The oral argument today in Evenwel v. Abbott was revealing in many respects, as such arguments often are, in highlighting which points in the briefs command the most attention from the Justices. However, as is also often the case, the oral argument can serve to obscure certain points, as advocates attempt to use their time to complicate issues that really are quite simple. Such is the case, I believe, with the debate in Evenwel over whether redistricting on some basis other than total population is possible. On this question, I should note I am hardly a disinterested observer, having filed an amicus brief based on my experience as a Special Master and court-appointed redistricting expert, and having written about the issue in Politico, Scotusblog, and the Washington Post.
To be clear, the issue is not whether one is physically able to draw districts on the basis on some data source other than the census population numbers. Of course, one could. The question is whether the available data are consistent with the theory of one person, one vote – even according to the Appellants’ novel interpretation of that rule. There is a fundamental incongruity between the interpretation of the Equal Protection Clause proposed by the Appellants and the data they use to demonstrate a constitutional violation and suggest as a remedy.
First, no dataset of eligible voters exists. All of the debate about the American Community Survey’s (ACS) citizen voting age population (CVAP) data is somewhat beside the point. Even that dataset does not pretend to capture the relevant statistic — eligible voters — that the Appellants contend is necessary for compliance with one person, one vote. The ACS data do not account for 168,000 people in Texas prisons or the roughly 350,000 others who are disfranchised because of a felony conviction. Nor do they account for the roughly half million eligible voters who live abroad but can vote in Texas elections. These large categories of persons do not amount to rounding errors or statistical nit-picks – even if the ACS were accurate, it does not satisfy Appellants own constitutional standard for one person, one vote compliance.
Second, the problem with the ACS CVAP data is not that you are unable to put it into a computer and use it for redistricting. You could do that with any dataset. Rather, what you would be putting into the computer at the time of redistricting are old estimates of what the population was, not what it is. Everyone, including Appellants, agrees that the yearly ACS surveys are not usable for redistricting because they only sample 2.5% of households and are released only for communities in excess of 65,000 people. Therefore, they say, you should use the averages of the ACS surveys from the previous five years – that is, use five-year old data to redistrict for a ten-year redistricting cycle! At the time of the 2011 redistricting, for example, you would have had to use data averaged from the 2006 to 2010 surveys. Some of the 17 year olds not counted in the 2006 CVAP numbers, for example, would be 23 years old by the time of the first elections held under the new lines.
Third, the granularity of the data and the presence of error margins introduce novel questions regarding one person, one vote. Again, the point here is not whether one is able to redistrict on the basis of survey results: You could draw lines to equalize between districts the number of Donald Trump supporters according to presidential preference surveys. Rather, the idea of a survey – as compared to a census – is in tension with the notion of one-person, one-vote. The linedrawer who draws a district using the ACS survey data is, in effect, saying that he or she is 90 percent certain (the ACS’s reported confidence intervals) that the population in that district is within the margin of error as reported by the survey. As explained in my amicus brief, these error margins are sometimes quite large, especially at the local level. Such errors will increase further in size, if one needs to break up the block groups into blocks, as one often would do, in order to accommodate any number of other redistricting considerations. (Even the five-year averages of ACS data are not released below the block group level.) When it comes to ACS citizenship data, moreover, some of the data are imputed by the Census Bureau, itself: For five percent of respondents in the 2013 ACS, the Bureau “allocated” (i.e., filled in) the answer to the citizenship question.
Finally, and perhaps most obviously, the ACS can be eliminated at any time. This is not a theoretical possibility: the House of Representatives voted to eliminate it in 2012. Further, because of the government shutdown in 2013 and a reduction in funding in 2004, ACS surveys in those years suffered from a high margin of error.
I hope the ACS stays around, as it is an invaluable tool to inform public policy, including the redistricting process. But the possibility that it might disappear highlights the irony in the EvenwelAppellants’ argument. They maintain that the only constitutionally required population dataset – namely, the census enumeration of persons – is actually forbidden by the Constitution for use in redistricting. That just cannot be right.