28 BRNOVICH v. DEMOCRATIC NATIONAL COMMITTEE Opinion of the Court voters who voted on election day cast an out-of-precinct ballot. Ibid. For non-minority voters, the rate was around 0.5%. Ibid. (citing Tr. Exh. 97, at 3, 20–21). A policy that appears to work for 98% or more of voters to whom it applies—minority and non-minority alike—is unlikely to render a system unequally open. The Court of Appeals attempted to paint a different picture, but its use of statistics was highly misleading for reasons that were well explained by Judge Easterbrook in a §2 case involving voter IDs. As he put it, a distorted picture can be created by dividing one percentage by another. Frank, 768 F. 3d, at 752, n. 3. He gave this example: “If 99.9% of whites had photo IDs, and 99.7% of blacks did,” it could be said that “ ‘blacks are three times as likely as whites to lack qualifying ID’ (0.3 ÷ 0.1 = 3), but such a statement would mask the fact that the populations were effectively identical.” Ibid. That is exactly what the en banc Ninth Circuit did here. The District Court found that among the counties that reported out-of-precinct ballots in the 2016 general election, roughly 99% of Hispanic voters, 99% of African-American voters, and 99% of Native American voters who voted on election day cast their ballots in the right precinct, while roughly 99.5% of non-minority voters did so. 329 F. Supp. 3d, at 872. Based on these statistics, the en banc Ninth Circuit concluded that “minority voters in Arizona cast [outof-precinct] ballots at twice the rate of white voters.” 948 F. 3d, at 1014; see id., at 1004–1005. This is precisely the sort of statistical manipulation that Judge Easterbrook rightly criticized, namely, 1.0 ÷ 0.5 = 2. Properly understood, the statistics show only a small disparity that provides little support for concluding that Arizona’s political processes are not equally open. The Court of Appeals’ decision also failed to give appropriate weight to the state interests that the out-of-precinct rule serves. Not counting out-of-precinct votes induces