Cite as: 581 U. S. ____ (2017) 27 Opinion of the Court discovered that doing so made essentially no difference: For example, only 18% of the region’s white Democrats wound up in District 12, whereas 65% of the black Democrats did. See id., at 332. The upshot was that, regardless of party, a black voter was three to four times more likely than a white voter to cast his ballot within District 12’s borders. See ibid. Those stark disparities led Ansolabehere to conclude that “race, and not party,” was “the dominant factor” in District 12’s design. Id., at 337.14 His report, as the District Court held, thus tended to confirm the plaintiffs’ direct evidence of racial predominance. See 159 F. Supp. 3d, at 620–621. The District Court’s assessment that all this evidence proved racial predominance clears the bar of clear error review. The court emphasized that the districting plan’s own architects had repeatedly described the influx of African-Americans into District 12 as a §5 compliance measure, not a side-effect of political gerrymandering. —————— 14 Hofeller did not dispute Ansolabehere’s figures, but questioned his inference. Those striking patterns, the mapmaker claimed, were nothing more than the result of his own reliance on voting data from the 2008 Presidential election—because that information (i.e., who voted for Obama and who for McCain) tracked race better than it did party registration. See App. 1101, 1111–1114; cf. Cromartie II, 532 U. S. 234, 245 (2001) (recognizing that “party registration and party preference do not always correspond”). As we have just recounted, however, the District Court had other reasons to disbelieve Hofeller’s testimony that he used solely that electoral data to draw District 12’s lines. See supra, at 25–26. And Ansolabehere contended that even if Hofeller did so, that choice of data could itself suggest an intent to sort voters by race. Voting results from a “single [Presidential] election with a Black candidate,” Ansolabehere explained, would be a ��problematic and unusual” indicator of future party preference, because of the racial dynamics peculiar to such a match-up. App. 341; see id., at 342– 343. That data would, indeed, be much more useful as a reflection of an area’s racial composition: “The Obama vote,” Ansolabehere found, is “an extremely strong positive indicator of the location of Black registered voters” and, conversely, an “extremely strong negative indicator of the location of White registered voters.” Id., at 342; see id., at 2546–2550.