Statistics and the Voting Rights Act

This is the sixth in a series about the oral argument in Shelby County v. Holder.

Nina Totenberg at NPR had a rather provocatively-titled piece after the oral argument in Shelby County: "In Voting Rights Arguments, Chief Justice Misconstrued Census Data." The piece goes on to explain that Mr. Roberts "was wrong" when he made the claim that Massachusetts had the worst ratio of African-American to white voter turnout, and that Mississippi had the best. The Secretary of State of Massachusetts, William Galvin, agreed that it was "disturbing" "misinformation" and "wrong."

If I may, both Ms. Totenberg and Mr. Galvin are either intentionally misrepresenting Chief Justice Roberts's (and the lower court's dissenting opinion's) data, or they are unaware of an important distinction they've elided over.

For Chief Justice Roberts (I think), the concern is the coverage formula. And the coverage formula was reauthorized in 2006. And the last available voter data was 2004. It's unsurprising, then, that the lower court's dissenting opinion, at 11-14, look at the voting data from 2004. It specifically refers to this Census data, Table 4a.

Within that table, one can see that the turnout rate (and we'll set aside the "ratio" for now, which is the point Mr. Roberts raised, but not the one that Ms. Totenberg addressed; for that, I'll return briefly below) for African-Americans in Mississippi in 2004 was 66.8%, MoE 5.2. In Massachusetts, it was 43.5%, MoE 9.6. So assuming one wants to stretch the MoE, the low end of MS would have been 61.6%, and the high end of MA would have been 53.1%. Ms. Totenberg's calculation to "factor in the margins of error at their extremes" would result in the same confidence that MA African-American turnout was worse than MS.

As to the citizen voting-age population question, one can run a quick check in the MA data to see that it would rise from 43.5% to 46.5%, while MS would remain largely the same--and I'm fairly confident that even a change in the MoE would not put MA in a statistical range in which it would be better than MS.

Now, this is important data because it is 2004 data, the data that Congress would have used (and, taking into account time and space, absent a DeLorean, could have used) when it reauthorized the coverage formula.

Ms. Totenberg and Mr. Galvin use the 2010 Census data, which is not the data that Congress would have had at its disposal in reauthorization.

Mr. Galvin "assumes" it is the 2010 data Mr. Roberts discusses, and is not terribly careful if he says the "only thing we could find" was the 2010 Census, or that "academics" at other institutions "could find no record," when the record is in the lower court dissent itself.

Ms. Totenberg, to her credit, links to the lower court dissent--but then ignores the actual 2004 Census data cited, instead choosing to cite the 2010 Census data, which was not used in the lower court dissent (and which, I assume, was not cited by Mr. Roberts).

On top of this, Rob Anderson (Pepperdine) has identified still more errors in Ms. Totenberg's analysis at a basic statistical level.

Now, granted, I understand that one could argue that the question is too narrow, that citing solely the returns from a single election (i.e., 2004) is not enough to sink the coverage formula, that the effectiveness and turnout rates today are important in the Court's analysis, etc. And I think Politifact has a very fair take on the relevance of the data Mr. Roberts cited (although I would take issue of putting too much weight in Mr. Roberts's use of the "present tense" to decide which dates he used, as opposed to putting more weight in the record developed below).

Indeed, I'm not terribly sure how relevant the fact Mr. Roberts raises is. As I've mentioned, I think Mr. Roberts is as guilty as many on the Court of using oral argument to raise irrelevant points.

But, these stories glibly rejecting a point Mr. Roberts made at oral argument by using a point he didn't make do not advance the conversation in any meaningful way. If writers and politicians are going to parse oral argument so seriously, I think it's only fair to weigh seriously what was actually said rather than what they wish was said, and then address the relevance.

A version of this post originated as an email on the Election Law listserv.