Algorithmic Redistricting: Prove the Map, Don't Just Draw It

The strongest objection: Rucho already told federal courts they will not hear partisan-gerrymandering claims, and Callais just made the surviving Voting Rights Act claims harder to win, requiring a race-neutral illustrative map that also meets the state's own goals and present-day evidence of intent rather than a bare showing of discriminatory effect. A mandatory disclosure requirement generates more evidence for exactly the claim courts have said they won't adjudicate. A critic could reasonably ask what a published statistical outlier score accomplishes beyond a stronger press release, when the body most equipped to act on partisan-fairness evidence has already declined the invitation.

The position holds anyway, for three specific reasons. First, most state courts that have addressed partisan gerrymandering under their own constitutions, unlike the federal courts Rucho bound, have found limits there; ensemble evidence is exactly the kind of evidence Pennsylvania's and North Carolina's courts have already accepted from expert witnesses in redistricting litigation, and a standardized, public version of that evidence lowers the cost of bringing that case in every state with a receptive constitution, not just the ones wealthy enough to hire their own modelers. Second, Callais narrowed Section 2 by demanding present-day evidence of intent rather than a bare effects showing, which makes statistical outlier analysis more relevant to that claim, not less: showing a map deviates further from a race-neutral, criteria-compliant baseline than chance would predict is precisely the kind of current-map evidence a plaintiff needs to establish that a map's shape reflects a deliberate present-day choice rather than an incidental one, and generating the litigation-specific version of that baseline without race as a criterion, while also matching a state's own specified political goals where it has stated any, is what keeps it usable as illustrative-map evidence under Callais's full requirement, rather than excludable on the same ground the state's own racial gerrymander was or dismissed as failing to match the state's stated partisan aims the way Texas's does. Third, human commissions already use statistical fairness evidence today, independent of what any court does with it: Michigan's commission built lopsided-margins, mean-median, and efficiency-gap scoring directly into its own mapping software during the 2021-2022 cycle, evidence aimed at the commission's own legitimacy and internal deliberation, not at a future lawsuit.

One honest, narrow loose thread remains. Callais was decided ten weeks before this issue's last review, and no court has yet ruled on whether ensemble-outlier evidence satisfies its new intentional-discrimination standard in practice, on the three-way split this position uses (a neutral public baseline, a race-aware audit, and a separate litigation-specific construction matching a state's political goals) to keep the public disclosure layer immune from a state gaming its own baseline while still producing evidence usable under Callais, or on how precisely the litigation-specific version should encode a state's specified political goals, since partisan balance and incumbent protection resist the same clean mathematical specification population, contiguity, and compactness get. That is a live, specific, currently unanswered set of questions this position will need to revisit as courts and technical practice both develop, not a reason to doubt the argument built on everything that is already settled.

The most concentrated cost falls on incumbent legislators and parties currently drawing maps without independent scrutiny, in states controlled by both parties: Texas's Republican legislature and Illinois's, Maryland's, New York's, and Oregon's Democratic ones alike lose the informational advantage of being the only party that knows, before anyone else does, how far its map departs from a neutral baseline. This position accepts that loss as its entire point. A map-drawer who currently benefits from that asymmetry is precisely who this proposal is written to constrain.

States and redistricting bodies bear a modest compliance cost to run the analysis, mitigated directly by this position's federal technical-assistance grants and its reliance on GerryChain, an already-built, open-source tool rather than a system any state would need to build from scratch.

A more diffuse and less fully resolved cost falls on communities of interest whose shared identity isn't racial or language-minority and isn't well captured by population, contiguity, or compactness data: an economically interdependent region, a shared occupational community, a rural-urban boundary. Even a version of this tool that correctly includes racial data has no built-in way to weigh that kind of community. This position doesn't leave that gap unaddressed: it requires any federally funded ensemble tool to operate alongside a structured public community-of-interest input channel, modeled on Colorado's existing Redistricting Online public comment portal, so the quantitative baseline has a qualitative counterweight rather than standing alone as the final word on what a defensible map looks like.

Advocacy groups and under-resourced litigants who currently pay for expert witnesses to produce the equivalent analysis case by case are the clearest beneficiaries of this shift. Standardizing the evidence as a public output is the specific mechanism that narrows the resource gap between a well-funded state defending its map and a challenger who cannot currently afford the same modeling.