews Episcopal Church precinct, (APSX 202), the drafters divided majority-white blocks ,in a straight line that improved the shape of the district.
The dissent asserts that race predominated because the drafters did not consider it a “problem” if a district had a high black population percentage, but they nevertheless split precincts, which reduced the black population in the district. (Dissent at 1403). But Representative McClen-don, whose deposition testimony the dissent cites, did not testify that he was not concerned about high black population percentages. He testified that a black population increase of five percent did not constitute packing. (Doc. 125-4 at 106, 109-10). We find that race did not predominate in the design of House District 97.
We also find that race did not predominate in House District 98, which decreased in black population percentage from 65.22’ percent to 60.02 percent. (Def. Ex. 406 at 659; Def. Ex. 403 at 424). District 98 is centered on the towns of Prichard and Saraland. It extends north along waterways to gain population from small towns up to and including Mount Vernon. (See U.S. Census Bureau, State Legislative District Reference Map: State House District 98 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01098/DC10SLDL01098_ 001.pdf). The northern extension transfers population from overpopulated District 102 to underpopulated District 98. We also know that some portions of Districts 98 and 102 were exchanged at the request of the incumbents, whose motivations the plaintiffs do not challenge. Their wishes provide a race-neutral explanation for the northern part of the shape. (Doc. 217 at 225-26).
The alternative plans do not convince us that race predominated in the design of District 98. Both plaintiffs moved west instead of north to reach more population:
House District 98 in Democratic Conference Plan A and Act 602
[[Image here]]
*1327(Doc. 287-17 at 3).
House District 98 in Black Caucus 1% Plan and Act 602
[[Image here]]
(APSX 531 (Black Caucus district shaded purple)). Both districts are more compact and regular, but the plaintiffs do not explain how their choices prove that race predominated in the enacted district, other than the arguments about precinct splits that we address below.
The black population percentages in District 98 are nearly identical across the various plans. No plan had a District 98 with a black population of less than 60 percent, and several of the earlier plans— which did not even follow the ± 1% population deviation — came closer to the previous percentage than the state did.
2010 Census Total Black Population Percentages in District 98 Under Various Plans
[[Image here]]
From these statistics, we infer that demographics and legitimate districting criteria explain the black population percentage in District 98.
The plaintiffs again argue that the drafters split precincts to increase the black population percentage, but their argument again fails to prove that race predominated. First, in Mt. Vernon Civic Center precinct, the drafters split the precinct along zero-population blocks and between majority-white blocks on either side of the border. We can find no racial pattern:
Close-Up of Mt. Vernon Civic Center Precinct
*1328[[Image here]]
(APSX 193). Second, in the split of Turner-ville Community precinct between District 98 and majority-white District 102, the drafters put similar black population percentages in each district — seven percent in District 98 and three percent in District 102. (Def. Ex. 405 at 640, 646). Third, in First Baptist Church of Axis precinct, the split followed zero-population blocks except where District 98 absorbed majority-white blocks,
First Baptist Church of Axis Precinct in Act 602
*1329[[Image here]]
*1330(APSX 183). Fourth, in Havenwood Baptist Church precinct, the split smoothed out an irregular precinct boundary by placing majority-white blocks in District 98.
Zoom of Havenwood Baptist Church Precinct in Act 602
[[Image here]]
(APSX 186). Fifth, the legislature split both College Park Baptist Church and Chickasaw Auditorium precincts in three, but District 98 borders only another majority-black district in each precinct, and the plaintiffs do not identify any pattern of racial predominance in those splits. (APSX 179, 180). Sixth, we cannot find a racial pattern in the split of Shelton Beach Rd. Baptist Church precinct, which sorts almost exclusively majority-white and zero-population blocks along the border and places only a 20 percent black, population in District 98.
Shelton Beach Rd. Baptist Church Precinct in Act 602
*1331[[Image here]]
(APSX 200). Moreover, both Plan A and the 1% Plan split this precinct with a 15.6 percent black population in District 98. (Doc. 296-6; APSX 633). Seventh, the legislature split Saraland Civic Center precinct among Districts 96, 97, and 98. The plaintiffs identify no pattern of racial predominance in the split between majority-black Districts 97 and 98, and the split between Districts 96 and 98 placed roughly equal black population percentages in each district — 8 percent in District 98 and 10 percent in District 96. (Def. Ex. 405 at 637, 640). Lastly, in the splits of Little Welcome Baptist Church, Joseph Dotch Comm. Center, Vigor High School, Whitley School, and 100 Black Men of Greater Mobile precincts, House District 98 borders other majority-black districts, and the plaintiffs identify no racial pattern. (APSX 174, 188, 190, 207, 208). None of these precinct splits establish that race predominated.
The split of Satsuma City Hall precinct is mildly suspicious, but it is insufficient to prove that race predominated. We observe that the shape is irregular along part of the border and tends to put majority-black blocks in District 98 and majority-white blocks in District 96.
Satsuma City Hall Precinct in Act 602
*1332[[Image here]]
(APSX 198). We also observe that the split placed 796 people in District 98, 51 percent of whom were black, and 3,431 people in District 96, 5 percent of whom were black. (Def. Ex. 405 at 637). But part of the border is formed by a majority-white block on the District 98 side opposite a majority-black block on the District 102 side, which suggests that race did not dictate the shape of this split. The black population from this precinct in District 98 amounts to less than a percent of the total population of the district, and if we remove the precinct entirely, the black population percentage increases negligibly from 60.02 to 60.19. We find that race did not predominate in the design of House District 98.
We find that race predominated in the design of House District 99, which is based in northern Mobile, on the basis of three precinct splits and the shape of the district. The black population decreased from 73.35 percent to 65.61 percent, (Def. Ex. 406 at 659; Def Ex. 403 at 424), and the shape is not outrageous. It extends to the northwest to pick up population from overpopulated District 102, likely because every other direction was blocked. The districts to the northeast, east, and southeast were underpopulated, and the residence of the incumbent in District 101 prevented significant expansion to the south. (Def. Supp. Ex. 2). District 99 maintains the same core as under the 2001 plan, al*1333though it is not quite as compact as its predecessor and contains a bottleneck in the middle.
[[Image here]]
(Ala. Reapportionment Office, supra).
Nor is the black population percentage in the district evidence of racial predominance. Some of the alternative plans had comparable proposed black population percentages, and some had lower percentages.
2010 Census Total Black Population Percentages in District 99 Under Various Plans
[[Image here]]
But the drafters badly missed any supposed target by nearly eight percentage points. Without further evidence, that some of the plans had lower black population percentages is insufficient to prove that race predominated.
With their alternative plans, the plaintiffs established that District 99 could be drawn more compactly and regularly, but these plans do not prove that race predominated in the enacted district.
House District 99 in Democratic Conference Plan A and Act 602
*1334[[Image here]]
(Docs. 287-17 at 4).
House District 99 in Black Caucus 1% Plan and Act 603
[[Image here]]
(APSX 532).
Most of the precinct splits are not suspicious. In the splits of Joseph Dotch Comm. Center, Figures Recreation Center, Murphy High School Library, Augusta Evans School, and Pleasant Valley Methodist Church precincts, District 99 borders another majority-black district with no racial pattern. (APSX 175, 182, 188, 194, 195). In St. John United Methodist Church pre*1335cinct, the drafters put two majority-black blocks along the border in majority-white District 101 and three majority-white blocks along the border in District 99.
St. John United Methodist Church Precinct in Act 602
[[Image here]]
(APSX 203). The drafters split College Park Baptist Church precinct three ways, The split with majority-black District 98 has no apparent racial pattern, and the split with majority-white District 102 divided an area of majority-white blocks to put 157 people in District 102. (Doc. 30-41 at 646). None of those 157 people were black, but this split is not visually bizarre and at best provides slight evidence that race predominated,
College Park Baptist Church Precinct in Act 602
[[Image here]]
(APSX 180). In Azalea City Church of Christ, the drafters put 46 percent black population on the District 99 side and 15 percent black population on the District 102 side. But they did so using a straight line.
*1336Azalea City Church of Christ Precinct in Act 602
[[Image here]]
(APSX 176).
In Moffett Road Assembly of God precinct, the drafters split the precinct along a large, sparsely populated block of two people.
Moffett Road Assembly of God Precinct in Act 602
[[Image here]]
*1337(APSX 192). Although the division is not quite along a zero-population block, it might as well be. This split does not provide evidence that race predominated.
Finally, in Friendship Missionary Baptist Church precinct, the drafters divided the precinct along another natural break in popolation. They put 1,798 people in District 99, 46 percent of whom were black, and 178 people in District 101, 4 percent of whom were black. (Def. Ex. 405 at 643, 645). But the split divides a densely populated area in the southwest from a densely populated area in the east of the precinct.
Friendship Missionary Baptist Church
[[Image here]]
*1338(APSX 185). A map from the Census Bureau confirms that most of the roads in the district are in the east or the southwest, with relatively few near the split.
Census Bureau Map in the Vicinity of Friendship Missionary Baptist Church Precinct
[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 99 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLD_RefMap/ lower/st01_al/sldl01099/DC10SLDL01099_ 001.pdf). The Black Caucus plaintiffs also split this district in an even more racial fashion, putting a population that was 93 percent black in District 99 and a population that was 36 percent black in majority-white District 100. (APSX 633). Yet the Black Caucus plaintiffs maintain that race did not predominate in their plan. This split is not evidence that race predominated.
Three split precincts — Semmes First Baptist, University City Church of Christ, and Little Welcome Baptist Church — exhibit clear patterns of racial sorting, and Alabama offers no explanation for these patterns. The split of Semmes First Baptist Church precinct placed 870 people in District 99, 50 percent of them black, and 6,332 people in District 102, 12 percent of them black. (Def. Ex. 405 at 641, 646).
Close-Up of Semmes First Baptist Church Precinct in Act 602
*1339[[Image here]]
(ASPX 199). The split of University Church of Christ precinct put 1,485 people in District 99, 62 percent of them black, and 2,081 people in District 101, 29 percent of them black. (Def. Ex. 405 at 642, 645).
University Church of Christ Precinct in Act 602
[[Image here]]
*1340(APSX 203). And the split of Little Welcome Baptist Church precinct put 3,432 people in District 99, 66 percent of them black; 1,440 people in District 98, 86 percent of them black; and 115 people in District 101, 17 percent of them black. (Def. Ex. 405 at 641-42, 644).
Close-Up of Little Welcome Baptist Church Precinct in Act 603
[[Image here]]
(APSX 190). These three precinct splits place a significant number of black people in District 99 on the basis of race. On the basis of these precinct splits and the shape of the district, we find that race predominated in the design of House District 99.
We further conclude that District 99 does not survive strict scrutiny. Once again, the only argument that Alabama makes is based on the comments of Senator Sanders and Representative Jackson. And once again, these comments do not provide a strong basis in evidence because this district is located in a different part of the state. They also do not provide a strong basis in evidence because the district is more urban — 93 percent of the population lives in the city of Mobile. (Def. Supp. Ex. 3). Because the state has not provided a strong basis in evidence for its use of race, we must enjoin the use of District 99 in future elections.
Finally, we find that race did not predominate in House District 103. The district, which decreased from 69.64 percent black to 65.06 percent black, (Def. Ex. 406 at 659; Def. Ex. 403 at 424), kept the core of its shape.
2001 District Lines
*1341[[Image here]]
(Ala. Reapportionment Office, supra). The district now extends farther to the northwest and the south.
The black population percentages in the alternative plans are similar to the percentage in the enacted plan. Of the plans with a majority-black District 103, none have a black population more than three percentage points lower than the enacted district.
2010 Census Total Black Population Percentages in District 103 Under Various Plans
[[Image here]]
*1342This similarity is evidence that legitimate districting choices and demographics produced the black population percentage in District 103.
At most, the alternative plans prove that the drafters could have drawn the district without splitting as many precincts. The plaintiffs both drew significantly different districts, and they do not explain why the overall decisions were compelled by traditional districting criteria other than precinct splits.
House District 103 in Democratic Conference Plan A and Act 602
[[Image here]]
(Doc. 287-17 at 8).
House District 103 in Black Caucus 1% Plan and Act 602
[[Image here]]
*1343(APSX 533).
The precinct splits do not prove that race predominated. The legislature split Pleasant Valley Methodist Church, Rock of Faith Baptist Church, and St. Andrews Episcopal Church precincts with another majority-black district using no racial pattern. (APSX 195-96, 202). We observe that the 1% Plan split Rock of Faith Baptist Church precinct as well. (APSX 196).
Three splits with majority-white districts also are not suspicious. First, in Bay of the Holy Spirit Church precinct, the legislature drew a border between Districts 101 and 103 along a smooth line of zero-population blocks. The only deviation places a majority-&kc/c block in the majority-white district, even though the majority-black district missed its supposed tar- ^
Bay of the Holy Spirit Church Precinct in Act 602
[[Image here]]
(APSX 177). Without further explanation, this split does not establish that race predominated. Second, in Hollingers Island School precinct, the legislature used a straight line to divide majority-white blocks between Districts 103 and 105.
Hollingers Island School Precinct
[[Image here]]
*1344(APSX 187). The split placed a roughly equal percentage of black population in District 103 (6 percent) and District 105 (4 percent). (Def. Ex. 405 at 648, 650). Third, in Dodge School precinct, the legislature placed a small cluster of majority-black and zero-population blocks in District 103. The split put only 100 people in District 103, 84 percent of them black, and 4,377 people in District 101, 31 percent of them black. (Id. at 645, 647). We find no evidence in this precinct that race predominated because the split follows a road that forms the precinct boundary in The Mug Cafe precinct to the north. As this map illustrates, the split smooths out irregular precinct lines and improves the shape of the district.
Dodge School Precinct in Act 602
[[Image here]]
*1345(APSX 181).
Census Bureau Map in the Vicinity of Dodge School Precinct
[[Image here]]
(U.S. Census Bureau, State Legislative District Reference Map: State House District 103 (Alabama), http://www2.census. gov/geo/maps/dclOmap/SLDJEtefMap/ lower/st01_a]/sldl01103/DC10SLDL01103_ 001.pdf).
The Mug Cafe Precinct in Act 602
*1346[[Image here]]
(APSX 204).
Three other precinct splits in District 103 provide some evidence that race predominated. In The Mug Cafe precinct, pictured above, the legislature drew an irregular line to place a cluster of majority-black and zero-population blocks in District 103. The split put 100 people in District 103, 84 percent of them black, and 4,337 people in District 101, 31 percent of them black. (Def. Ex. 405 at 645, 647). Second, in Kate Shepard School precinct, the drafters brought populated blocks into District 103 in two areas.
Kate Shepard School Precinct in Act 602
*1347[[Image here]]
(APSX 189). Although the drafters added an unnecessary majority-white block of 271 people the split placed 659 people in District 103, 46 percent of them black, and 2,315 people in District 104, only 8 percent of them black. (Def. Ex. 405 at 648-49). Third, in First Independent Methodist Church precinct, the drafters used irregular lines to place most of the majority-black blocks in District 103.
First Independent Methodist Church Precinct in Act 602
*1348[[Image here]]
(APSX 184). The split placed 121 people in District 103, 98 percent of them black, and 5,150 people in District 104, only 22 percent of them black. (Def. Ex. 405 at 648-49). Alabama offers no explanations for these splits.
We find that these splits are not enough to prove that race predominated. The number of black people that they put in District 103 amounts to less than 2 percent of the total population of the district, and if we remove the precincts entirely, the black population percentage of the district would remain almost unchanged at 65.21 percent. This change is all the more negligible because District 103 fell more than four points short of the previous black population percentage. We find that race did not predominate in the design of House District 103.
CONCLUSION
We GRANT judgment for the plaintiffs with respect to Senate District 20, Senate District 26, Senate District 28, House District 32, House District 53, House District 54, House District 70, House District 71, House District 77, House District 82, House District 85, and House District 99, and we ENJOIN the use of these twelve districts in future elections. We GRANT judgment for the defendants with respect to the other 24 challenged districts. A sep*1349arate order setting a status conference will follow.
APPENDIX
[[Image here]]
THE ALABAMA LEGISLATURE
STATE OF ALABAMA
REAPPORTIONMENT COMMITTEE GUIDELINES
FOR CONGRESSIONAL, LEGISLATIVE, AND STATE BOARD OF EDUCATION REDISTRICTING
May 2011
Pursuant to the Constitution of the United States and the Constitution of the State of Alabama, the Alabama State Legislature is required to review 2010 Federal Decennial Census data provided by the U.S. Bureau of the Census to determine if it is necessary redistrict Alabama’s congressional, legislative, and State Board of Education districts because of population changes since the 2000 Census. Accordingly, the following guidelines for congressional, legislative, and State Board of Education redistricting have been established by the Legislature’s Permanent Joint Legislative Committee on Reapportionment, (hereinafter referred to as the “Reapportionment Committee”).
I. POPULATION
The total Alabama resident state population of 4,779,736 persons, and the population of defined subunits thereof, as reported by the 2010 Census, shall be the permissible data base used for the development, evaluation, and analysis of proposed redistricting plans. It is the intention of this provision to exclude from use any census data, for the purpose of determining compliance with the one person, one vote requirement, other than that provided by the United States Census Bureau.
II. EQUAL POPULATION REQUIREMENT: ONE PERSON-ONE VOTE
The goal of redistricting is equality of population of congressional, legislative, and State Board of Education districts as defined below.
1. Congressional Districts
The Apportionment Clause of Article 1, Section 2, of the United States Constitution requires that the population of a state’s congressional districts in a state be “as nearly equal in population as practicable.” Accordingly, Congressional redistricting plans must be as mathematically equal in population as is possible.
2. Legislative And State Board of Education Districts
In accordance with the Equal Protection Clause of the Fourteenth Amend*1350ment to the United States Constitution, legislative and State Board of Education districts will be drawn to achieve “substantial equality of population among the various districts.”
a. Any redistricting plan considered by the Reapportionment Committee will comply with all relevant case law regarding the one person, one vote principle of the equal protection clause of the 14th Amendment of the United States Constitution, including but not limited to the cases of Larios v. Cox, 300 F.Supp.2d 1320 (N.D. Ga. 2004) aff'd sub nom Cox v. Larios, 542 U.S. 947, 124 S.Ct. 2806, 159 L.Ed.2d 831 (2004), and White v. Regester, 412 U.S. 755, 93 S.Ct. 2332, 37 L.Ed.2d 314 (1973). When presenting plans to the Reapportionment Committee, proponents should justify deviations from the ideal district population either as a result of the limitations of census geography, or as a result of the promotion of a consistently applied rational state policy.
b. In keeping with subpart a, above, a high priority of every legislative and State Board of Education redistricting plan must be minimizing population deviations among districts. In order to ensure compliance with the most recent case law in this area and to eliminate the possibility of an invidious discriminatory effect caused by population deviations in a final legislative or State Board of Education redistricting plan, in every redistricting plan submitted to the Reapportionment Committee, individual district populations should not exceed a 2% overall range of population deviation. The Reapportionment Committee will not approve a redistricting plan that does not comply with this requirement.
III. VOTING RIGHTS ACT
1. Districts shall be drawn in accordance with the laws of the United States and the State of Alabama, including compliance with protections against the unwarranted retrogression or dilution of racial or ethnic minority voting strength. Nothing in these guidelines shall be construed to require or permit any districting policy or action that is contrary to the United States Constitution or the Voting Rights Act of 1965.
2. Redistricting plans are subject to the preclearance process established in Section 5 of the Voting Rights Act.
IV. CRITERIA FOR CONGRESSIONAL, LEGISLATIVE, AND STATE BOARD OF EDUCATION DISTRICTS
1. All congressional, legislative, and State Board of Education districts will be single-member districts that comply with the population-equality standards discussed above.
2. A redistricting plan will not have either the purpose or the effect of diluting minority voting strength, shall not be retrogressive, and shall otherwise comply with Sections 2 and 5 of the Voting Rights Act and the Fourteenth and Fifteenth Amendments to the Constitution.
3. No district will be drawn in a manner that subordinates race-neutral districting criteria to considerations that stereotype voters on the basis of race, color, or membership in a language-minority group.
*13514. All legislative and congressional districts will be composed of contiguous and reasonably compact geography •
5. The following legislative redistricting requirements prescribed by the Alabama Constitution shall be complied with:
a. Sovereignty resides in the people of Alabama, and all districts should be drawn to reflect the democratic will of all the people concerning how their governments should be restructured.
b. House and Senate districts shall be drawn on the basis of total population.
c. The number of Senate districts is set by statute at 35 and, under the Alabama Constitution, may not exceed 35.
d. The number of Senate districts shall be not less than one-fourth or more than one-third of the number of House districts.
e. The number of House districts is set by statute at 105 and, under the Alabama Constitution, may not exceed 106.
f. The number of House districts shall not be less than 67.
6. The following redistricting policies contained in the Alabama Constitution shall be observed to the extent that they do not violate or conflict with requirements prescribed by the Constitution and laws of the United States:
a. Each House and Senate district should be composed of as few counties as practicable.
b. Every part of every district shall be contiguous with every other part of the district. Contiguity by water is allowed, but point-to-point contiguity and long-lasso contiguity is not.
c.Every district should be compact.
7.The following redistricting policies are embedded in the political values, traditions, customs, and usages of the State of Alabama and shall be observed to the extent that they do not violate or subordinate the foregoing policies prescribed by the Constitution and laws of the United States and of the State of Alabama:
a. Contests between incumbent members of Congress, the Legislature, and the State Board of Education will be avoided when ever possible.
b. The integrity of communities of interest shall be respected. For purposes of these Guidelines, a community of interest is defined as an area with recognized similarities of interests, including but not limited to racial, ethnic, geographic, governmental, regional, social, cultural, partisan, or historic interests; county, municipal, or voting precinct boundaries; and commonality of communications. Public comment will be received by the Reapportionment Committee regarding the existence and importance of various communities of interest. The Reapportionment Committee will attempt to accommodate communities of interest identified by people in a specific location. It is inevitable, however, that some interests will be advanced more than others by the choice of particular district configurations. The discernment, weighing, and balancing of the varied factors *1352that contribute to communities of interest is an intensely political process best carried out by elected representatives of the people.
c. Local community and political leaders and organizations and the entire citizenry shall be consulted about new district lines.
d. In establishing congressional and legislative districts, the Reapportionment Committee shall give due consideration to all the criteria herein. However, priority is to be given to the compelling state interests requiring equality of population among districts and the Voting Rights Act of 1965, as amended, should the requirements of those criteria conflict with any other criteria.
V. PLANS PRODUCED BY LEGISLATORS
1. The confidentiality of any Legislator developing plans or portions thereof will be respected. The Reapportionment Office staff will not release any information on any Legislator’s work without written permission of the Legislator developing the plan, subject to paragraph two below.
2. A proposed redistricting plan will become public information upon its introduction as a bill in the legislative process, or upon presentation for consideration by the Reapportionment Committee.
3. Access to the Legislative Reapportionment Office Computer System, census population data, and redistricting work maps will be available to all members of the Legislature upon request. Reapportionment Office staff will provide technical assistance to all Legislators who wish to develop proposals.
4. In accordance with Rule 23 of the Joint Rules of the Alabama Legislature (2011) all amendments or revisions to redistricting plans, following introduction as a bill, shall be drafted by the Reapportionment Office.
5. Drafts of all redistricting plans which are presented for introduction at any session of the Legislature, and which are not prepared by the Reapportionment Office, must be presented to the Reapportionment Office for review of proper form and for entry into the Legislative Data Bank.
VI. REAPPORTIONMENT COMMITTEE MEETINGS AND PUBLIC HEARINGS
1. All meetings of the Reapportionment Committee and its sub-committees will be open to the public and all plans presented at committee meetings will be made available to the public.
2. Minutes of all Reapportionment Committee meetings shall be taken and maintained as part of the public record. Copies of all minutes shall be made available to the public.
3. Transcripts of all public hearings shall be made and maintained as part of the public record, and shall be available to the public.
4. The Reapportionment Committee will hold public hearings at different locations throughout the State in order to actively seek public participation and public input.
5. All interested persons are encouraged to appear before the Reapportionment Committee and to give their comments and input regarding con*1353gressional, legislative, and State Board of Education redistricting. Reasonable opportunity will be given to such persons, consistent with the criteria herein established, to present plans or amendments redistricting plans to the Reapportionment Committee, if desired, unless such plans or amendments fail to meet the minimal criteria herein established.
6. Notices of all Reapportionment Committee meetings will be posted on the fifth, sixth, seventh, and eighth floors of the Alabama State House, the Reapportionment Committee’s website, and on the Secretary of State’s website. Individual notice of Reapportionment Committee meetings will be sent by email to any citizen or organization who requests individual notice and provides the necessary information to the Reapportionment Committee staff. Persons or organizations who want to receive this information should contact the Reapportionment Office.
VII. PUBLIC ACCESS
1. The Reapportionment Committee seeks active and informed public participation in all activities of the Committee and the widest range of public information and citizen input into its deliberations. Public access to the Reapportionment Office computer system is available every Friday from 8:30 a.m. to 4:30 p.m. Please contact the Reapportionment Office to schedule an appointment.
2. A redistricting plan may be presented to the Reapportionment Committee by any individual citizen or organization by written presentation at a public meeting or by submission in writing to the Committee. All plans submitted to the Reapportionment Committee will be made part of the public record and made available in the same manner as other public records of the Committee.
3. Any proposed redistricting plan drafted into legislation must be offered by a member of the Legislature for introduction into the legislative process.
4. A redistricting plan developed outside the Legislature or a redistricting plan developed without Reapportionment Office assistance which is to be presented for consideration by the Reapportionment Committee must:
a. Be clearly depicted on maps which follow 2010 Census geographic boundaries;
b. Be accompanied by a statistical sheet listing total population and minority population for each district and listing the census geography making up each proposed district;
c. Stand as a complete statewide plan for redistricting, or, if presenting a partial plan, fit back into the plan which is being modified, so that the proposal can be evaluated in the context of a statewide plan (i.e., all places of geography must be accounted for in some district);
d. Comply with the guidelines adopted by the Reapportionment Committee.
5. Electronic Submissions
a. Electronic submissions of redistricting plans will be accepted by the Reapportionment Committee.
b. Plans submitted electronically must also be accompanied by the paper materials referenced in this section.
*1354c. See the Appendix for the technical documentation for the electronic submission of redistricting plans.
6. Census Data And Redistricting Materials
a. Census population data and census maps will be made available through the Reapportionment Office at a cost determined by the Permanent Legislative Committee on Reapportionment.
b. Summary population data at the precinct level and a statewide work maps will be made available to the public through the Reapportionment Office at a cost determined by the Permanent Legislative Committee on Reapportionment.
c. All such fees shall be deposited in the state treasury to the credit of the general fund and shall be used to cover the expenses of the legislature.
Appendix.
ELECTRONIC SUBMISSION OF REDISTRICTING PLANS REAPPORTIONMENT COMMITTEE-STATE OF ALABAMA
The Legislative Reapportionment Computer System supports the electronic submission of redistricting plans. The electronic submission of these plans must be on either a flash drive or CD ROM. The software used by the Reapportionment Office is the Esri Redistricting Online (RO) Solution.
The electronic file should be in DOJ format (Block, district # or district #, Block). This should be a two column, comma delimited file containing the FIPS code for each block, and the district number. The Esri RO Solution has an automated plan import that creates a new plan from the block/district assignment list.
Web services that can be accessed directly with a URL and ArcView Shape-files can be viewed as overlays. A new plan would have to be built using this overlay as a guide to assign units into a blank RO Solution plan. In order to analyze the plans with our attribute data, edit, and report on, a new plan will have to be built in the RO Solution.
In order for plans to be analyzed with our attribute data, to be able to edit, report on, and produce maps in the most efficient, accurate and time saving procedure, electronic submissions are REQUIRED to be in DOJ format.
Example (DOJ FORMAT BLOCK, DISTRICT #) SSCCCTTTTTTBBBB, D SS is the 2 digit state FIPS code
CCC is the 3 digit county FIPS code TTTTTT is the 6 digit census tract code
BBBB is the 4 digit census block code , a comma goes before the district number
DDDD is the district number
(The above format is also acceptable with a blank space in place of the comma).
Contact Information:
Legislative Reapportionment Office
Room 811, State House
11 South Union Street
Montgomery, Alabama 36130
(334) 242-7941
For questions relating to reapportionment and redistricting, please contact:
Ms. Bonnie Shanholtzer
Supervisor
Legislative Reapportionment Office
*1355district@al-legislature.gov
Please Note: The above e-mail address is to be used only for the purposes of obtaining information regarding redistricting. Political messages, including those relative to specific legislation or other political matters, cannot be answered or disseminated to members of the Legislature. Members of the Permanent Legislative Committee On Reapportionment may be contacted through information contained on their Member pages of the Official Website of the Alabama Legislature.
THOMPSON, District Judge, concurring in part and dissenting in part:
I concur with the majority that 12 of the districts at issue do not pass constitutional muster: Senate Districts 20, 26, and 28 and House Districts 32, 53, 54, 70, 71, 77, 82, 85, and 99. I also concur that 12 districts are constitutional: Senate Districts 18, 19, and 33 and House Districts 19, 58, 59, 72, 76, 78, 84, 98, and 103.
However, for the reasons given below, I respectfully dissent from the majority’s conclusion that an additional 12 districts are constitutional: Senate Districts 23 and 24 and House Districts 52, 55, 56, 57, 60, 67, 68, 69, 83, and 97. Two of the districts for which the majority concluded that race predominated, SD 23 and HD 68, cannot survive strict scrutiny. In the other ten districts, for which the majority held that race did not predominate, I conclude that race predominated and they cannot survive strict scrutiny.
In this case, involving the constitutionality of Alabama’s 2012 House and Senate plans, the Supreme Court vacated the judgment previously adopted by the majority and remanded it back to our three-judge panel to consider racial-gerrymandering with- respect to individual districts for the purposes of resolving plaintiffs’ equal-protection claim. Ala. Legislative Black Caucus v. Alabama, — U.S. -, 135 S.Ct. 1257, 1274, 191 L.Ed.2d 314 (2015). That opinion provided us guidance about the importance of direct evidence demonstrating the State’s policy of using “mechanical racial targets” and a roadmap for assessing the districts challenged by plaintiffs. |d. at 1267, 1271. Specifically, as to racial predominance, the Court provided a roadmap for our district-specific analysis, instructing us to examine evidence indicating the State’s policy of achieving racial targets had an impact on “at least some of [a district’s] boundaries,” such as through the district’s shape, the racial composition of the population added to the district, or the preservation of the percentage of the population that was black. Id. at 1271. The Court also directed our attention to split precincts, especially where the population was “clearly divided on racial lines.” Id. Finally, the Court instructed that certain factors could not be considered traditional race-neutral factors overcoming the use of race, including the State’s equal-population objective and factors “not mentioned in the legislative redistricting guidelines.” Id. at 1271-72. To survive strict scrutiny, the Court instructed that the State must have a “ ‘strong basis in evidence’ in support of the (race-based) choice that it has made.” Id. at 1274. The Court further concluded that the State “expressly adopted and applied a policy of prioritizing mechanical racial targets above all other districting criteria (save one-person, one-vote),” id. at 1267, and described that approach as asking “the wrong question with respect to narrow tailoring.” Id. at 1274.
Because the majority does not adhere to all of the Court’s guidance in its analysis of the strict-scrutiny and racial-predominance prongs of plaintiffs’ equal-protection claim, *1356I cannot accept the majority’s analysis. Our disagreement is not merely a question of evidentiary disputes; the majority commits clear legal errors in its analysis of both the two districts where it found strict scrutiny satisfied and the ten districts where I disagree with the majority’s conclusions as to racial predominance.
I. Two Additional Districts
Because I partially agree with the majority as to SD 23 and HD 68, I will begin with them. The majority correctly concludes that race predominates with respect to these two districts. But I do not agree that the State has met its burden to demonstrate that its use of race survives strict scrutiny.
SD 23 and HD 68 are drawn in the largely rural West Black Belt region of Alabama. Each district consists of a majority-black population that was represented by a black legislator during redistricting. In each, the State hit its racial target— based on preserving the same percentage of the black percentage when applying the 2010 census to the 2001 district lines— almost to a tee: in SD 23, the drafters took a district that had been 64.76 % black under the old lines and drew it with a 64.84 % black population in the new plan; in HD 68, the drafters took a district that had been 62.55 % black and left it 64.56 % black. The State met its targets despite adding significant new population to each district to achieve its equal-population objective.
For evidentiary support on the strict-scrutiny prong, the majority relies principally on comments made by incumbent legislators during public hearings in the redistricting process. But these comments are inadequate to provide a strong basis in evidence. In the first comment, Representative Thomas Jackson, incumbent for HD 68, while expressing concern about packing too many blacks into majority-black districts, stated that instead Alabama’s majority-black districts “could be sixty-two percent or sixty-five percent” black. Reapportionment Comm. Oct. 13, 2011 Hrg. Tr. (doc. no. 30-23) at 8 (emphasis added). And that is it: Jackson offered no demographic support whatsoever for his comment. A few days later, Senator Henry Sanders, incumbent for SD 23, while expressing concern about drawing majority-black districts with too few blacks, suggested they “ought not to be less than 62 percent” black. Reapportionment Comm. Oct. 18, 2011 Hrg. Tr. (doc. no. 30-28) at 6 (emphasis added). The only demographic support Sanders offered for his comment is this: “Sometimes a lot of people don’t vote. Sometimes a lot of people can’t vote.... Sometimes a lot of folks are discouraged for one reason or another.” Id. (emphasis added). Moreover, Sanders unequivocally directed his comments at “majority African-American districts,” id. (emphasis added); that is, in the plural and thus generally.
As I will explain later and in detail, these generic and conclusory “could be” and “sometimes” comments provide inadequate evidentiary support for government action based on race under strict scrutiny. Strict scrutiny is designed to ensure that, before declaring a State’s use of race constitutional, the court must be reasonably confident that, as I believe the United States aptly put it in an amicus brief recently filed with the Supreme Court, the State “relied on information about district-specific characteristics — localized assessments drawn from knowledge of actual constituents, rather than stereotypes.” Brief for United States as Amicus Curiae at 32, Bethune-Hill v. Va. State Bd. of Elections (U.S. 2016) (No. 15-680), 2016 WL 4916822, at *32. Because the record is insufficient for this court to determine, one way or the other, that Alabama did not *1357rely on stereotypes about black voting behavior, the two districts cannot survive strict scrutiny and the majority’s conclusion otherwise is clearly erroneous.
A. StricWScrutiny Standard
Where a court concludes that race predominated in the design of a district, the court must further determine if that district can survive strict scrutiny — that is, “whether the [district] is narrowly tailored to further a compelling governmental interest.” Shaw v. Reno (“Shaw I”), 509 U.S. 630, 658, 113 S.Ct. 2816, 125 L.Ed.2d 511 (1993). Shaw I adopted, for racial-gerrymandering claims in the redistricting context, the two strict-scrutiny prongs from ordinary equal-protection analysis: first, “any racial classification ‘must be justified by a compelling governmental interest,’ ” and, second, “the means chosen by the State to effectuate its purpose must be ‘narrowly tailored to the achievement of that goal.’” Wygant v. Jackson Board of Education, 476 U.S. 267, 274, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (plurality opinion) (citations omitted).
As to the first prong, the parties do not dispute that the State held a compelling interest in compliance with a correct interpretation of § 5 of the Voting Rights Act, 42 U.S.C. § 1973c. Accordingly, my review centers on the second prong — the narrow-tailoring requirement.
To satisfy the narrow-tailoring prong here, the State must show that it had a strong basis in evidence for the way it used race in the design of a district. Ala. Legislative Black Caucus, 135 S.Ct. at 1274. Elaborating on the standard, the Supreme Court explained that the State “may have a strong basis in evidence to use racial classifications to comply with a statute when [it] ha[s] good reasons to believe such use is required, even if a court does not find that the actions were necessary for statutory compliance.” Id. (quoting Brief for the United States as Amicus Curiae 29). Put differently, while the State need not show that a compelling interest actually required it to use racial classifications, the State must show a strong basis in evidence for believing that the racial classification it used was required.
In adopting the strong-basis-in-evidence standard to satisfy the narrow-tailoring prong, the Supreme Court directed us to the standard for strict scrutiny that has been applied in both the voting-rights and racial-preference contexts. See Ala. Legislative Black Caucus, 135 S.Ct. at 1274 (“[W]e agree with the United States that a court’s analysis of the narrow tailoring requirement insists only that the legislature have a ‘strong basis in evidence’ in support of the (race-based) choice that it has made.”) (citing Brief for United States as Amicus Curiae 29 and Ricci v. DeStefano, 557 U.S. 557, 585, 129 S.Ct. 2658, 174 L.Ed.2d 490 (2009)).
In Ricci, an affirmative-action case that confronted a race-discrimination claim under Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., the Supreme Court concluded that, to survive strict scrutiny, the government needed more than a “good-faith belief’ that its racial preferences were necessary for statutory compliance. 557 U.S. at 581-82, 129 S.Ct. 2658. Even though the government reached its decision on the recommendation of an industry expert, that expert’s “few stray (and contradictory) ... statements” could not qualify as a strong basis in evidence, especially because he had not “studied the [issue] at length or in detail” and had not seen relevant data. Id. at 591, 129 S.Ct. 2658 (quotations and internal alterations omitted).
Beyond the Court’s recent application of the standard in Ricci, the strong-basis-in-evidence standard originally derived from *1358equal-protection cases in the racial-preference context. The plurality opinion in Wy-gant explained that a governmental actor using racial classifications for a remedial purpose must have “a strong basis in evidence for its conclusion that remedial action was necessary,” which, in turn, required that government actors proceed with “extraordinary care” and assemble “convincing evidence” before taking such action. 476 U.S. at 277, 106 S.Ct. 1842.
The same standard was applied in City of Richmond v. J.A. Croson Co., 488 U.S. 469, 109 S.Ct. 706, 102 L.Ed.2d 864 (1989), which determined that a race-based measure, supported by the testimony of government officials that was neither geographically specific nor accompanied by an explanation of the basis of their opinions, could not pass strict scrutiny, because such “highly conclusionary” and “generalized assertion[s]” lacked the rigor necessary to establish a strong basis in evidence. Id. at 500, 109 S.Ct. 706.
The majority asserts that the strict-scrutiny standard developed in racial preference cases has no application to this context. But well before the Supreme Court’s decision in this case, the strong-basis-in-evidence standard was imported into the racial-gerrymandering context. See Shaw I, 509 U.S. at 656, 113 S.Ct. 2816 (citing and applying strong-basis-in-evidence standard from Wygant and Croson). Similarly, in Bush v. Vera, Justice O’Connor, who authored the plurality opinion, explained in a separate concurrence that a State’s “ ‘strong basis in evidence’ need not take any particular form,” but made clear that it “[could not] simply rely on generalized assumptions” about voting behavior, such as “the prevalence of racial bloc voting.” 517 U.S. 952, 994, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996) (O’Connor, J., concurring). She concluded that the State’s reliance on a recent federal court ruling that had made extensive, geographically specific factual findings about voting behavior and was corroborated by expert testimony provided a strong basis in evidence. Id. at 994-95, 116 S.Ct. 1941 (O’Connor, J., concurring).1
In Shaw v. Hunt (“Shaw II”), the Supreme Court concluded that the State’s evidentiary burden on strict scrutiny can be met only by using evidence it actually considered at the time of redistricting; after-the-fact justification does not count. 517 U.S. 899, 908 n.4, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996) (stating that strong-basis-in-evidence standard requires interrogation of a legislature’s “actual purpose,” not “speculation about what ‘may have motivated’ ” it). The Court refused to credit the State’s proffer of historical and social science reports, generated only after the challenged districts had been drawn, to prove that it had a strong basis in evidence for its creation of two majority-black districts. Id. at 910, 116 S.Ct. 1894.
Shaw II also required a district at issue to be designed in a manner tailored to the evidence justifying it. The Supreme Court held that, although North Carolina had established a strong basis in evidence to draw a remedial district based on its failure to give effect to minority voting strength in one region of the State, that evidence did not establish a broad license to draw a remedial district elsewhere in the State. Id. at 917,116 S.Ct. 1894.
The majority observes that strict scrutiny does not require that a State engaged in redistricting must precisely guess what the Voting Rights Act demands; the State *1359is entitled to “a limited degree of leeway.” Bush, 517 U.S. at 977, 116 S.Ct. 1941. I fully agree. But here the question put is simply whether the record is adequate to determine that the State has not made its race-conscious decision based on stereotypes about black voting behavior and instead has based its decision on localized, district-specific characteristics.
Stated simply, while the strong-basis-in-evidence standard adopted by the Supreme Court in this case, Ala. Legislative Black Caucus, 135 S.Ct. at 1274, “need not take any particular form,” Bush, 517 U.S. at 994, 116 S.Ct. 1941 (O’Connor, J., concurring), it does require the State to point to concrete and specific evidence, which it actually relied upon before reaching its decision and which is sufficient for the court to interrogate the basis for the State’s approach. Further, the race-based decision must be narrowly tailored to address the harm to citizens in the district at issue, and it must be fully supported by the evidence. If the State cannot establish a strong basis in evidence consistent with these standards, redistricting undertaken with race as the predominant consideration is not narrowly tailored and cannot survive strict scrutiny.
B. SD 23 and HD 68
The majority’s striet-scrutiny conclusion for SD 23 and HD 68 rested primarily on the suggestions made by the districts’ incumbent legislators, Representative Jackson and Senator Sanders, which, the majority believes, gave the State a strong basis in evidence to draw majority-black districts within the range of a 62 % to 65 % black population. The majority correctly recognizes the legislators’ comments cannot provide a strong basis in evidence for majority-black districts across the entire State, as the State had argued. Nonetheless, the majority, relying on those same legislators’ suggestions, resuscitated the legislators’ two districts, transforming the State’s statewide argument into one specifically about SD 23 and HD 68 solely by virtue of the fact that Jackson and Sanders served as incumbents in those districts.
For the reasons discussed below, this approach is profoundly mistaken. The legislators’ comments on which the majority relies do not exhibit the “extraordinary care” demanded in strict-scrutiny analysis. Wygant, 476 U.S. at 277, 106 S.Ct. 1842. Instead, on their face, they betray that even the speakers themselves lacked precise recommendations: the appropriate percentage of black population “could be sixty-two percent or sixty-five percent” black, according to Representative Jackson, the incumbent for HD 68; the percentage in all majority-black districts “ought not to be less than 62 percent,” according to Senator Sanders, the incumbent for SD 23. Reapportionment Comm. Oct. 13, 2011 Hrg. Tr. (doc. no. 30-23) at 8; Reapportionment Comm. Oct. 18, 2011 Hrg. Tr. (doc. no. 30-28) at 6 (emphases added). But Sanders referred to all majority-black districts in Alabama — so we know that his recommendation was not tailored for SD 23 and HD 68. Jackson’s suggested “could be” percentages provide neither a precise suggestion nor a reason to think his passing comment was tailored to the districts at issue. Sanders’s 62 % minimum sets a bright-line rule for all majority-black districts based on things that happen, by his own words, only “sometimes”: he went on to explain that, “Sometimes a lot of people don’t vote. Sometimes a lot of people can’t vote.... Sometimes a lot of folks are discouraged for one reason or another.” In the districts where, “sometimes,” a high proportion of the black population exercises its right to vote, one would expect a more tailored approach to adopt a lower percentage of black population. Sanders did not tell us whether those *1360conditions exist in his district, and the record does not disclose it. What Jackson and Sanders told us “could be” relevant to the districts at issue “sometimes”; but without district-specific evidence, the record gives us no way to know. These comments do not provide a strong basis in evidence sufficient to satisfy strict scrutiny.
i. Background
In October 2011, before draft redistricting maps were available, the Permanent Legislative Committee on Reapportionment held 21 public hearings across the State to solicit public input. At one such hearing, on October 13 in Thomasville, Representative Jackson cautioned against packing: “I know they’ve compacted those districts, put the minorities in one district they could put them in, but we didn’t feel that was fair.” Reapportionment Comm. Oct. 13, 2011 Hrg. Tr. (doc. no. 30-23) at 8. Jackson’s concern about packing apparently derived from the recently completed redistricting process for Congressional and school-board districts. Suggesting that the drafters had not taken into account statements he had made at public hearings during that process, Jackson asked if the public hearings reflected a genuine attempt to solicit input: “[I]s this something that we’re just going around getting people to come out and say you’re doing this to make us feel good? Because when it’s to come for these — the vote for this stuff ... you-all had your minds all made up and you did what you was going to do.” Id. at 7-8.
Jackson then urged the drafters not to pack too many black people into a single district:
“REPRESENTATIVE JACKSON: ...
You know, and having a minority district over there,' it’s got to be ninety-nine percent minority.
“HEARING OFFICER: Right.
“REPRESENTATIVE JACKSON: It could be sixty-two percent or sixty-five percent. That’s what we’re trying to tell our good friends before on that. Nobody listens too well to us.”
Id. at 8 (emphasis added).
At another public hearing on October 18, 2011 in Selma, Senator Sanders urged the drafters not to place too few black people into majority-black districts:
“One of many concerns is we are not to have any less African-American — the majority African-American districts than you have, and that those districts ought not be less than 62 percent. And I just want to say why 62 percent, ought not to be less than 62 percent. Many times a population of a district is not reflective of the voters at all in that district. Sometimes a lot of people don’t vote. Sometimes a lot of people can’t vote. They might be in prison or other kinds of institutions. Sometimes a lot of folks are discouraged for one reason or another. So I would hope that 62 percent is a minimal for the majority African-American district.”
Reapportionment Comm. Oct. 18, 2011 Hrg. Tr. (doc. no. 30-28) at 6 (emphasis added). In essence, Sanders expressed concern about cracking, a gerrymandering tactic that involves spreading the population of a voting bloc among many districts to deny its members a sufficiently large population to elect the candidate of their choice in any particular district. Sanders also expressed a procedural concern, requesting a further hearing after the drafters had proposed a concrete redistricting plan because he did not believe a hearing absent one held much value: “Whenever you are talking in the abstract, it really doesn’t mean anything. It just doesn’t mean anything.” Id. In June 2012, after a draft redistricting plan had been shared *1361with senators, Sanders proposed an alternative plan that included a 57.75 % black population in his district, SD 23. Sanders Senate Plan, C-47, at 5. Sanders’s alternative plan suggests he did not believe the “sometimes” conditions applied in his district.
The State presented no evidence that the drafters of the redistricting plan relied upon the legislators’ statements in drafting SD 23 or HD 68. At his deposition, Representative Jim McClendon, the principal legislative architect of the House districts,2 attested that he did not remember testimony at the public hearings about packing, and did not otherwise testify about Jackson or Sanders’s statements.3 At trial, Senator Gerald Dial, the principal legislative architect of the Senate districts, recalled Sanders’s statement about a 62 % minimum black population for majority-black districts,4 but did not instruct Hinaman to incorporate that advice.5 Dial did reject certain alternative plans for Senate districts, including Sanders’s own plan, because “[i]t didn’t even meet the requirements [Sanders] had said would keep [the majority-black districts] 62 percent at least.” Tr. Vol. I (doc. no. 215) at 77:14-15. But there is no evidence that Dial actually considered Sanders’s statement when constructing the enacted majority-black Senate districts; to the contrary, the ultimately-enacted plans also included majority-black districts below Sanders’s suggested 62 %.6
For his part, Hinaman testified that he discussed with Dial and McClendon information they learned from meetings with black legislators, but the nature of the advice consisted of geographic-specific requests to incorporate or exclude specific areas. Tr. Vol. Ill (doc. no. 217) at 151:4-16. Hinaman testified that he took “some” comments from public hearings into consideration in his initial plan and “tried to get an overall feel for them as a group of all the hearings,” but that “[i]t’s not something you could implement all of their comments.” See Tr. Vol. Ill (doc. no. 217) at 148:23-25; Hinaman Dep., APX 75 (doc. no. 134-4) at 70:20-25.
In short, the statements relied on by the majority to conclude that strict scrutiny *1362has been met consist of imprecise statements made by legislators addressing different topics at public hearings that neither legislator viewed as a serious forum for input. On top of that, the State presented no evidence that the drafters actually incorporated Sanders’s or Jackson’s guidance into the design of the relevant districts.
ii. Analysis
As stated, the majority holds that the Jackson and Sanders comments provided the State a strong basis in evidence to conclude that it should draw both SD 23 and HD 68 within a range of 62 to 65 % black. See ante, at 1106 (“We conclude that [SD 23] satisfies strict scrutiny because the state had a strong basis in evidence to believe that ability to elect for purposes of the Voting Rights Act required black population percentages of 62 to 65 percent in this area. The drafters drew a district in that range.”); id. at 304 (applying same approach to HD 68). With this holding, the majority errs in at least eight ways.
First, the comments were not geographically specific. As the majority acknowledges, see ante, at 1062, neither Sanders nor Jackson offered a proposal specific to the conditions in SD 23 and HD 68.7 Con-clusory, geographically imprecise descriptions have “little probative value” in equal-protection analysis. Croson, 488 U.S. at 500-01, 109 S.Ct. 706 (rejecting reliance on legislator’s statement that racial discrimination existed “in this area, and the State, and around the nation” because it was not specific to the relevant city); accord Shaw II, 517 U.S. at 902, 116 S.Ct. 1894 (concluding that remedy tailored to one region of the State could not justify remedy applied to a different area of the State). Indeed, without district-specific demographic evidence, there is no way to be sure that the two districts are not based on stereotypes about black voting behavior.
Second, the legislators proposed imprecise remedies. Because Jackson was concerned about packing, his suggestion that a majority-black district “could be sixty-two percent or sixty-five percent” — in contrast to his earlier reference to a hypothetical district that is 99 % black — is most logically read as a casual reference to theoretically possible lower percentages that would sustain majority-black districts. In contrast, Sanders proposed a flat 62 % floor, suggesting that, “sometimes,” a higher or lower percentage of total black population would be necessary to achieve a district-specific tailored approach.
Third, the majority fails to recognize that Jackson and Sanders made proposals that, on their face, do not address the precise question: Jackson was concerned about racial packing and Sanders about cracking, or spreading the black population too thin across districts. Neither focused on the relevant question: “To what extent must we preserve existing minority percentages in order to maintain the minority’s present ability to elect the candidate of its choice?” Ala. Legislative Black Caucus, 135 S.Ct. at 1274. Without under*1363lying demographics, we do not know if they would have reached the same result if they had provided an answer to the relevant question.
Fourth, and relatedly, there is no evidence that the remedy proposed by Sanders was tailored to the State’s compelling interest in achieving § 5 compliance instead of a different goal, such as a legitimate self-interest in re-election.8 As Dial testified, “I never had a senator [in a majority-minority district] tell me to not put too many minorities in his or her district.” Tr. Vol. I (doc. no. 215) at 96:14-15. The narrow tailoring requirement demands a fit between the means and the particular end justifying the State’s use of race — not some altogether different end.
Fifth, that Sanders felt comfortable proposing a noticeably lower percentage (57.75 %) in his own Senate district just months after recommending a 62 % floor in all majority-black districts, see Sanders Senate Plan, C-47, at 5, suggests one of two things: either his original recommendation was never intended to apply to his own district, or it was not based on a “reasoned, principled explanation.” Fisher v. Univ. of Texas at Austin, — U.S. -, 136 S.Ct. 2198, 2208, 195 L.Ed.2d 511 (2016) (citation omitted). The majority insists that the legislature was “entitled” to rely on Sanders’s oral public hearing comments instead of his written plan because “the legislature chose to gather input about its redistricting effort through public hearings.” Ante, at 1110. But without underlying data or analysis, it is impossible to determine whether it was reasonable to rely on either proposal.
Moreover, the majority’s acceptance of the legislature’s unreasoned dismissal of Sanders’s alternative plan belies its attempt to bolster Sanders’s credibility. The majority declares Sanders “an expert in the politics of the Black Belt and Senate District 23” and recites a litany of his experience. See ante at 1108. But if the majority truly believes Sanders is so qualified, why does it accept the legislature’s unexplained rejection of Sanders’s written plan for his own district? Sanders’s knowledge and credibility is not disputed; rather, I question the court’s ability to evaluate, while applying the rigorous strict-scrutiny standard, the State’s purported reliance on one of his suggestions but dismissal of the other, without evidence in the record adequately explaining the basis of that choice. See ante at 1041 (citing Dial’s testimony that the State rejected Sanders’s own Senate plan solely because it did not comply with Sanders’s public hearing comments). Indeed, that Sanders himself said that abstract public hearing testimony “doesn’t mean anything,” Reapportionment Comm. Oct. 18, 2011 Hrg. Tr. (doc. no. 30-28) at 6, should at least give a court applying strict scrutiny pause about the State’s supposed preference.
Sixth, the State presented no evidence that it actually relied on the legislators’ comments when designing the relevant districts.9 The State can hold a strong basis in *1364evidence only if that evidence reflects its “actual purpose,” and not mere “speculation about what ‘may have motivated’ ” it. Shaw II, 517 U.S. at 908 n.4, 116 S.Ct. 1894.
Seventh, in the absence of evidence of whether and how the drafters actually used the legislators’ comments, the majority assumes that the drafters determined that Sanders’s statement set a 62 % floor and Jackson’s passing reference set a 65 % ceiling, and the majority apparently finds this range applicable to both Senate and House majority-black districts in the West Black Belt. See ante, at 1106-07, 1241-42, 1270-71,1297. But there is no evidence the State ever applied, or even contemplated, that range. Nor is there evidence that the State understood the comments as applicable solely to the West Black Belt region; in fact, Dial said the exact opposite: he understood Sanders’s comments applied to “all [majority-black] districts” across the State. Tr. Vol. I (doc. no. 215) at 37:13-25-38:1. The final redistricting plans provide ample evidence that, despite Dial’s understanding, the drafters did not adhere to this imagined range: five of eight majority-black Senate districts10 and 18 of 28 majority-black House districts, including two of the West Black Belt’s six House districts, fell outside the range.11
Eighth, and finally, the comments were not “sufficiently measurable to permit judicial scrutiny of the policies adopted to reach them,” Fisher, 136 S.Ct. at 2211, so that the court can reasonably assure itself that the State did not rely on racial stereotypes. This is, after all, strict scrutiny. The record reflects no data, studies, or even anecdotal accounts relied on by Jackson or Sanders to formulate their suggestions.12 We must be able to ask: How did the legislators formulate their ideas? What time period and geographic scope did they address? What data did they rely upon as the basis for their comments? How did the State understand and implement the suggestions? Here, with only a “few stray ... statements” plucked from a voluminous hearing record, we are left with only guesses. Ricci, 557 U.S. at 591, 129 S.Ct. *13652658. (I suspect that, if answers to these questions were in the record, Sanders and Jackson might -have admitted that their opinions were formed less by careful analysis and more by an off-the-cuff invocation of the now-discarded rule of thumb which had long considered a 65 % total black population necessary for black voters to elect the candidate of their choice. See, e.g., Barnett v. City of Chicago, 141 F.3d 699, 702-03 (7th Cir. 1998) (“[I]t is a rule of thumb that blacks must be at least 65 percent of the total population of a district in order to be able to elect a black.”). To say the least, their comments come suspiciously quite close to, if not coincide with, this racially stereotypical rule, a circumstance which should raise a red flag and warrant our inquiry into the underlying district-specific demographic bases, if any, for the comments before we can rely on them.)
C. Racial Targets
While I have observed that the comments of Sanders and Jackson are inadequate to assess whether the State used impermissible stereotyping about black voting behavior, it cannot be overlooked that affirmative evidence in the record demonstrates that racial stereotyping did occur. The record makes abundantly clear that, in redrawing the majority-black districts at issue here, the State of Alabama attempted to achieve mechanical racial targets based on a thoughtless repetition of the black percentage of each district’s population under district lines as drawn in 2001. Since that approach is profoundly inconsistent with the State’s burden to demonstrate that its use of race in each district was carefully considered, narrowly tailored, and supported by a strong basis in evidence, it serves as an important complement to the district-specific strict-scrutiny analysis.13
An assessment of narrow tailoring first requires a precise understanding of the State’s compelling interest. To narrowly tailor its approach to a compelling interest in compliance with the retrogression requirement of § 5 of the Voting Rights Act, the State’s approach should have been “reasonably necessary” under § 5. Miller v. Johnson, 515 U.S. 900, 921, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995) (“[Compliance with federal antidiscrimination laws cannot justify race-based districting where the challenged district was not reasonably necessary under a constitutional reading and application of those laws.”); accord Shaw II, 517 U.S. at 911, 116 S.Ct. 1894 (holding that State’s remedy, creating an additional majority-black district, was not narrowly tailored because it was “not required under a correct reading of § 5”). Although the State’s approach need not have been “actually necessary” to comply with § 5, the State must still have held a “strong basis in evidence” to “believe” its *1366use of race was required. Ala. Legislative Black Caucus, 135 S.Ct. at 1274.
Here, the State’s interest in § 5 compliance “does not require a covered jurisdiction to maintain a particular numerical percentage. It requires the jurisdiction to maintain a minority’s ability to elect a preferred candidate' of choice.” Ala. Legislative Black Caucus, 135 S.Ct. at 1272 (citing 52 U.S.C. § 10304(b)). The State should have known the correct requirement during drafting: Department of Justice Guidelines provided that its § 5 pre-clearance determinations are not based “on any predetermined or fixed demographic percentages.... Rather, in the Department’s view, this determination requires a functional analysis of the electoral behavior within the particular jurisdiction or election district.... [CJensus data alone may not provide sufficient indicia of electoral behavior to make the requisite determination.” Id. (quoting Guidance Concerning Redistricting Under Section 5 of the Voting Rights Act, 76 Fed. Reg. 7471 (2011)). In simple terms, the State should have asked itself: “To what extent must we preserve existing minority percentages in order to maintain the minority’s present ability to elect its candidate of choice?” Ala. Legislative Black Caucus, 135 S.Ct. at 1274.
By any measure, the approach actually taken by the State was woefully misguided: the drafters applied “mechanical racial targets,” id. at 1267, even though information about black electoral behavior, which would have furthered the State’s ability to conduct a functional, district-specific analysis of black citizens’ ability to elect the candidates of their choice, was readily available. See Ala. Legislative Black Caucus v. Alabama, 989 F.Supp.2d 1227, 1345 (M.D. Ala. 2013) (Thompson, J., dissenting) (describing drafters’ failure to consult readily available political data); accord Bush, 517 U.S. at 967, 116 S.Ct. 1941 (plurality opinion) (concluding that it was “evidentially significant that at the time of the redistricting, the State had compiled detailed racial data for use in redistricting, but made no apparent attempt to compile, and did not refer specifically to, equivalent data regarding communities of interest.”). Randy Hinaman, the State’s technical adviser, explained that he instead “tried to draw those [majority-black] districts as close to the [black percentage] numbers as possible and practicable as they were in the 2001 plan to avoid problems with potential Section 5 preclearance.” Tr. Vol. Ill (doc. no. 217) at 164:4 — 7. Senator Gerald Dial acknowledged that he did not examine the voting behavior of individual districts when guiding the drafting of Senate districts. Tr. Vol. I (doc. no. 215) at 136:8-14. Because the State mechanically maintained minority percentages in majority-black districts, the Court has already recognized that “the legislature asked the wrong question with respect to narrow tailoring.” Ala. Legislative Black Caucus, 135 S.Ct. at 1274.
District-specific evidence corroborates that the State’s misguided approach extended to the two districts at issue here. Hinaman touted his achievement of drawing an SD 23 that was “very close to getting back to the identical numbers that [it was] 10 years ago,” by which he meant that the resulting “total black percentage” was identical. Tr. Vol. Ill (doc. no. 217) at 186:19-25-187:1. Although the State added a net 23,399 people to SD 23, Hinaman managed to maintain almost identically the black percentage, from 64.76 % under the old lines to 64.84 % under the new ones. Senate Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 5; Comparison of Minority Percentages (doc. no. 263-2) at 4. Indeed, the majority acknowledges that, “The plaintiffs argue with some force that, given the extent of the changes to the *1367districts, the only way the drafters could have maintained black population percentages as close as they did was through a policy of racial targets.” Ante, at 1106 (emphasis added).
Under-populated HD 68 similarly reflects the application of a mechanical racial target. The State managed to keep the black population to within two percentage points — increasing it from 62.55 % to 64.56 % — even though the State added a net 8,835 people to address substantial underpopulation. Comparison of Minority Percentages (doc. no. 263-2) at 2; House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4. Stated differently, the State increased HD 68’s total population by 24 % while maintaining a black percentage within just 2 % of the prior plan.
With its mistaken conception of § 5 compliance, the State reduced the likelihood that it would reach the right result— in other words, that it would tailor its race-based action as narrowly as possible under the circumstances. But asking the “wrong question,” Ala. Legislative Black Caucus, 135 S.Ct. at 1274, also meant that the information the State actually reviewed and the analysis it conducted when tailoring its race-based approach had little, if any, correlation with the approach “reasonably necessary” for statutory compliance. This leaves the court — required to conduct our analysis using strict scrutiny — in a difficult position. We have no demographic evidence before us specific to SD 23 or HD 68 about black voting behavior, past election results, or other relevant factors necessary for determining how to preserve black citizens’ ability to elect the candidate of their choice in those districts — because the State never considered that kind of information. We have no evidence about how the State weighed that information — because it did not. We have no evidence about how the State considered alternative approaches — because it did not. In other words, the types of evidence the State would need to establish a strong basis in evidence are simply absent from the record because they were absent from its decision-making in the first instance. Under these circumstances, I cannot conceive how the majority could conclude that the State had a strong basis in evidence and the State’s approach was narrowly tailored.
Similarly, other courts have made short work of the State’s narrow-tailoring arguments where the State failed to make even the barest showing that it has conducted an appropriate inquiry. Reviewing a redistricting plan adopted by the Virginia legislature, it was easy for a three-judge court to conclude that the plan was not narrowly tailored where the legislature used “a [black voting-age population] threshold, as opposed to a more sophisticated analysis of racial voting patterns,” to draw district boundaries. Page v. Virginia State Bd. of Elections, No. 3:13-cv-678, 2015 WL 3604029, at *18 (E.D. Va. June 5, 2015). Likewise, when the South Carolina legislature mechanically drafted each majority-black district with a 55 % black population, another three-judge court found that the legislature’s approach was not narrowly tailored and could not survive strict scrutiny. Smith v. Beasley, 946 F.Supp. 1174, 1210 (D.S.C. 1996). That court recognized that consideration of the relevant facts “was not done in the present cases because of the insistence that all majority-minority districts have at least 55 % [black voting-age population] with no evidence as to registration or voter turnout,” facts that “should be considered when district lines are drawn” for a district to be narrowly tailored. Id.
Because the State of Alabama did not conduct an even minimally sufficient analysis about what was “reasonably necessary” *1368for it to achieve § 5 compliance, this court’s ability to scrutinize the State’s use of race has been severely, if not fatally, hampered.
Moreover, it could be reasonably argued that, because Sanders’s and Jackson’s “could be” and “sometimes” comments were generic and conclusory and were not supported by district-specific demographics, they constitute exactly the type of stereotyping about black voting behavior that strict scrutiny is intended to prohibit.
[[Image here]]
The point is not that Sanders, Jackson, or the State of Alabama reached the wrong result. It might be the case that a range of 62 to 65 % is appropriate to maintain black voters’ opportunity to elect the candidate of their choice in SD 23 and HD 68, and thus a design of those districts within that range could theoretically be narrowly tailored to meet the State’s interest in § 5 nonretrogression. Nor does it follow from my conclusion that the legislature was obligated to reject the suggestions of Sanders and Jackson, as the majority suggests. See ante, at 1108. Rather, the State, having opted to utilize race in drawing the boundaries of SD 23 and HD 68, needed to do so with sufficient support to satisfy strict scrutiny. But the State has not presented — and the record does not otherwise contain — sufficient evidence to conclude that the State’s approach was narrowly tailored. Absent an adequate record, this court cannot evaluate the State’s use of race in drawing SD 23 and HD 68. And that inability is fatal, for “blind judicial deference to legislative or executive pronouncements of necessity has no place in equal protection analysis.” Croson, 488 U.S. at 501, 109 S.Ct. 706.
Because the State has not met its burden of demonstrating that its use of race was narrowly tailored, I cannot give these districts my imprimatur.
II. Ten Additional Districts
I now turn to ten additional districts for which the majority concluded mistakenly that race did not predominate. In the racial-predominance inquiry, a plaintiff must “show, either through circumstantial evidence of a district’s shape and demographics or more direct evidence going to legislative purpose, that race was the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.” Ala. Legislative Black Caucus, 135 S.Ct. at 1267 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475) (emphasis added). As the standard’s name suggests, the use of race must be not merely present, but predominant; the drafters’ mere “consciousness of race” does not suffice. Bush, 517 U.S. at 958, 116 S.Ct. 1941 (plurality opinion). Race predominates when “race for its own sake, and not other districting principles, was a legislature’s dominant and controlling rationale in drawing district lines.” Miller, 515 U.S. at 913, 115 S.Ct. 2475. The use of racial targets “provides evidence” that race predominated, but must be accompanied by a showing that the use of a target “had a direct and significant impact on the drawing of at least some of [a district’s] boundaries.” Ala. Legislative Black Caucus, 135 S.Ct. at 1267, 1271. When determining whether race or race-neutral factors predominated, a court should take a holistic approach, reviewing the entirety of the drafters’ decisions that resulted in the district’s boundaries.
While the racial-predominance standard is rightfully demanding, the majority’s application of it errs in several ways, causing it to conclude mistakenly that race did not predominate in ten districts. First, I will explain several significant legal errors that *1369led to significant factual errors in the majority’s analysis. Then, I will explain why the evidence supports a racial-predominance finding in each of those ten districts.
A. Racial-Predominance Standard
i. Race-Neutral Districting Criteria
First, the majority goes astray when it categorically states: “When the plaintiffs proceed with only indirect evidence that race predominated and the design of a district can be explained by traditional dis-tricting criteria, the plaintiffs have not satisfied their burden of proof.” Ante, at 1044. The majority erroneously finds dispositive the mere possibility that traditional race-neutral criteria could explain a district’s boundaries. This categorical approach does not comport with the Supreme Court’s guidance.
The injury underlying a racial gerrymandering claim stems from the racial classification itself, not the classification’s outward manifestations. Miller, 515 U.S. at 913, 115 S.Ct. 2475 (observing that the “racial purpose of state action, not its stark manifestation” is “the constitutional violation”). The harms stemming from a racial gerrymander “include being personally subjected to a racial classification, as well as being represented by a legislator who believes his primary obligation is to represent only the members of a particular racial group.” Ala. Legislative Black Caucus, 135 S.Ct. at 1265 (brackets, ellipses, citations, and internal quotation marks omitted). Therefore, if a plaintiff can prove that the State predominantly relied on race when drawing district lines, a cognizable injury exists even if that classification did not distort the district’s shape or otherwise violate traditional redistricting principles. Miller, 515 U.S. at 910-914, 115 S.Ct. 2475.
Race predominates when race dwarfs other considerations and functions as an overriding factor determining the placement of district lines. See Shaw II, 517 U.S. at 906-907, 116 S.Ct. 1894 (rejecting the argument that predominance cannot occur “where a State respects or complies with traditional districting principles,” because the legislature’s ability to “address[ ] these interests does not in any way refute the fact that race was the legislature’s predominant consideration”) (brackets, citation, and internal quotation marks omitted).
Indeed, race sometimes strongly correlates with race-neutral criteria, such as partisan politics and communities of interest, including “shared broadcast and print media, public transport infrastructure, and institutions such as schools and churches.” Bush, 517 U.S. at 964, 116 S.Ct. 1941 (plurality opinion). In other words, a boundary line actually drawn for race-based reasons could easily be explained after-the-fact by pointing instead to party affiliation or a railroad boundary. Were our analysis focused merely on identifying ‘race-neutral justifications’ for a district’s boundaries, we would frequently be able to find them, even in the most egregious circumstances where race served as the one and only factor actually motivating the drafters.
Moreover, the majority’s requirement of “direct evidence” when “the design of a district can be explained by traditional districting criteria” cannot be reconciled with the thrust of the Supreme Court’s decisions in Miller or this case. In Miller, the Court stated that a plaintiff may establish racial predominance “either through circumstantial evidence of a district’s shape and demographics or more direct evidence going to legislative purpose.” Miller, 515 U.S. at 916, 115 S.Ct. 2475 (emphasis added). The Court proceeded to rely heavily on circumstantial evidence of the district’s shape and its use *1370of land bridges, together with the relevant racial demographics, in reaching its conclusion that race predominated, even though the State offered “various alternative explanations” for the district based on traditional districting principles. Id. at 918-19, 115 S.Ct. 2475. In our case, the Court reiterated the principle that circumstantial evidence alone may be enough to establish predominance. Ala. Legislative Black Caucus, 135 S.Ct. at 1267. This analysis forecloses the majority’s insistence that race predominates only when a district contains direct evidence of race-based decisionmaking or explanations based on traditional districting criteria are impossible.
As an analytical matter, direct evidence is not an inherent aspect of the constitutional harm or an essential element of a racial gerrymandering claim. Circumstantial evidence of racial intent can be just as potent as direct evidence, and the mere fact that traditional race-neutral districting principles ‘might’ explain what the legislature did does not mean that the legislature in fact predominantly utilized those principles. For example, if the evidence shows the jurisdiction set a particularly high racial target and the drafter states that he moved large numbers of voters in and out of the district to reach that target, with those moved in being almost all black and those moved out being almost all white, a court could conclude from this circumstantial evidence that race predominated, even if the district were relatively compact and consistent with other traditional redistricting principles. See Ala. Legislative Black Caucus, 135 S.Ct. at 1263 (indicating that stark demographic evidence is significant when observing that the legislature sought to maintain a black population of 72.75 % in a challenged district that added 15,785 new individuals, only 36 of whom were white).
Similarly, a showing that the legislature relied on racial data and did not consider non-racial information that might otherwise explain the challenged district’s lines could establish predominance even in the absence of direct evidence based on the drawing of district lines. See, e.g., Bush, 517 U.S. at 966-967, 116 S.Ct. 1941 (plurality opinion) (emphasizing that the State had detailed racial data available but not other data concerning traditional district-ing principles).
Accordingly, racial-predominance analysis should involve a holistic determination whether “the legislature subordinated traditional race-neutral districting principles ... to racial considerations.” Ala. Legislative Black Caucus, 135 S.Ct. at 1270 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475). The key word here is “subordinated.” That is to say, because the statewide evidence establishes that the legislature’s approach prioritized race, see, e.g., Tr. Vol. I (doc. no. 215) at 63:4-9, 124:16-22 (Dial describing avoiding retrogression as the “ultimate goal” and “first priority”), then so long as there is some evidence, direct or circumstantial, corroborating the application of that overarching approach to a district, it is appropriate to find that race predominated over race-neutral criteria. A demonstration that race-neutral factors might also explain some of a district’s boundaries does not necessarily negate that conclusion.
The majority’s analysis of SD 24 demonstrates its error: It credits the State’s argument that the district’s expansion into a portion of Pickens County that was 74 % black and exclusion of a portion of the county that was 75 % white did not count as evidence of racial predominance because the State did so using smooth lines that hugged the state border. See ante, at 1114. But that use of valid race-neutral criteria does not end the analysis. Instead, we *1371must consider evidence of the State’s application of its policy to the district and its role in the overall design of the district; in SD 24, the plaintiffs pointed to both preservation of the racial target — exceeded by just +0.44 % — and testimony from two legislators that the State was keenly aware of its mechanical targets when drawing at least some of the district’s boundaries. See infra at 1390-91. Indeed, Dial frankly admitted that SD 24 expanded into Pickens County because the district “had to have more minorities.” Tr. Vol. I (doc. no. 215) at 48:19-25. Viewed as a whole, the evidence clearly establishes racial predominance even though the State’s use of valid race-neutral criteria also affected some of the district’s boundaries.
Instead of halting the predominance analysis where traditional race-neutral criteria can explain some of a district’s boundaries, I would consider whether there is evidence, direct or circumstantial, that race affected some of the district’s boundaries and then determine whether race or race-neutral criteria predominated over those criteria when looking at the district as a whole. In equal-protection analysis, we must ask why the drafters actually made the choices they did — that is, what was their motive — regardless of the range of other possible explanations; that a drafter could have drawn a boundary in a particular way to achieve ends other than race does not prove that racial considerations did not in reality predominate over race-neutral ones in the State’s ultimate design of the district. Shaw II, 517 U.S. at 908 n.4, 116 S.Ct. 1894 (stating that strict scrutiny requires interrogation of a legislature’s “actual purpose,” not “speculation about what ‘may have motivated’ ” it). Where two possible explanations exist, those possibilities must be viewed in the context of the State’s policy of meeting racial targets if the evidence reflects that it had an overarching role in the design of the district.
ii. State’s Policy of Meeting Racial Targets
Moreover, the majority’s statement and repeated application in its district analyses that “When the plaintiffs proceed with only indirect evidence that race predominated and the design of a district can be explained by traditional districting criteria, the plaintiffs have not satisfied their burden of proof,” overlooks, or at least places too little weight, on the State’s express policy of meeting racial targets. In other words, while this policy, by itself, is insufficient direct evidence to establish racial predominance, it is still important evidence in making a determination as to whether race predominated as to any district.
As the Supreme Court found, when drawing the 2012 House and Senate plans, the State employed a policy of attempting to draw majority-black districts in such a way that the percentage of black population in the new districts matched the percentage under the district’s old lines. Ala. Legislative Black Caucus, 135 S.Ct. at 1267. The State prioritized that policy “above all other districting criteria (save one-person, one-vote).” Id.
To be sure, as stated, the existence of a racial target does not, by itself, establish predominance. A racial target may end up playing little or no role in how a particular district’s lines are actually constructed. And even when a racial target is one motivating factor in the drawing of some lines, race may still not predominate over other non-racial factors in the design of the district as a whole.
Nevertheless, the existence of a racial target does not drop out of the picture in a court’s analysis whether the racial-predominance standard is met. It “provides evidence,” Ala. Legislative Black Caucus, 135 S.Ct. at 1267 — and sometimes very com*1372pelling evidence — of the State’s use of race. But all racial targets should not be treated equally. They may vary in nature, the circumstances of their creation, and the circumstances of their application.
A legislature might use a racial target to preserve (not too many and not too few) the same percentage of minority voters in a district. A legislature might use a racial target as a floor that could be low (for example, at or slightly above 50 % minority) or high (for example, above 60 % minority). A legislature might use a racial target as a general guide or range or it might insist on mathematical precision such that the target becomes a driving force behind the district’s boundaries. Each of these factors impacts the eviden-tiary weight to be given the existence and use of a racial target.
Here, the targets were set at high percentages of minority population; they were most constraining (not too many and not too few); and reflected a rigorous intent to reach as many majority-black districts as possible. Ample testimony from the drafters provides direct evidence that the State had a policy of maintaining, to the extent practical, black percentages in every majority-black district. Hinaman explained that he “tried to draw those [majority-black] districts as close to the [black percentage] numbers as possible and practicable as they were in the 2001 plan.” Tr. Vol..Ill (doc. no. 217) at 164:4-7. Senator Dial testified that avoiding retrogression was the “ultimate goal” and “first priority.” Tr. Vol. I (doc. no. 215) at 63:4-9, 124:16-22. Dial instructed Hinaman that, “We didn’t want to regress any of the [majority-black] districts” and, “We wanted to make sure the [majority-black] districts stayed as they were and we did not do away with any and that the population, as they grew, that they grew into the same proportion of minorities that they originally had or as close to it as we could get it.” Dial Dep., APX 66 (doc. no. 125-3) at 17:7-22 (emphasis added). Representative McClendon explained the same approach for House districts: “[W]e tried to look at the 2010 census, overlay it on the districts, and try not to change the percentages of the citizens, the black citizens, in a district any more than we had to. Tried to keep them in about the same proportion as they were.” Tr. Vol. Ill (doc. no. 217) at 222:8-12. Hinaman explained his focus on maintaining the black percentage in each district, which superceded the race-neutral criterion of keeping precincts whole: He “may add a white precinct, a majority white precinct and a majority African American precinct; but if you look at the end number, if it did not retrogress the overall end number [in that district], then they were added in. If for some reason they retrogressed that number, then ... [he] would split precincts.” Tr. Vol. Ill (doc. no. 217) at 144:2-7.
Indeed, where other three-judge panels have applied Ala. Legislative Black Caucus to Shaw claims, they have each recognized that racial targets deserve significant weight. See Covington v. North Carolina, 316 F.R.D. 117, 140 (M.D.N.C. 2016) (three-judge court) (concluding, after describing legislative purpose to achieve racial target of 50 % plus 1 in majority-black districts, that “[f]or all the challenged districts, the overwhelming statewide evidence provides decisive proof that race predominated.” (emphasis added)); Harris v. McCrory, 159 F. Supp. 3d 600, 610 (M.D.N.C. Feb. 5, 2016) (three-judge court), prob. juris, noted, — U.S. -, 136 S.Ct. 2512, 195 L.Ed.2d 838 (2016) (concluding that race predominated where “overwhelming evidence in this case shows that a BVAP [black voting age population-floor, or a racial quota, was established in [the challenged districts]”); Bethune-Hill v. Va. State Bd. of Elections, 141 *1373F.Supp.3d 505, 524 (E.D. Va. 2015) (three-judge court), prob. juris, noted, — U.S. -, 136 S.Ct. 2406, 195 L.Ed.2d 779 (2016) (“[Ala. Legislative Black Caucus] could not be clearer that use of racial BVAP floors constitutes evidence — albeit significant evidence — of predominance.”).14
Moreover, one must also not lose sight of the big picture presented by circumstantial evidence. Here, the drafters hit their racial target within three percent in 21 of 36 majority-black districts, and even more precisely within one percent of the target in 16 of those districts, including all ten districts for which I conclude that race predominated but the majority does not. If something happens once, it could be a coincidence; if it happens again and again and again — indeed, most of the time — that is unmistakable evidence of intent. The fact that, over and over again, the drafters hit their racial target provides very compelling circumstantial evidence that it is what they intended to do.
Finally, my approach — which considers the nature of the racial target, the circumstances of its creation, and the circumstances of its application — does not result in a “per se” finding of racial predominance or a “rebuttable presumption,” ante at 1056, 1197, as I agree with the majority that race did not predominate in 12 districts in large part because this circumstantial evidence does not apply in those districts. Indeed, I am not saying that my approach compels a finding of racial predominance; I just think the evidence should not be ignored.
in. Alternative Plans
The majority states that, “If an alternative district has identical lines, we take the plaintiffs’ offering of that plan as a concession that race did not predominate in the enacted district.” Ante, at 1049. The majority overstates its case.
In racial-discrimination cases, whether the defendant intentionally discriminated turns on information the defendant considered before its decision. Thus, in McKennon v. Nashville Banner Pub. Co., 513 U.S. 352, 115 S.Ct. 879, 130 L.Ed.2d 852 (1995), the Supreme Court rejected an after-the-fact justification for the challenged action because “[t]he employer could not have been motivated by the knowledge it did not have and it cannot now claim that the employee was fired for the nondiscriminatory reason.” Id. at 360, 115 S.Ct. 879. While the inquiry is more challenging when a Shaw claim depends on circumstantial evidence of legislative intent, the same equal-protection principle applies: what ultimately matters is the State’s intent at the time it drew district lines. See, e.g., Shaw II, 517 U.S. at 910, 116 S.Ct. 1894 (refusing to credit State’s reliance on historical and social science reports to justify its race-based, approach because reports were created after districts had been drawn). Alternative maps generated only after the enacted districts have been drawn do not necessarily preclude a finding that race predominated where the plaintiff has otherwise met its burden. Because, as I have indicated previously, application of race-neutral criteria can sometimes create the same result as racial ones, see supra at 1369, alternative plans do not always dispositively establish that race did not predominate, especially in the face of clear evidence, direct or circumstantial, of the legislature’s intent.
*1374Relying on Easley v. Cromartie, 532 U.S. 234, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001), the State argues that the plaintiffs “ ‘must show at the least that the legislature could have achieved its legitimate political objectives in alternative ways that are comparably consistent with traditional districting principles’ and ‘that those dis-tricting alternatives would have brought about significantly greater racial balance.’ ” Def. Remand Br. (doc. no. 263) at 19 (quoting Easley, 532 U.S. at 258, 121 S.Ct. 1452). The State’s argument distorts Easley’s evidentiary rule, applicable where the racial-predominance inquiry depends entirely on circumstantial evidence, and transforms it into a legal element of all Shaw claims, including those with direct evidence of racial purpose.
This case is quite different from Easley, a case in which the State defended a racial-predominance claim with persuasive evidence that political gerrymandering, rather than race, was the predominant motive of the legislature, and had also demonstrated that race and political party closely correlated. Easley concerned a North Carolina Congressional district that had been' redrawn after a finding that race unconstitutionally predominated in Shaw II. The primary issue was whether the legislature was predominately motivated by race or partisan' politics when it redrew the district. The district court’s conclusion that race predominated depended on an exhaustive examination of circumstantial evidence, including precinct splits, voter registration data, and alternative decisions the State could have made, to parse whether the legislature was predominately motivated by race or by creating a safe partisan district. Reviewing the district court’s factual findings, the Supreme Court emphasized that “[cjaution is especially appropriate in this case” because “the State ha[d] articulated a legitimate political explanation for its districting decision, and the voting population is one in which race and political affiliation are highly correlated.” Easley, 532 U.S. at 242, 121 S.Ct. 1452. In those circumstances, plaintiffs “must show at the least that the legislature could have achieved its legitimate political objective in alternative ways that are comparably consistent with traditional districting principles. That party must also show that those districting alternatives would have brought about significantly greater racial balance.” Id. at 258, 121 S.Ct. 1452.
Easley does not apply to a ease, like this one, where direct evidence of legislative intent indicates that the State “expressly adopted and applied a policy of prioritizing mechanical racial targets,” Ala. Legislative Black Caucus, 135 S.Ct. at 1267, rather than pursuing a policy that prioritized partisan politics. While the State now attempts to justify its actions on political grounds, it previously relied on evidence of legislative intent which reflected prioritization of compliance with its understanding of § 5 nonretrogression. Indeed, instead of presenting evidence that the legislature pursued a legitimate political gerrymander, the State has consistently highlighted evidence showing that the Republican-led drafters accommodated the requests of Democrats. See Def. Remand Br. (doc. no. 263) at 19 n. 1 (cataloguing testimony that drafters incorporated requests of incumbent Democrats). The Supreme Court recognized as much in our case, citing Easley only as a contrast to the relevant predominance standard of Miller, and noting that Easley explained “the plaintiffs burden in cases, unlike these, in which the State argues that politics, not race, was its predominant motive.” Ala. Legislative Black Caucus, 135 S.Ct. at 1267 (emphasis added). Three-judge panels have agreed that Easley does not apply in circumstances comparable to those here. See, e.g., Covington, 316 F.R.D. at 139 n.21 (three-judge *1375court) (citing evidence that State defended partisan results of redistricting on grounds of Voting Rights Act compliance as “easily distinguish[ing] this case from the Cromar-tie cases, where there was substantial direct evidence supporting the State’s ‘legitimate political explanation for its districting decision’ ” (quoting Easley, 532 U.S. at 242, 121 S.Ct. 1452)).
iv. District-Specific Evidence
The majority incorrectly disregards district-specific circumstantial evidence that reflects actions that reduced the proportion of, or inconsistently allocated, black residents. Rather than looking for corroboration that the State brought its racial target to bear in a particular district, the majority treats the plaintiffs’ claim as if it were based on a packing concept, citing any actions the drafters took to reduce the black population to repeatedly conclude that race did not predominate in a particular district. But because this case is not about packing, but rather is about preserving majority-black population percentages (not too many and not too few), the majority’s approach is wrong.
For example, despite committee guidelines to keep precincts whole, the majority considers precinct splits suspicious only if the precinct is split between a majority-black and majority-white district where the majority-black district received a greater percentage of black people. See ante, at 1059 (concluding that precinct splits alone do not create inference of racial predominance, in part because several splits “placed a higher black population percentage in a majority-white district than a majority-black district”). Frequently, however, where the State achieved a racial target exactly, its calculation of the racial makeup of that district without split precincts demonstrates that it would not have been able to hit its target without splits. For example, in HD 52, the State accomplished its racial target (60.11 %) almost exactly (60.13 %); without the use of six split precincts, the State would have exceeded its target significantly, raising the black percentage to 64.50 %. Def. Precinct Split Analysis (doc. no. 263-3) at 5. Nonetheless, the majority accorded no weight to the precinct splits because they did not demonstrate a “clear pattern.” See ante, at 1197. Although the splits reduced the black percentage of the population, they ultimately helped the drafters reach their target; the precinct splits therefore provide evidence that race predominated when viewed in the context of the statewide policy.
The crucial problem with the drafters’ approach is not that they used non-traditional districting criteria to create improperly majority-minority districts, but that they disregarded traditional race-neutral criteria to achieve racial targets in their maintenance of existing majority-minority districts. When conducting a district-specific analysis of racial predominance, that statewide goal must guide our approach. Circumstantial evidence that the State ignored traditional race-neutral districting criteria to inch closer toward its racial target — even if it lowered the percentage of black population in the district — serves as powerful evidence that the State’s racial target was brought to bear in a district.
Historically, the Court has identified evidence of racial predominance in a Shaw claim where a legislature’s disregard of traditional race-neutral criteria placed more minority than white citizens in a challenged district. But, in each prior case, the racial-gerrymandering claim involved a challenge to the creation of a majority-minority district, which necessarily involves efforts to add high percentages of minority population. See Shaw I, 509 U.S. at 635-636, 113 S.Ct. 2816 (citing evidence that district “winds in snakelike fashion *1376through tobacco country, financial centers, and manufacturing areas until it gobbles in enough enclaves of black neighborhoods” (internal quotation marks and citation omitted)); Miller, 515 U.S. at 917, 115 S.Ct. 2475 (affirming racial-predominance conclusion based in part on evidence “of narrow land bridges to incorporate within the district outlying appendages containing nearly 80 % of the district’s total black population”); Bush, 517 U.S. at 973, 116 S.Ct. 1941 (plurality opinion) (relying on evidence that district’s “many narrow corridors, wings, or fingers ... reach out to enclose black voters, while excluding nearby Hispanic residents.” (internal citation and quotation marks omitted)).
In our case, in contrast, the plaintiffs’ racial-predominance theory rests on direct evidence concerning the drafters’ method of maintaining (not too many and not too few) the percentage of blacks in majority-minority districts using mathematically precise ways. Here, a racial-predominance finding could be supported by evidence that the drafters applied a race-based approach to achieve a racially precise allocation of residents in existing majority-black districts. In some districts, the drafters needed to add large amounts of new black population to achieve that goal; but to achieve that goal in other districts, the drafters needed to add new white population so they could reduce the black population percentage, or they needed to sweep in and out large amounts of populations of both races. In other words, evidence that the State swooped large amounts of black population, into a district is not the only type of evidence that supports a racial-predominance finding here.
To be sure, the Supreme Court’s model analysis based on SD 26 also looked to the. exceptionally high percentages of black population added to the district using nontraditional districting criteria. The Court described the drafters’ “remarkable feat” of adding a mere 36 white individuals to the district despite adding a total population of 15,785. Ala. Legislative Black Caucus, 135 S.Ct. at 1271. The Court also noted that precincts split between a majority-black and a majority-white district were “clearly divided on racial lines,” with heavily black portions of the precincts added to majority-black SD 26 and heavily white portions of the precincts added to majority-white SD 25. Id.
But the Court also recognized the importance of the drafters’ approach to determining the racial composition of existing majority-minority districts. The Court approvingly cited the majority’s prior conclusion that the legislature “ ‘preserved’ ‘the percentage of the population that was black,’ ” Ala. Legislative Black Caucus, 135 S.Ct. at 1271 (quoting Ala. Legislative Black Caucus, 989 F.Supp.2d at 1306). SD 26 had a remarkably high 75.13 % black population, the district with the second highest black percentage of any majority-black district in Alabama. See District Statistics Report, Def. Ex. 400 (doc. no. 30-39) at 3. As such, in that district, preserving the black percentage of the population required obvious disregard of race-neutral criteria to add sufficient black population. But preserving a black percentage with mechanical precision does not only require adding black population; it requires achieving a racial balancing that sometimes involved the use of precinct or county splits, irregular appendages, and land bridges to discard ‘excess’ black population that would have otherwise left the State too far afield of its target. Where a district’s shape is “unexplainable on grounds other than the racial quotas established for those districts,” evidence of “utter disregard for traditional redistricting criteria” provides strong evidence that race predominated — regardless of whether that evidence reflects an additive or subtractive *1377effect on the minority population in that district. Bush, 517 U.S. at 976, 116 S.Ct. 1941 (plurality opinion) (citation and quotation marks omitted).
By any measure, the State routinely disregarded race-neutral districting criteria to achieve a racial target, even if its failure to apply race-neutral criteria reduced the minority percentage in the district,
v. “Significant Number of Voters”
Finally, the majority used an overly rigid numerical minimum in its predominance analysis. To establish racial predominance, the plaintiffs “must show that ‘race was the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.” Ala. Legislative Black Caucus, 135 S.Ct. at 1270 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475) (emphasis added). The Supreme Court has never quantified a “significant number of voters.” Frequently, the Court has relied on dramatic statistics that made a racial-predominance conclusion obvious in that case, but provides little clarity for three-judge panels required to navigate “the complex interplay of forces that enter a legislature’s redistricting calculus.” Miller, 515 U.S. at 915-16, 115 S.Ct. 2475. For example, in Miller, the Court approvingly cited the panel’s conclusion that the use of a narrow land bridge to incorporate 80 % of a district’s total black population made it “exceedingly obvious” that the drafters deliberately used race. M. at 917, 115 S.Ct. 2475. In Bush, the plurality found “utter disregard of ... local election precincts” where 60 % of the challenged districts’ residents lived in split precincts. 517 U.S. at 974, 116 S.Ct. 1941.
In our case, the Court likewise did not clarify the statistical minimum extent of racial impact. The Court’s model analysis concluding that race predominated in SD 26 relied on seven precinct splits, “with the population in those precincts clearly divided on racial lines,” without quantification of that racial division. Ala. Legislative Black Caucus, 135 S.Ct. at 1271. The Court also cited the relative proportion of new population, noting that adding just 36 white individuals out of 15,785 total residents was a “remarkable feat.” Id.
Other three-judge panels have concluded that race predominated in districts where race explained the placement of only a small percentage of a district’s population. In fact, facing evidence that a white incumbent legislator was excluded from a district because of her race, the panel contemplated the possibility that this alone was enough for race to predominate: “It is not clear whether the addition or removal of one person on the basis of race could be sufficient to establish that race predominated in the drawing of a district.” Covington, 316 F.R.D. at 149 n.28 (three-judge court). Nonetheless, the panel concluded that a “significant number of voters” had been excluded where direct evidence established that the legislature arbitrarily increased a recommended minority percentage by less than one percentage point (0.58 %). Id. Another panel in that district described a 6.89 % increase in the bláck voting age population as “whopping,” and, where it corroborated direct evidence of legislative intent to achieve a racial quota, sufficient to establish racial predominance. Harris, 159 F.Supp.3d at 618 (three-judge court).
Sometimes, my racial-predominance analysis considers the number or proportion of voters in a given district that were placed within or without the district because race-neutral criteria had been ignored. But the degree the drafters hit their racial target also matters here. The Supreme Court approvingly cited, as evidence of racial predominance, the majority’s previous finding that SD *137826 “preserved” its percentage of minority population, where the district had a 72.69 % black population and was redrawn with a 75.13 % black population. Ala. Legislative Black Caucus, 135 S.Ct. at 1271. In other words, the Court agreed that coming within 2.44 % of a racial target provided circumstantial evidence of racial predominance. In each of the districts I discuss below, the State came within less than 1 % of hitting its racial target.
Although the Court and prior panels offer statistical guideposts for the predominance analysis, I proceed not with unwavering reliance on a calculator and statistically precise determinations, but with adherence to first principles. Race-conscious redistricting that separates a “significant number of voters” or “preserves” the minority percentage in a district obtains constitutional significance because it signals state-sponsored endorsement of the idea that voters may be identified and sorted principally based upon race. In Shaw I, the Supreme Court applied equal-protection principles to redistricting because it recognized that legislative action that classifies citizens based on race is inconsistent with principles of equality and stigmatizes individuals solely by reason of their membership in a racial group. 509 U.S. at 643, 113 S.Ct. 2816. Á district warrants a racial predominance-finding where race had a significant enough impact to realize these harms.
B. Application of Racial-Predominance Standard
Having laid out my disagreements with the majority’s approach to racial predominance, I will now turn to the individual districts and, applying the Supreme Court’s guidance in this and past cases, explain why I would reach different conclusions from the majority’s on a number of the challenged districts.
In assessing each district, I will follow the three-step approach the Supreme Court used in assessing SD 26: First, I start with direct evidence reflecting the “primary redistricting goal [ ] to maintain existing racial percentages in each majority-minority district, insofar as feasible”; second, I look at circumstantial evidence in the record to see whether “this goal had a direct and significant impact on the drawing of at least some of [the district’s] boundaries”; and, third, I look for evidence of race-neutral factors actually considered by the drafters and determine whether those factors, or race-conscious decisions, predominated in the overall design of the district. Ala. Legislative Black Caucus, 135 S.Ct. at 1271.
In the discussion of individual districts that follows, I focus in particular on the following corroborating circumstantial evidence: (1) how closely the drafters hit the prior percentage of black population in each district; (2) the frequency of county and precinct splits, and the comparative racial composition of those splits; (3) the ratio of blacks to whites added to the district; and (4) the shape of the district. See Ala. Legislative Black Caucus, 135 S.Ct. at 1271 (assessing each factor). Like the majority, I also weigh the net population change relative to the level of underpopulation — in other words, I examine whether the State reshuffled the district’s inhabitants significantly more than its equal-population objectives required.
Moreover, with regard to the ten districts for which I conclude that race predominated, it must be kept in mind that evidence that the drafters hit, within one percentage point, their admitted target of maintaining existing racial percentages paints a clear picture that the racial percentage in each district was specifically intended. As previously stated, if something happens once it may be coincidence, *1379but, if it happens again and again and again, that is unmistakable evidence that it was intended.
i. Jefferson County (HDs 52, 55, 56, 57, and 60)
I conclude that race predominated in five Jefferson County House Districts: HDs 52, 55, 56, 57, and 60. At the heart of assessing the drafters’ approach to race in Jefferson County — and indeed, across the State — was the decision to extract a majority-black House district from the County (HD 53) and move it far north to create a new majority-black district composed entirely of Madison County residents. Although the majority and I agree that race predominated in HD 53, the majority bases its decision solely on factors specific to the district’s new location in Madison County. In doing so, the majority fails to account for testimony from the drafters that serves as direct evidence that race played a central role in their decision to unnecessarily reshuffle residents of Jefferson County House districts based on an incorrect interpretation of § 5 compliance. Under the 2001 map, HD 52, 55, and 60 were each contiguous to HD 53 and, as a result, experienced substantial unnecessary population change under the new plan as each inherited black population from HD 53. HD 56 and 57, although located southwest of HD 53 in the 2001 map and separated from it by HD 52, also experienced substantial population change because of the significant reshuffling of Jefferson County’s black population.
Beyond this direct evidence, significant corroborating evidence that race predominated in the drafters’ approach may be found in each district: precise matches of racial targets; substantial reliance on split precincts, without which those targets would not have been achievable; the movement of significantly more population in and out of the district than necessary; and sometimes irregularly shaped districts. These factors contribute to my conclusion that race predominated in five Jefferson County House districts not recognized by the majority.
a. Impact of Racial Predominance in HD 53 on Surrounding Districts
Although, as stated, I agree with the majority that race predominated in the drawing of HD 53, I reach that conclusion on different grounds. Race played a significant role in the drafters’ decision to move HD 53 from Jefferson County. That transfer, which enabled House districts within Jefferson County to achieve mathematically precise racial targets, should be given significant weight in the predominance analysis of its surrounding districts.
As I previously explained, the drafters created for themselves an enormous challenge through their very specific (but incorrect) understanding of § 5 compliance, combined with a no-greater-than-2 % ideal population-size deviation. See Ala. Legislative Black Caucus, 989 F.Supp.2d at 1314-17 (Thompson, J., dissenting). Because of the reapportionment committee’s decision to tighten significantly the deviation, many majority-black districts required the addition of significant black population to avoid ‘under-population.’
Nowhere was the effect of that unnecessary challenge felt more acutely than Jefferson County, where compliance with the drafters’ approach required the addition of approximately 70,000 black people to the county’s existing majority-black House districts — -the equivalent of a district and a half. Tr. Vol. Ill (doc. no. 217) at 132:22-25-133:1. Because Hinaman “knew that most of the — if not all of the minority neighborhoods were already included in those districts,” he concluded that “trying to repopulate them to get them back to deviation was going to retrogress most if not all of them.” Id. at 133:1-5; see also *1380McClendon Dep., APX 67 (doc. no. 125-4) at 100:12-14 (testifying that “[a]ll of the districts — Jefferson County minority districts were underpopulated, and we had to get the people from somewhere”). To overcome that self-inflicted challenge, Hinaman recommended and the legislature ultimately adopted a plan that moved HD 53 from Jefferson County to Madison County, reducing by one the number of majority-black House districts in Jefferson County. Hinaman then used the black Jefferson County residents living within former HD 53 to add population to the remaining majority-black districts in the area.
The majority concludes that the relocation of HD 53 “provides no evidence that race predominated,” although it finds that race still predominated in the drawing of the district because of five suspicious precinct splits specific to the district’s new location in Madison County. See ante, at 1166, 1180-85. For the majority, HD 53’s move can be explained solely on the ground of “massive underpopulation.” Id. at 1192. To the drafters, however, the tight deviation that resulted in such dramatic under-population in Jefferson County was inseparable from their mechanical view of retrogression.
The drafters’ incorrect approach to retrogression meant that even Jefferson County’s majority-black districts with a population well-above 50 % black, such as HD 55 (73.55 % black) and HD 57 (68.42 % black), required a massive influx of black population to avoid even the slightest reduction in the percentage of black population. The drafters appeared to be completely oblivious to the ripple effects caused by unnecessarily high requirements for black population in any given district— and completely oblivious to the level of black population actually necessary to elect black voters’ candidates of choice. When Representative McClendon was asked if he had been advised that certain percentages of minority population in specific districts would be too high and could raise concerns about packing, McClendon testified that he had not. McClendon Dep. (doc. no. 125-4) at 106:10-18 (“I haven’t been advised that if you go over a certain number [of minorities out of the total population in a district], you’re in trouble.”). Senator Dial likewise testified that he “did not consider any [black percentage] too high,” and agreed that his objective was to construct “guaranteed black safe districts,” without any apparent concern about packing. Tr. Yol. I (doc. no. 215) at 56:5-12, 18-21.
*1381[[Image here]]
Jefferson County House Districts, 2001 Map, showing HD 53 at its center. Def. Ex. 480 at 2 (excerpt).
[[Image here]]
*1382Jefferson County House Districts after redistricting, reflecting departure of HD 53. Def. Ex. 479 at 2 (excerpt).
District-specific evidence corroborates the drafters’ intent to flood Jefferson County House districts with new black population to preserve the existing percentages: six of the eight majority-black Jefferson County House districts reached a black percentage of the population within one percentage point of the target, reflecting an unmistakable pattern that stretches credulity to view as a coincidence.15 Despite these efforts to achieve mathematical precision, the record reflects not even a modest effort by the drafters to determine the actual percentage necessary for black voters in Jefferson County to have the opportunity to elect their candidate of choice. See 52 U.S.C. § 10304(b). Had the drafters undertaken this analysis, they would have concluded that black voters in Jefferson County districts would have safely maintained the opportunity to elect their candidate of choice with a lower percentage of black population, in part because of support from other minorities and whites for the preferred candidate of black voters. Lichtman Witness Summary (doc. no. 168-1) ¶ 25 & Table 7 (concluding, as to illustrative alternative districts, that 55.7 % black population would have maintained safe black opportunity to elect in Jefferson County).16 Further, the drafters’ approach made no attempt to preserve all of Jefferson County’s majority-black districts by eliminating unnecessarily high black percentages. The ALBC plaintiffs’ 1 % House Plan demonstrates that doing so was possible: it preserved all nine majority-black House districts in Jefferson County, including HD 53, by reducing the percentages of black population in certain districts. See ALBC Plfs. Comparison of Black Percentages and Deviations, APSX 492. For example, the plan drew HD 52 with a 56.64 % black population (instead of 60.13 % in the State’s plan); HD 55 with a 57.85 % black population (instead of 73.55 % in the State’s plan); and HD 60 with a 53.49 % black population (instead of 67.68 % in the State’s plan). Id.
Because “race was the predominant factor motivating the legislature’s decision to place a significant number of voters ... without” HD 53, it was also the predominant factor motivating the decision to place those same voters “within” the remaining majority-black districts of Jefferson County. Ala. Legislative Black Caucus, 135 S.Ct. at 1270 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475) (emphasis added).
b. HD 52 (Central Birmingham)
The State’s devotion to a racial target is convincingly demonstrated in HD 52. The State matched the percentage of black population nearly identically: from 60.11 % black in 2001 to 60.13 % in 2012. That feat is extraordinary in and of itself, but takes on added significance given the extent of the changes to the district. Although the district was under-populated by a relatively modest 2,362 people, see District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 4 — the least under-populated of any majority-black Jefferson County district — the State nonetheless moved 19,284 people in and out of the district and still achieved virtually the exact same racial proportion. House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4. In other words, the drafters moved eight times more people than necessary to address under-popu*1383lation, and shifted a population 43 % of the size of the district’s total population (45,-083) to accomplish its racial target. Id.; District Statistics Report, Def. Ex. 403 (doc. no. 30-36) at 4. The State’s quota required an additional 1,157 black people; the drafters added a net total of 1,134. House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4. This feat “represents racial sifting down to the finest level, a racial exactitude that would be admirable in its skill if it were not illegal.” Ala. Legislative Black Caucus, 989 F.Supp.2d at 1330 (Thompson, J., dissenting).
Nonetheless, the majority concludes that the percentage match was not “suspicious” because the alternative plans were close enough to the same racial percentage, and two proposals even reflected a higher percentage of black population than the .enacted plan. See ante, at 1194-95. But there is a crucial distinction between a district with a 55.64 % black population under the ALBC 1 % Plan or 57.90 % black population under the Democratic Conference Plan and the ultimately enacted plan with a 60.13 % black population. Beyond the principle that alternative plans created after the challenged redistricting shed little light on the drafters’ intent, the difference in percentages in the alternative plans— compared to the near-identical black percentage in the enacted plan — belies this argument. While the enacted plan was only 0.02 % removed from the targeted percentage of black population, achieving a mathematical precision that creates a strong inference of intent, the plaintiffs’ post-remand plans were 4.47 % and 2.21 % away from the target, which suggests only that demographic realities in Jefferson County led to a result in the same ballpark.
The State also used six split precincts in HD 52 to achieve the nearly identical racial percentage. The majority concludes that split precincts in HD 52 “prove nothing with respect to race,” explaining that the two precincts split between majority-black and majority-white districts contain nearly identical percentages of black population, while the remaining precincts divided between majority-black districts display no clear pattern. Ante, at 1196. But the six precinct splits incorporated into HD 52 are notable not because the State used them to pack black people into existing majority-black districts — as the majority’s disregard for this evidence assumes — but because, without them, the State would have been too far afield of its racial target. The State reports that unsplitting all of the split precincts in HD 52 would increase the black percentage in HD 52 from 60.13 % to 64.50 %. Def. Precinct Split Analysis (doc. no. 263-3) at 5. As the majority has itself concluded as to another district, precinct splits that “appear[ ] to make the difference between hitting and missing the target” contribute to a finding of racial predominance. See ante, at 1197 (explaining basis for finding racial predominance as to HD 54). Because the drafters were not concerned if majority-black districts were packed with too many black people, see supra, at 1380, the only logical inference to explain their use of split precincts to lower the black population percentage is that the race-neutral districting principle and committee guidelines that required keeping precincts whole were subordinated to the goal of meeting the drafters’ racial target.
As HD 52 demonstrates, the crucial improper feature of the drafters’ approach is not that it used non-traditional districting criteria to place significantly more black people into majority-black districts (although it may well have done that in certain districts), but that it used non-traditional criteria to achieve racial targets. Because the majority weighs evidence of *1384the former but not evidence of the latter m its predominance analysis, it fails to identify clear examples of racial predominance. By any measure, that latter approach— using non-traditional criteria to achieve a racial target — was brought to bear in the drawing of HD 52.
c. HD 55 (Northwest Birmingham)
The story is similar in HD 55. The drafters inherited a district that was 73.55 % black, and they spit out a district that remained exactly 73.55 % black See Minority Percentages Comparison (doc. no. 236-2) at 2. Even though the district borders HD 52 (which has a black population more than 13 % lower than HD 55), and even though HD 55 experienced substantial geographic and population change, its proportion of black residents did not. The State managed to accomplish its racial target to a tee even though HD 55 was among the most under-populated districts, 9,949 people (or 21.86 %) short of the ideal population. District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 5. The State moved in and out 28,143 people to reach its target, with a net total of 6,967 blacks (73 %) and 730 whites (8 %) added to the district. Minority Percentages Comparison (doc. no. 236-2) at 2; House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4.
[[Image here]]
Geographically, the new district became substantially more irregular, transformed into a thin, snake-like band that runs primarily north-south except for the addition of a portion of former HD 53 in its new southeastern corner. Map, Def. Ex. 479, at 2. HD 55 includes four precincts that had been encompassed by the former HD 53 and are now shared with other Jefferson County majority-black districts: Center Street Middle School, Glen Iris Elementary School, Legion Field Lobby, and Ramsey Alternative High School. 2001 House District Precinct Splits, Def. Ex. 413, at 31; 2012 House District Precinct Splits, Def. Ex. 405, at 106-07.
The State attributes the high black percentage in HD 55 to black population growth specific to “this portion” of Jefferson County. Def. Remand Br. (doc. no. 263) at 120. But the State overlooks that the new HD 55 is substantially different geographically under the 2012 plan than it was in 2001. While the district previously extended further west along the natural boundary of Birmingport Road (State *1385Highway 269), the district’s western stretches were engulfed by HD 16 and HD 57. In addition, the southern portion of the district now extends further east into what had been HD 53. Compare 2014 Map, Def. Ex. 479, to 2001 Map, Def. Ex. 480. Of the 19 precincts that comprise the new HD 55, 12 precincts were not within the district in 2001. 2012 House District Precinct Splits, Def. Ex. 405, at 106-07; 2001 House District Precinct Splits, Def. Ex. 413, at 32-33. The State’s attempt to justify a race-neutral approach, like the lines of HD 55, rests on shaky ground.
Because it is impossible to explain the State’s preservation of the identical racial percentage in HD 55 for reasons other than the State’s laser focus on race and because applying that percentage target significantly affected the district’s lines, I conclude that race predominated over traditional race-neutral factors in the drawing of HD 55.
d. HD 56 (Southwest Birmingham)
Once again, the drafters hit their racial quota nearly spot on: HD 56 was 62.13 % black before redistricting and 62.14 % under the new plan. The drafters drew a district just four black people removed from its target. In light of the 14,241 people that entered or exited the district, that mathematical ..precision qualifies as stunning circumstantial evidence that the statewide policy was brought to bear in HD 56. The drafters shifted substantial residents in and out of the district despite its relatively manageable under-population of 4,457 people — movement of more than three times the population size necessary to cure its under-population. While doing so, the State added a net total of 2,489 blacks to the district compared to just 906 whites: 62.12 % of its new population was black, matching the district’s 62.13 % racial target almost exactly. House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4.
Like other Jefferson County majority-black districts, the legislature placed many black people “within” HD 56 because of the transfer of HD 53 to Madison County. See Ala. Legislative Black Caucus, 135 S.Ct. at 1270 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475). Although HD 56 was not contiguous to the former HD 53 and thus its new boundaries did not encompass any of that transferred district’s residents, HD 56 swallowed many precincts from neighboring HD 52, which in turn replenished its supply from the previously contiguous, transferred HD 53. See 2012 House District Precinct Splits, Def. Ex. 405, at 99-101, 108-09; 2001 House District Precinct Splits, Def. Ex. 413, at 31; 2001 Map, APX 40 (doc. no. 60-26); 2012 Map, APX 41 (doc. no. 60-27).
The majority makes much of the fact that alternative plans drawn by the plaintiffs appeared geographically similar to the enacted HD 56, while the proposed black percentages fell within four points of the enacted plan. See ante, at 1206-08. Even according some weight to the alternative plans, there is a difference at the magnitude of 100 between the enacted plan, which falls 0.01 % away from the target, and the alternative plans, with even the tightest fit falling 1 % away from the target. One plan and one plan only — the State’s — demonstrates remarkable proximity to a racial target; the alternative plans, even if in the same ballpark, are not so precise that they are “unexplainable other than by race.” Miller, 515 U.S. at 917, 115 S.Ct. 2475.
The majority and I agree that the shape and compactness of HD 56 are not suspicious. We also agree that the manmade boundaries apparently used to split the Hunter Street Baptist Church precinct, see Precinct Map, APSX 114, relied on by *1386the ALBC plaintiffs, likewise demonstrate another valid race-neutral criterion that defines at least some of the district’s boundaries. See Katherine Inglis Butler, Redistricting in a Post-Shaw Era: A Small Treatise Accompanied by Districting Guidelines for Legislators, Litigants, and Courts, 36 Univ. Rich. L. Rev. 137, 218 (2002) (“[P]recinct boundaries generally correspond to known landmarks and encompass reasonably compact geographic areas.”). Nonetheless, HD 56’s four split precincts allowed the district to hit its racial target: without doing so, the district would have been 61.34 % black, or 0.79 % above the target. Def. Precinct Split Analysis (doc. no. 263-3) at 9. Splitting precincts allowed the drafters to minimize that discrepancy to an almost unnoticeable 0.01 %.
The plaintiffs have established that the State matched its target racial percentage while addressing significant under-population, in the process shifting many more people within and without the district than necessary. This provides sufficient evidence that race had a “direct and significant impact” on the drawing of HD 56’s boundaries. Ala. Legislative Black Caucus, 135 S.Ct. at 1271.
e. HD 57 (West Birmingham)
HD 57 provides yet further evidence of the drafters’ success at reaching their racial target in Jefferson County following the transfer of HD 53. With a racial target of 68.42 %, the State managed to enact a plan with a 68.47 % black population in the district, leaving it just 20 black people over an exact match. District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 5; District Statistics Report, Def. Ex. 403 (doc. no. 30-36) at 5. The drafters did so despite under-population of 9,322 people below the ideal population, or 20.48 % of the entire district. District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 5. To accomplish those goals, the drafters moved 21,590 people in and out of the district. House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4.
[[Image here]]
Geographically, HD 57 had been compact under the old plan, but the drafters constructed an elongated district that runs north and south adjacent to and to the *1387west of HD 55. See Map, Def. Ex. 479. In shape, the new district appears something akin to the naturally occurring landmass of the United Kingdom, with southwestern portions alternately jutting into and away from majority-white HD 15 (instead of toward Ireland), and a southeastern tip that extends so far east it might have obviated the need for the Channel Tunnel. I do not agree with the majority that its shape remains “compact” and is “not suspicious,” see ante, at 1210-11, a criterion required under the guidelines. Reapportionment Comm. Guidelines (doc. no. 80-4) at 3 (requiring, pursuant to Alabama Constitution, that “[ejvery district should be compact.”).
The majority notes but accords little, if any, weight to the “almost unchanged” black population percentage, focusing its attention instead on alternative plans that purportedly prove the resulting percentage is not suspicious. Were my analysis to focus on alternative plans, I would reach the opposite conclusion: the State’s plan (at 68.47 % black) met its target (68.42 %) so precisely while the alternative plans fell several percentage points away (72.51 % and 66.1 %), providing strong evidence that race predominated in the State’s plan. Comparison of New Plans with 2012 Plans, Def. Supp. Ex. 64 (doc. no. 296-1) at 2.
The majority discounts the five precinct splits used to construct HD 57, including a split of the Pleasant Grove First Baptist Church precinct with majority-white HD 15. Through this split, the drafters placed a 51 % black population in majority-black HD 57 and a 22 % black population in majority-white HD 15, but the majority found that racial allocation non-problematic because the split was accomplished through a “relatively smooth J-shaped line.” Ante, at 1212. The reapportionment committee’s guidelines do not support that result: valid criteria include respecting “voting precinct boundaries,” Reapportionment Comm. Guidelines (doc. no. 30-4) at 3-4, but does not include an exception if those boundaries can be split through smooth lines. Accordingly, because this criterion, although race-neutral, was “not mentioned in the legislative redistricting guidelines,” it may not factor into our analysis. See Ala. Legislative Black Caucus, 135 S.Ct. at 1271-72. In any event, regardless of whether the split was “smooth” or not, the State’s neglect of a race-neutral criterion (keeping precincts whole) had a sizeable effect on the racial composition of the district. Further, had the State not split any precincts, HD 57 would have been 66.59 % black, almost two percentage points below its goal. Def. Precinct Split Analysis (doc. no. 263-3) at 10. Although these precinct splits are not dispositive, they provide further circumstantial evidence of the drafters’ racial intentions.
The State’s only attempt to offer a race-neutral explanation for HD 57 pointed to a map of the district’s 2001 lines in McClen-don’s notebook, which includes hand-drawn lines sweeping two new areas into the district. See McClendon Notebook, Def. Ex. 459, at 125. According to McClendon’s testimony about his notebook generally, the hand-drawn additions could reflect input from incumbents.17 The State, noting that the resulting district had a “slightly higher” percentage of black population than “Hinaman’s so-called target,” Def. Remand Br. (doc. no. 263) at 122, apparently suggests that Hinaman would have met a racial target exactly but for the interjection of feedback from legislators. That the State’s technical adviser would *1388have hit the racial target precisely by default only further demonstrates — not detracts from — the central role that race played in the formulation of the district’s lines. At any rate, because the final district population fell a statistically insignificant 20 black people above its target, this justification is unconvincing. Accordingly, I conclude that race predominated in the drawing of HD 57.
f. HD 60 (North Birmingham)
HD 60 provides further proof that Hina-man’s transfer of HD 53 out of Jefferson County allowed the State to meet its racial targets in Jefferson County. The State drafted a district approximately one-quarter percentage point over its target, taking a district that had been 67.41 % black and leaving it 67.68 % black ( + 0.27 %). Especially when considered in light of the drafters’ need to add 8,817 people to meet its ideal population target (19.37 % of the total district population), this feat is impressive. District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 5.
This district is an outlier in the State’s otherwise inefficient resolution of its under-population dilemma: it efficiently moved only 9,170 people in or out of the district despite adding 8,380 people to address under-population. House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4. But that apparently aberrational approach can be explained because the drafters quite simply tacked on a southeastern tip to the district that had previously belonged to HD 53, resulting in an irregularly shaped district that snakes south to encompass more black population. See Map, Def. Ex. 480. For the reasons previously discussed, the displacement of former residents of HD 53 to meet the State’s incorrect view of § 5 retrogression provides substantial evidence that race predominated in the surrounding Jefferson County districts, including HD 60.
[[Image here]]
*1389Although the drafters split 11 precincts to construct HD 60, the majority analyzes only the two precincts split between the district and majority-white districts, ignoring the nine precincts split between majority-black districts. Because the majority’s analysis considers those two precinct splits separately, it concludes that the effect was negligible. See ante, at 1218-19. The State has indicated, however, that had it not used any split precincts, HD 60 would have been 72.5 % black — more than 5 % above its racial target. Def. Precinct Split Analysis (doc. no. 263-3) at 13. The split of Gardendale Civic Center precinct is particularly troubling: the drafters placed a 47.3 % black population in majority-black HD 60 but only a 6.1 % black population in majority-white HD 51, carving out a small sliver at the southern end of the precinct to include in HD 60. 2012 House District Precinct Splits, Def. Ex. 405, at 98, 116; Map, APSX 108. The use of split precincts, including those split with other majority-black districts, contributes to an inference that race predominated in HD 60.
[[Image here]]
Gardendale Civic Center Precinct, reflecting small sliver added to HD 60 in orange. Map, APSX 108.
The State suggests that HD 60 passes constitutional muster because it preserved the core of the old district and protected its incumbent. Def. Remand Br. (doc. no. 268) at 125. The first justification, core preservation, “is not directly relevant to the origin of the new district inhabitants,” Ala. Legislative Black Caucus, 135 S.Ct. at 1271, and thus cannot preclude a predominance finding. At any rate, protecting the district’s core cannot explain the addition of an irregularly shaped southeastern tip with heavily-black census blocks from the former HD 53, see Maps, ADC Supp. Ex. 16B & 16C, which resulted from the State’s decision to shift population among Jefferson County majority-black House districts to meet its equal-population goals consistently with its approach to retrogression.
As to the State’s asserted race-neutral justification of incumbency protection, the evidence in the record gives no indication *1390that anything about that goal was unique or predominated in the drafters’ approach to HD 60. The incumbent’s residence in the mid-central portion of the district, a portion of the district which remained constant from the 2001 map, cannot explain HD 60’s precise match to the racial target, the use of 11 precinct splits, or the addition of an irregularly shaped southeastern tip swallowed from the former HD 53. See House Addresses, Def. Supp. Ex. 1; 2014 House Districts Map, Def. Supp. Ex. 2.
Because the plaintiffs have advanced sufficient evidence suggesting racial predominance that cannot be explained on race-neutral grounds, I find that race predominated in the drawing of HD 60.
ii. West Black Belt (SD 24 and HDs 67 and 69)
Direct evidence specific to districts in the Western Black Belt establishes that the drafters equated under-population in majority-black districts with the need to add heavily-black population to the districts. Beyond direct evidence, the majority fails to place appropriate weight on circumstantial evidence that the State’s racial policy was brought to bear' in these districts, meeting racial targets with mathematical precision and unnecessarily splitting counties and precincts in search of new heavily-black population to add to the district. For these reasons, I conclude that race predominated in the drafters’ approach to SD 24, HD 67, and HD 69.
a. SD 24 (Choctaw, Clarke, Green, Hale, Marengo, Pickens, Suffolk, and Tuscaloosa Counties)
Overwhelming direct and circumstantial evidence reflects the State’s use of race to draw SD 24: the drafters said they were targeting a racial percentage for the district, and they in fact met that target. In light of this direct evidence, I disagree with the majority’s analysis that the racial-predominance inquiry in this district requires a “close call.” See ante, at 1111. In fact, the State’s direct evidence has made our job quite easy. See Shaw II, 517 U.S. at 906, 116 S.Ct. 1894 (placing substantial weight on “direct evidence of the legislature’s objective”).
Discussing SD 24’s under-population, Dial testified that it “had to have more minorities.” Tr. Vol. I (doc. no. 215) at 48:19-23. Just like Hinaman’s decision to move HD 53 from Jefferson County because of his conclusion that under-population in majority-black districts could only be remedied by adding more black population, Dial’s testimony serves as an admission that he also correlated under-population with the racial composition of the residents purportedly required to address the under-population.18
In addition, Senator Marc Keahey, the incumbent for neighboring SD 22, offered uncontroverted testimony that Dial’s mechanical view of retrogression limited the options for drawing SD 24’s boundaries.19 Dial would consider any amendments proposed by Keahey under two criteria: first, affected incumbent senators agreed to the amendment, and, second, the amendment would not retrogress the minority population, by which Keahey understood Dial to mean lowering by any amount the percentage of black population in the district from *1391the level of the 2000 census used to draft the 2001 lines.20 Tr. Vol. I (doc. no. 215) at 192:2-24. Keahey proposed, among other amendments, smoothing the southern boundary of SD 24 at the Marengo County border so that it did not extend into a portion of Clarke County — a recommendation consistent with the committee’s guidelines of keeping counties whole. Id. at 199:24-25-200:1-4. The incumbent senator for affected SD 24 agreed to the amendment, meeting the first of Dial’s two stated criteria for approval, but Dial rejected it because of race: he stated that it would retrogress majority-black districts. Id. at 198:2-5, 200:7-8. Keahey later learned that Dial’s understanding of retrogression measured a reduction in the black population using 2010 census data applied to the 2001 lines, consistent with overwhelming evidence of the statewide policy. Id. at 198:8-15. Because Dial rejected the incumbent’s proposal based on his now-discredited understanding of retrogression, Keah-ey’s testimony provides direct evidence that the statewide policy was brought to bear in SD 24.
The majority agrees that the testimony of Dial and Keahey shows that the drafters “considered race in the drafting of the district” but nonetheless concludes that the evidence falls short of establishing predominance. Ante, at 1113. But because the drafters “considered race” when solving under-population, significantly more black people were added to the district than white people (28,241 compared to 19,606), while significantly more white people were removed from the district compared to black people (14,099 compared to 10,828). Because the drafters’ admitted use of race motivated the movement of “a significant number of voters within or without” the district, the testimony strongly supports a predominance conclusion. Ala. Legislative Black Caucus, 135 S.Ct. at 1267 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475).
By the time the legislature had completed redistricting, the drafters had accomplished their goal: SD 24, which had been 62.78 % black when applying the 2010 census to the old lines, stood at 63.22 % black, just 0.44 % above the target. Comparison of Minority Percentages (doc. no. 263-2) at 4. At trial, Hinaman boasted that, although SD 24 had been severely under-populated, it was ultimately “very close to getting back to the identical numbers that [it was] 10 years ago,” by which he meant identical “[b]y total black percentage.” Tr. Vol. Ill (doc. no. 217) at 186:19-25-187:1.
*1392[[Image here]]
SD 24 and its eight counties, six of which were split. Map, Def. Supp. Ex. 5.
By adding a northern portion of Clarke County and a southwestern portion of Pickens County to SD 24, the drafters violated the traditional districting criterion of keeping counties whole. Although the State maintains that changes to the district lines “largely resulted from the wishes of incumbent senators in the area,” Def. Remand Br. (doc. no. 263) at 70, it offered no evidence disputing Keahey’s testimony about his rejected request pertaining to Clarke County.21 The State argued merely that it needed to add slivers of Clarke and Pickens Counties because of under-population and geographically mandated limitations, since the district faced Mississippi on its western border and under-populated SD 23 on its eastern border.
Merely addressing under-population does not explain the substantial racial disparities in those portions of the counties added to SD 24. In Pickens County, which had previously belonged wholly to majority-white SD 21, the State added 7,303 people, of whom 74.01 % were black, to SD 24. District Splits, Def. Ex. 401 (doc. no. 30-40) at 8. It kept the northern and eastern portions of Pickens County — which just so happened to contain a 75.00 % white population — in majority-white SD 21. Id. at 7. The majority dismisses this racial division *1393because the shape of the SD 24 portion of Pickens County is tied to the western border of the State and has a “relatively smooth line” to the Greene County border, ante, at 1114, but the State did not describe these race-neutral explanations as its “actual purpose,” see Shaw II, 517 U.S. at 908 n.4, 116 S.Ct. 1894. Instead, the evidence shows that the State specifically intended to add black population: Dial testified that SD 24 “had to have more minorities, and so [it] had to grow as well,” so it “grew up into the north district,” referring to Pickens County. Tr. Vol. I (doc. no. 215) at 48:19-25. To the extent that creating a ‘smooth line’ creates a race-neutral inference, it cannot overcome the plaintiffs’ direct evidence that traditional districting factors were “subordinated to racial objectives.” Miller, 515 U.S. at 919, 115 S.Ct. 2475.
In Clarke County, which had previously belonged solely to majority-white SD 22, the drafters added a 61.13 % black northern slice to SD 24 and a 70.67 % black southeastern section to majority-black SD 23 while preserving in majority-white SD 22 a portion of the county that was only 22.02 % black. District Splits, Def. Ex. 401 (doe. no. 30-40) at 7-9. The State’s addition of heavily black portions of split counties to majority-black SD 24 without a race-neutral explanation provides further substantial evidence that race predominated over traditional districting criteria.
Precinct splits tell a similar story. Although the 21 precinct splits in SD 24 do not in and of themselves demonstrate that race predominated, several are quite striking. In particular, in two Choctaw County precinct splits with majority-white SD 22, the State placed hundreds of the precinct’s black residents in SD 24 but not a single one in SD 22. See District Splits, Def. Ex. 475, at 88, 109; Precinct Map, APSX 291 (split of Silas-Souwilpa-Isney-Toomey precinct with 850 blacks and 818 whites placed in SD 24, but no blacks and 237 whites in SD 22); Precinct Map, APSX 290 (split of Riderwood-Rock Springs precinct with 335 blacks and 339 whites placed in SD 24, but no blacks and 43 whites in majority-white SD 22). The majority concludes without explanation that it could find “no racial pattern” in the Silas precinct split, ante, at 1122, but splitting a precinct by placing black people on only one side of the line seems an obvious racial pattern to me.
*1394[[Image here]]
Silas-Souwilpa-Isney-Toomey precinct, which added significant black population to SD 24 but no black people in the irregularly-shaped portion drawn in majority-white SD 22. Precinct Map, APSX 291.
Moreover, these stark racial divisions cannot be explained solely by under-population, as the State attempts to do. SD 24 was 17,732 people short of the ideal population target; in redistricting, the State moved 70,988 people in and out of the district — four times the number of people necessary to address under-population. Pop. Summary Report, Def. Ex. 402 (doc. no. 80-41) at 1; Senate District Pop. Changes from 2010 to 2012, ADC Supp. Ex. 5. Combined with the stark racial divisions in the new geographical areas added to address under-population, this race-neutral explanation cannot overcome a finding that race predominated in drawing SD 24.
b. HD 67 (Dallas and Perry Counties)
I agree with much of the majority’s conclusions as to HD 67, but not the ultimate one. I find that race,predominated in the State’s drawing of this district because of sufficient circumstantial evidence that the State’s racial target “motivate[d] the legislature’s decision to place a significant number of voters within or without” the district. Ala. Legislative Black Caucus, 135 S.Ct. at 1267 (quoting Miller, 515 U.S. at 916). The district was 69.14 % black after the 2010 census; by adding just the right amount of Perry County that would address both under-population and its racial target, the State achieved a new district that was 69.15 % black. Race is the “[o]ne factor and one factor alone [that] explains the fact that” the drafters hit HD 67’s target so precisely, and is sufficient to establish that race predominated. Ala. Legislative Black Caucus, 989 F.Supp.2d at 1325 (Thompson, J., dissenting).
*1395[[Image here]]
As the majority explains, HD 67 fit entirely within Dallas County under the old lines. Most of Dallas County belonged to HD 67, with approximately 13 % of the county’s population in neighboring, majority-black HD 69. 2001 District Splits, Def. Ex. 412, at 7. The drafters moved all Dallas County residents to HD 67, serving the race-neutral principle of keeping the county whole. Nonetheless, with the new ideal population guidelines, the district remained 1,701 people short of the ideal population target. District Statistics Report, Def. Ex. 403 (doc. no. 30-36) at 6; County District Splits, Def. Ex. 404, at 15. Needing to move beyond Dallas County’s borders, the drafters added 1,258 residents from Perry County (59.54 % of whom were black), a rural Black Belt county that had previously been encompassed entirely by majority-black HD 72. County District Splits, Def. Ex. 404, at 16. To do so, the drafters added six precincts from Perry County, four of which were split. 2012 House District Precinct Splits, Def. Ex. 405, at 135-36. The drafters? final HD 67 remained 443 people short of its equal-population target (but within the -1.00 % allowable deviation), but just three black people above an exact match to its racial target. District Statistics Report, Def. Ex. 403 (doc. no. 30-36) at 6. That uncanny match provides strong circumstantial evidence that race predominated in the drafters’ exact placement of the Perry County boundary of HD 67.
The majority criticizes my reliance here on circumstantial evidence that the State met its racial target, see ante, at 1223-24; however, the majority misreads the Supreme Court’s instruction that a statewide policy of racial targets “provides evidence” that race motivated the drawing of a district’s lines, Ala. Legislative Black Caucus, 135 S.Ct. at 1267, as a prohibition on finding racial predominance where there is direct evidence that the State set a racial target and circumstantial evidence that it met the target precisely in a particular district. As I have explained, where the State sets a high racial target (here, a *139669.14 % black population), applies it with strict constraints (there must not be too many black people above the target, nor too few black people below the target), and does so as part of a rigorous intent to reach as many majority-black districts as possible (this is one of the 16 districts for which the drafters hit their racial target within one percent), that qualifies as strong evidence of racial predominance. Nevertheless, I agree with the majority that race did not predominate in 12 districts in large part because this circumstantial evidence does not apply in those districts.
I agree with the majority that race-neutral criteria explain some of HD 67’s boundaries, but not so much to negate the conclusion that race predominated over traditional districting criteria in the district as a whole. The State had valid, race-neutral reasons to use Perry County as opposed to a portion of a contiguous majority-white county, as some of the plaintiffs’ alternative plans proposed, so that it could conjoin two Black Belt counties, a coherent community of interest, into a single district.22
Nonetheless, several line-drawing decisions are unexplainable on race-neutral grounds. Maps at the precinct level along the border with HD 72 demonstrate that the State could have made boundary choices that would have resulted in smoother lines. See, e.g., Precinct Map, APSX 245 (National Guard Armory); Precinct Map, APSX247 (UCH Airport Armory split). Similarly, the new territory from Perry County was drawn with several uneven tentacles that swept in majority-black populations. Map, ADC Suppl. Ex. 17C. The State could have drawn an equally compact district with more or less of Perry County, but it instead drew the line to match its racial target exactly. This satisfies the racial-predominance standard.
c. HD 69 (Autauga County, Lowndes County; Montgomery County, and Wilcox County)
Once again, I find that race predominated primarily based on a disagreement with the majority about evidence of racial targets. To the majority, the “negligible” increase of the percentage of black population — from 64.16 % to 64.21 % black — is not suspicious, apparently because the drafters did not place a noticeably higher percentage of black people within the district. Comparison of Minority Percentages (doc. no. 263-2) at 2. But the State unquestionably used race to place people “within” the district according to a target, and that supports the plaintiffs’ showing of racial predominance. Miller, 515 U.S. at 916, 115 S.Ct. 2475. As I previously observed, this evidence of hitting a racial target is particularly convincing not only because it happened in this district, but because it occurred in 21 of the challenged districts. This obvious pattern provides clear evidence that race predominated.
In HD 69, although the State had substantial under-population to remedy (7,949 people), it swept with a much broader brush than necessary, moving in and out 24,373 people — more than half of the district’s population. District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 6; House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4. Most notably, the State’s shuffling removed heavily white rural precincts and added majority-black urban districts. The State removed several Autauga County precincts, even though this county was already split, and added portions of heavily-black Montgomery County. The drafters removed Autauga *1397County’s Billingsley Government Center and White City Fire Station precincts from HD 69, a total of 2,161 people that were only 14.48 % black, and placed them in majority-white HD 42, which was already 6.19 % over-populated. ADC Ex. W, C-41 (doc. no. 140-2) at 84-85; 2001 House District Precinct Splits, Def. Ex. 413, at 43; District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 4. To fill HD 69’s population gap, the drafters added 15,190 people from urban Montgomery County, 60.45 % of whom were black, to this otherwise rural district. Direct evidence demonstrates that the State merely assumed that re-population of a majority-black district required the addition of new population that was also majority black. Hinaman testified that “it would have been hard to [repopulate HD 69 without going into Montgomery County] without eliminating another black district,” Tr. Vol. Ill (doc. no. 217) at 160:4-5, but then said he did not know what prevented him from using population from other counties to re-populate the district. Id. at 161:4-6.
Nor could the State demonstrate that the ADC plaintiffs’ alternative, which would have repopulated HD 69 with rural population from the less heavily majority-black Butler County, was unworkable. Tr. Vol. Ill (doc. no. 217) at 160:12-25-161:1-6 (Hinaman stating “[t]here may be no impediment to [using Butler County], but I don’t know that — I did not do — you know, I don’t know if those numbers work out or not.”). The State suggests that it needed to use Montgomery County’s heavily-black population to re-populate HD 69 or else it risked losing a majority-black district, but in contiguous HD 90, Butler County’s heavily-black population was drawn into majority-white HD 90. The State’s generalized assertion that it had limited options in its pursuit of population for HD 69 because of rural under-population throughout the Black Belt districts carries little weight when it cannot explain its failure to use specifically proposed alternatives.
The drafters’ extension of HD 69 into Montgomery County also relied on a land bridge accomplished with precinct splits that even the majority acknowledges is “suspicious.” Ante at 1251. The Snowdoun Womens Club precinct splits HD 69 and majority-white HD 75. See Precinct Map, APSX 239. If one were to look at precinct-specific demographics alone, the split seems race-neutral: it resulted in 51 whites and three blacks in majority-black HD 69, no whites and three blacks in majority-black HD 76, and 202 blacks and 493 whites in majority-white HD 75. 2012 House District Precinct Splits, Def. Ex. 405, at 145, 158, 160. More consequentially, the split precinct map shows that the split — along with a further split in the contiguous Fitzpatrick Elementary School precinct — allowed the drafters to append a highly-irregular land bridge to the easternmost end of HD 69, which facilitated the inclusion of significant black census blocks from the urban Peter Crump School precinct. Precinct Map, APSX 239. By doing so, the drafters were able to add an additional 1,701 people to HD 69, nearly all (95.47 %) of whom were black. ADC Ex. W, C-41 (doc. no. 140-2) at 146. This serves as “quite compelling” evidence that the borders of HD 69 are “unexplainable other than by race.” Miller, 515 U.S. at 917, 115 S.Ct. 2475 (holding that “drawing of narrow land bridges to incorporate within the district outlying appendages ... considered in conjunction with its racial and population densities” provides evidence of racial gerrymandering).23
*1398[[Image here]]
Precinct splits used to build land bridge to scoop majority-black census blocks from the Peter Crump School precinct into HD 69. Precinct Map, APSX 239.
Because the drafters met a racial target while disregarding race-neutral criteria, I find that race predominated in drawing HD 69.
iii. HD 83 (East Black Belt: Lee and Russell Counties)
Race predominated in the drawing of HD 83 because the State shifted a significant number of voters in and out of the district, used split precincts to incorporate the vast majority of the district’s residents and exclude portions of precincts with high percentages of white population, and met its racial target. With a 56.92 % black target, the State drew a district 57.52 % black ( + 0.60 %), just 277 black people short of an exact match. To accomplish that, the State moved 18,466 people in and out of the district, even though the district was only under-populated by 4,482 people. House Districts Pop. Changes from 2010 to 2012, ADC Supp. Ex. 4.
The majority acknowledges that two precinct splits appear “suspicious,” but disregards them because they “do not account for the assignment of a significant number of people.” Ante, at 1310. The majority assesses the affected population as “less than half of a percent of the total population of the district,” but it counted only black people within one of the splits, the *1399Ladonia Fire Department precinct, see ante, at 1312. Because the precinct splits also facilitated the placement of a significant white population “-without” HD 83, that population must be taken into account as well. Miller, 515 U.S. at 916, 115 S.Ct. 2475. Specifically, the Ladonia Fire Department precinct placed an 85.9 % black population into HD 83 but only a 12.5 % black population into majority-white HD 80, while the Old Salem School precinct added a 42.3 % black population to HD 83 but only an 8.9 % black population to majority-white HD 38. 2012 House District Precinct Splits, Def. Ex. 405, at 78, 166, 172. The high proportion of black population added to HD 83 appears particularly suspicious because both precincts had previously been wholly encompassed by majority-white districts. 2001 House District Precinct Splits, Def. Ex. 413, at 24, 49. Combined, the two precinct splits' allowed the State to place an additional 204 black people into majority-black HD 83 while retaining 7,541 white people in majority-white districts, equivalent to 16 % of HD 83’s total population. 2012 House District Precinct Splits, Def. Ex. 405, at 172-73. That population rises to the level of a “significant number of voters” placed without the district on account of race. See Miller, 515 U.S. at 916, 115 S.Ct. 2475. Indeed, another three-judge panel concluded that 0.58 % of a district qualified as a “significant number of voters” sufficient to establish racial predominance. Covington, 316 F.R.D. at 149 n.28 (three-judge court).
Similarly, the split of Opelika B precinct allowed the State to place a large, overwhelmingly black population in HD 83 and a large, overwhelmingly white population in majority-white HDs 38 and 79. The drafters placed 18,201 people into HD 83, 58.81 % of whom were black, and 9,598 people into majority-white districts, only 10.58 % of whom were black. 2012 House District Precinct Splits, Def. Ex. 405, at 78, 165, 172. The placement of majority-black census blocks into HD 83 accounts for the odd shape of the northwestern limb of the district. Precinct Maps, APSX 140 & ADC Supp. Ex. 27F.
*1400[[Image here]]
Opelika B precinct split between majority-black HDs 82 and 83 and majority-white HDs 38 and 79. Census blocks with a majority-black population are shaded in green. Precinct Map, APSX 140 (excerpt).
The State suggests that an inconsistent racial pattern in the Opelika B precinct split negates a showing that a large proportion of the precinct’s population was allocated according to race, but the supposed inconsistencies are in fact consistent with the drafters’ statewide approach to racial targets. In contrast to the heavily black population added to HD 83, the drafters placed a largely white portion of the precinct into majority-black HD 82 (2,140 people total, only 19.44 % of whom were black). 2012 House District Precinct Splits, Def. Ex. 405, at 171. But extending HD 82 to the majority-black census blocks of that part of the precinct — not contiguous to the majority-black census blocks of the precinct included in HD 83 — would have placed HD 82 well over the +/- 1 % ideal population target. Id.; Precinct Map, APSX 140. Unlike many of the other majority-black districts, the drafters had already exceeded the ideal population of HD 82 with a deviation 0.74 % above ideal. District Statistics Report, Def. Ex. 403 (doc. no. 30-36) at 7. A racial-predominance finding does not require a showing that the drafters utilized racial considerations to the maximum extent possible; it only requires a showing that race predominated over other, race-neutral criteria.24
*1401In addition to the three already mentioned precincts, the drafters split eight other precincts in HD 83. In all, a substantial 75.7 % of the district’s population resides in split precincts. 2012 House District Precinct Splits, Def. Ex. 405, at 172-73. The district also shares its two counties with numerous other districts, ignoring the guideline to keep counties whole; it shares parts of both Lee and Russell Counties with three other districts.
The State contends that HD 83 was drawn in a race-neutral manner because it “preserves the core of the former district” and “is compact.” Def. Remand Br. (doc. no. 263) at 153. While it is difficult to envision this odd-shaped district, essentially comprising five different ‘legs’ that jut out at various irregular points, as compact, the State accurately notes that the shape, driven by the need to join the population centers of Opelika and Phenix City, does not vary significantly from its 2001 lines. With that said, the plaintiffs have estab-hshed that where the drafters needed to add population, they did so in a race-conscious manner, selectively extending the boundaries of the district to include majority-black census blocks. For example, ⅛ a north-central limb, the drafters extended the length of the limb further north into Lee County by drawing boundaries closely tailored to several majority-black census blocks. See Precinct Maps, ADC Supp. Ex. 27C & 27D.
[[Image here]]
HD 83. Map, Def. Ex. 479 (excerpt).
*1402[[Image here]]
New north-central limb of HD 83 reflecting addition of heavily-black census blocks. Map, ADC Supp. Ex. 27C.
For all of these reasons, I must conclude that race predominated in the design of HD 83.
iv. HD 97 (Mobile County)
The drafters hit their racial target on the head in Mobile County’s HD 97, retaining the previous 60.66 % black population exactly. Comparison of Minority Percentages (doc. no. 263-2) at 2. Tellingly, the State accomplished this feat despite adding a net total of 9,665 people, addressing a district that had been 22.22 % under-populated. District Statistics Report, Def. Ex. 403 (doc. no. 30-36) at 8; District Statistics Report, Def. Ex. 406 (doc. no. 30-37) at 8. The State offers no explanation for achieving this exact racial match using race-neutral methods, and the majority apparently does not take this precision into account.25 The exact racial match provides strong circumstantial evidence that the drafters pursued the statewide policy in HD 97.
The plaintiffs have demonstrated, and the State has not rebutted, that race, and not an equal-population objective, drove the drafters to meet HD 97’s racial target. For example, the redrawn HD 97 remained under-populated by 0.99 %, just barely meeting the State’s allowable minus-one-percent total population deviation. District Statistics Report, Def. Ex. 403 (doc. no. 30-36) at 8. Neighboring HD 96, a majority-white district, was over-populated by 0.99 %, just barely meeting the allowable plus-one-percent total population deviation. Id. Nonetheless, the northernmost tip of HD 97 — referred to by the majority as the “bishop’s mitre” — was drawn in a highly irregular shape through a racially divided split of the Saraland Civic Center Precinct, avoiding majority-white census blocks that were instead drawn into HD 96. Had the State prioritized compactness, it could have added additional majority-white census blocks from *1403HD 96 into HD 97 to smooth out the district’s shape and bring both districts closer to the ideal population target. See Precinct Map, APSX 197. Of course, had the State added additional majority-white census blocks to HD 97, it would not have hit the racial target. The State offered no explanation for its apparent prioritization of a racial target above regularly-shaped district lines and its ideal population target. Instead, it noted only that the Sara-land precinct split added more whites than blacks to the district. Def. Remand Br. (doc. no. 263) at 160-61. But where the resulting district has achieved an exact racial match, a suspicious precinct split provides evidence of racial predominance even if it allocated more whites than blacks to a majority-black district.
[[Image here]]
Saraland Civic Center precinct split. Map, APSX 197.
The same principle holds true for the district as a whole. The State’s unsplitting exercise demonstrates that, had it not included a single precinct split in HD 97, its majority-black population would have increased to 64.8 % of the total population. Def. Precinct Split Analysis (doc. no. 263-3) at 28. Given the testimony of the principal architect for the House districts that he had not “been advised that if you go over a certain [percentage of minority population]” it would be a problem, see McClendon Dep. (doc. no. 125-4) 106:10-18, the use of precinct splits to reduce the percentage to a result identical to the racial target strengthens the conclusion that race predominated over traditional dis-tricting principles, including keeping precincts whole.
Because the plaintiffs demonstrated that the drafters applied a racial target to the drawing of HD 97 and disregarded race-neutral criteria to do so, I conclude that race predominated.
C. Strict Scrutiny
As I have concluded that race predominated over traditional districting criteria when the State drafted ten additional districts, I must also determine whether the districts would survive strict scrutiny. Because my racial-predominance analysis determined that the drafters applied a statewide policy of reaching racial targets for each of the ten identified districts, I also conclude, for the reasons explained above, *1404that the State did not narrowly tailor its use of race in these districts. See supra at 1365-68. Moreover, the State has not submitted evidence or arguments specific to the identified districts sufficient to carry its burden of showing that its approach was narrowly tailored. Accordingly, I conclude that, although the State had a compelling interest in complying with § 5 of the Voting Rights Act, its approach to SD 24 and HDs 52, 55, 56, 57, 60, 67, 69, 83, and 97 was not narrowly tailored to meet that interest and therefore could not survive strict scrutiny.
III. Twelve Remaining Districts
In addition to the 12 districts deemed unconstitutional racial gerrymanders by the majority and the additional 12 I deem unconstitutional, the plaintiffs have sought to establish that race predominated in the State’s drawing of an additional 12 majority-black districts. As to these remaining 12 districts, I agree with the majority that the plaintiffs have not satisfied their burden to demonstrate district-specific proof of racial predominance in the remaining districts: SDs 18, 19, and 33 and HDs 19, 58, 59, 72, 76, 78, 84, 98, and 103.
IV. Conclusion
For the above reasons, I concur with the majority that 12 of the districts (Senate Districts 20, 26, and 28 and House Districts 32, 53, 54, 70, 71, 77, 82, 85, and 99) do not pass constitutional muster and that 12 districts (Senate Districts 18, 19, and 33 and House Districts 19, 58, 59, 72, 76, 78, 84, 98, and 103) do pass constitutional muster.
I respectfully dissent from the majority’s conclusion that an additional 12 districts (Senate Districts 23 and 24 and House Districts 52, 55, 56, 57, 60, 67, 68, 69, 83, and 97) are constitutional.
DONE, this the 20th day of January, 2017.
. Justice O’Connor and the other members of the plurality ultimately concluded that the new majority-black district was not narrowly tailored because it was not compact and was oddly shaped. Bush, 517 U.S. at 979, 995, 116 S.Ct. 1941.
. McClendon worked on House districts while Senator Gerald Dial worked on Senate districts. Tr. Vol. Ill (doc. no. 217) at 256:18-21; Affidavit of Gerald Dial, APX 63 (doc. no. 76-4) ¶ 8.
. McClendon was asked if "any of the legislators or any of the folks speaking at the public hearings express [sic] any concerns that the plan that was being proposed was going to result in pulling blacks from certain districts and concentrating them in black-majority districts?” McClendon testified that "if that came up, it was not a constant theme by any means" and it "wasn’t a high priority issue that I recall these folks talking about that.” McClendon Dep. (doc. no. 125-4) at 110:4-15. He could not “remember a specific conversation on that.” Id. at 111:9-14.
. The majority suggests that Dial "considered” Sanders’s opinion, see ante, at 1109, but Dial merely testified that he remembered hearing it. Tr. Vol. I (doc. no. 215) at 37:9-25-38:1 (Dial recalling that Sanders "felt like that the minority districts should be at a minimum 62 percent minorities”). Even if Dial actually ‘considered’ Sanders’s statement, consideration remains at least one step removed from reliance.
. When Dial was asked to identify the instructions he gave Hinaman, the suggestions of Jackson and Sanders were not among them. Tr. Vol. I (doc. no. 215) at 67:20-68:1, 69:16-25. Dial testified that he instructed Hinaman about "fill[ing] in the number of minorities to insure that we did not regress the minority districts,” but he also testified that his approach to doing so consisted of maintaining the black percentage of the population in each district. Id. at 67:23-25-68:1, 136:8-14.
. SD 18 (59.10 % black) and SD 28 (59.83 % black). District Statistics Report, Def. Ex. 400 (doc. no. 30-39) at 2, 3.
. As the majority acknowledges, Sanders referred to all majority-black districts in the State, and, consistently, Dial understood Sanders’s 62 % statement as applicable to "all districts,” in contrast to other instructions Sanders provided that were specific to SD 23. Tr. Vol. I (doc. no. 215) at 37:13-25-38:1. As for Jackson’s comments, the State argued he addressed "that part of the State” (without defining "that part”), Def. Remand Br. (doc. no. 263) at 54, and the majority understood him to refer to "his area.” See ante, at 1062. For my part, I have been unable to discern Jackson’s intentions, given his various vague and nonspecific references to "these House and Senate districts,” "those districts,” and "a minority district.” Reapportionment Comm. Oct. 13, 2011 Hrg. Tr. (doc. no. 30-23) at 8.
. Because Jackson focused on reducing the percentage of black voters, I assume that he was not motivated by incumbency protection. To this end, I take judicial notice that Jackson — who expressed concern about placing too many blacks within a district — has never faced a general election opponent, but Sanders — who expressed concern about establishing a minimum floor of black would-be supporters — usually has faced general election competition. See Alabama Votes, Elections Division Data, https://www.alabamavotes.gov/ ElectionsData.aspx.
. For the same reason, trial testimony from Alabama Democratic Conference member Dr. Joe Reed and prior testimony from plaintiffs' expert Dr. Theodore Arrington relating to recommendations for total black percentage in majority-black districts may not play any role *1364in the strict-scrutiny analysis. Compare ante, at 1107 (stating that Reed and Arrington testimony "confirm the reliability" of statements made by Jackson and Sanders). Because there is no evidence that the State relied on this testimony to craft district lines for SD 23 and HD 68 — and, of course, it could not possibly have relied on after-the-fact trial testimony— it cannot establish a strong basis in evidence. See Bush, 517 U.S. at 971 n. *, 116 S.Ct. 1941 (finding that legislature’s purported reliance on 1992 election results cannot justify challenged redistricting where those results "were not before the legislature when it drew the district lines.”).
.Two majority-black Senate districts fell below Sanders’s 62 % minimum: SD 18 (59.10 % black) and SD 28 (59.83 % black). Three fell above Jackson’s 65 % suggestion: SD 19 (65.31 % black), SD 26 (75.13 % black), and SD 33 (71.64 % black). See District Statistics Report, Def. Ex. 400 (doc. no. 30-39).
. Seven majority-black House districts fell below Sanders’s 62 % minimum: HD 19 (61.25 % black), HD 32 (60.05 % black), HD 52 (60.13 % black), HD 53 (55.83 % black), HD 54 (56.83 % black), HD 97 (60.66 % black), and HD 98 (60.02 % black). Eleven majority-black House districts fell above Jackson’s 65 % suggestion: HD 55 (73.55 % black), HD 57 (68.47 % black), HD 58 (72.76 % black), HD 59 (76.72 % black), HD 60 (67.68 % black), HD 67 (69.15 % black), HD 71 (66.90 % black), HD 76 (73.79 % black), HD 77 (67.04 % black), HD 78 (69.99 % black), and HD 103 (65.06 % black). See District Statistics Report, Def. Ex. 403 (doc. no. 30-36).
. The majority's reliance on testimony from Dr. Reed suffers from the same problems as the comments of Jackson and Sanders; that is, Dr. Reed offered only generalizations about the appropriate percentage of black population in all majority-black districts, without offering district-specific evidence.
. It is an open question whether direct evidence of a statewide redistricting approach that is incompatible with a narrowly tailored approach to the use of race could alone be sufficient to conclude that the State has failed to meet strict scrutiny in a particular district. The Supreme Court remanded for this panel to conduct a district-specific analysis with respect to racial predominance, but made no such statement as to strict scrutiny. Ala. Legislative Black Caucus, 135 S.Ct. at 1266. In fact, in its strict-scrutiny analysis, the Supreme Court concluded that both "the District Court and the legislature asked the wrong question with respect to narrow tailoring,” which "may well have led to the wrong answer.” Id. at 1274 (emphasis added). As strict scrutiny requires the court to interrogate the legislature’s actions leading up to its race-based decision, one could conclude that this wrong approach may well have been fatal. However, because the State has failed to establish it had a strong basis in evidence for the particular districts at issue, I need not reach that conclusion here.
. Although each panel recognized that racial targets deserved significant evidentiary weight, factors specific to each State's application of the targets impacted the ultimate predominance conclusions. For example, in Bethune-Hill, the 55 % BVAP target was relatively low and the panel found only that the target was “used in structuring the districts,” not that the legislature mechanically applied it. 141 F.Supp.3d at 519.
. HD 52 ( + .02 %); HD 54 ( + .10 %); HD 55 (identical match); HD 56 ( + .01 %); HD 57 ( + .05 %); HD 60 (+.27 %). Comparison of Minority Percentages (doc. no. 263-2) at 2.
. The majority previously credited Licht-man's testimony on this subject. See Ala. Legislative Black Caucus, 989 F.Supp.2d at 1271.
. McClendon Dep. (doc. no. 125-4) at 78:2-16 (testifying that his notebook contained drawings made by incumbent legislators re-fleeting requested changes to districts). McClendon did not testify specifically about the drawing for HD 57.
. No such automatic correlation was necessary. Had the drafters addressed SD 24’s under-population without adding a single black person to the district, it would have remained a majority-black district at 54.63 % black. ADC Remand Br., Table 1 (doc. no. 258) at 21.
. Keahey’s testimony was previously credited by the majority. See Ala. Legislative Black Caucus, 989 F.Supp.2d at 1260.
. As other testimony established, the drafters’ concept of retrogression looked at 2010 census data applied to the 2001 lines. See, e.g„ Tr. Vol. Ill (doc. no. 217) at 222:8-12.
. Although the State maintained that the drafters’ approach to Clarke County was constrained by an incumbent residing on the boundary of SD 22 and SD 23, the incumbent’s residence was swallowed up by SD 23 in a southern portion of the State far removed from the portion of the county added to SD 24. 2014 Senate Map, Def. Supp. Ex 5; Map, Def. Ex. 476. Thus, the incumbent’s location had no effect on SD 24’s boundaries.
. The analysis might be different if the plaintiffs had a § 2 claim based on packing too many black residents into HD 67, but such a claim is not currently before the court.
. Although the majority suggests the large hook extending into Montgomery County, which encompassed this land bridge, was required to avoid the HD 78 incumbent’s Mont*1398gomery residence, it cannot explain the drafters’ need to add a small but concentrated pocket of black people from Montgomery County to HD 69. Only race can do that. The additional 1,701 mostly black people gained from the Crump precinct is very close to the 2,161 mostly white people the State extracted from the Autauga County precincts. In other words, if the State did not unexplainably remove predominantly-white population from HD 69, it would not have needed to use a suspicious land bridge into Montgomery County at all.
. Further, because the predominance inquiry must focus on the drafters’ intentions when drawing district lines, I place little weight, unlike the majority, on the somewhat similar district boundary reflected in the ADC’s plan. ADC Alternative Plan Map (doc. no. 287-18) at 9. The record provides no indication that the alternative plan, which reflects some variation from the State plan, had any effect on the racial split in the heavily populated Opeli-ka B Voting Precinct.
. The State argued that “all [majority-black Mobile County House districts] have a lesser percentage of black population than Hina-man's alleged target.” Def. Remand Br. (doc. no. 263) at 159. This is, of course, incorrect as to HD 97.