This week, two major court cases concerning the right to an equal and effective vote revealed how crucial scientific integrity in the courts is going to be if voting rights are to be secured for all Americans. On Tuesday, a federal court threw out North Carolina’s Congressional districting plan as an unconstitutional partisan gerrymander, relying on extensive empirical models and statistical evidence that demonstrated both discriminatory intent and effect. On Wednesday morning, the Supreme Court of the United States heard oral arguments regarding Ohio’s “use-it-or-lose-it” voter list purging process, during which considerable time was dedicated to issues of data integrity and availability. Both cases illustrate the growing importance of our ability to measure equal justice under law, and the degree to which claims of voting rights violations are based on quantitative arguments.
Measuring intent and effect in gerrymandering
The North Carolina decision handed down Tuesday included an extended discussion of the courts’ “obligation to keep pace with the technological and methodological advances so it can effectively fulfill its constitutional role to police ever-more sophisticated modes of discrimination.” In their 205-page opinion, the court reprimanded defendants for arguing that claims should be dismissed simply because they “rely on new, sophisticated empirical methods that derive from academic research.”
The opinion explicitly relied on such methods for establishing the intent of the North Carolina legislative leaders to discriminate against voters of the opposing party. They did this through a combination of computer simulations that showed the near impossibility of the adopted plan being chosen without intending to discriminate, given the traits of that map compared to alternatives, and data visualizations that experts say illustrate the “signature” gerrymandering effect of partisan vote shares being non-linearly distributed across districts.
In finding a discriminatory impact, the court relied on both simulations and statistical tests of partisan asymmetry to demonstrate the near certainty that the governing party engineered itself at least one additional seat, and likely several additional seats, through the adopted plan.
Does a non-response “count” as a signal of any relevance?
In the Ohio voter list purge case, Justices repeatedly asked about available data from both sides that would provide an empirical context for the legal arguments. In particular, Justice Sotomayor asked about estimates of how many of those purged from voter lists had actually changed residence out of their districts, as the state assumes. Similarly, Justice Breyer stated that they were looking at “an empirical question” and inquired about the availability of “numbers, or surveys” of residential movement within the state, as well as estimates of what percentage of residents typically throw away mailed notifications, so as to get some grasp of what it means when a voter does not respond to a notification.
Indeed, a crucial challenge to the defense revolved around just what information was obtained from the notifications. Justice Alito pushed Plaintiff’s counsel to assess whether nothing additional was learned, such that the removal was based purely on non-voting, which is prohibited. Alito suggested that the state learned something, non-response, from the unreturned notifications, but counsel countered that no information about residency was obtained.
Neither side claimed that they could provide accurate numbers of the total disenfranchised, even though the arguments critically turn on the extent of discrimination taking place, and how those who do not return notifications should be classified so as to estimate what percentage of non-respondents had actually moved. Plaintiff’s argument rested partly on the claim that, because the percentage of Americans who move every year is low relative to the share of non-respondent voters in Ohio who were purged, the Ohio process necessarily results in false-positives, and must “vastly over purge” voters from registration lists.
An arms race: the technology of discrimination v the technology of empowerment
The amount of time spent on such questions in the Ohio arguments reaffirms the extent to which the identification and measurement of voter behavior is going to be central to challenging voting rights as we move forward. Further, estimating the impact of administrative procedures and eligibility requirements, while statistically difficult, is going to be of even greater importance, if we need to untangle compound effects in order to assess their performance.
Together, this week’s cases show that scientists, the courts, and the public need to advocate for greater scientific integrity, not only in the domain of legislative policymaking, but throughout the policy making process, including litigation, where improved methods and models can “provide a new understanding of how to give effect to our long-established governing principles.”