Fair: one rule for all 50 states, using nothing but population and legal boundaries. It never sees party, race or incumbents, not even to choose between maps. Each state gets the single plan that splits the fewest counties and is the most compact, so nobody, including us, picks the outcome.
Practical: every district is equal in population to within a few people, in one piece, and built from whole counties, towns and precincts. Every plan passes federal law and comes as a standard block file that election offices can load.
That's all of it. The details, and the proof that each rule is followed, are below.
Every figure is computed from 2020 Census blocks and compared with the maps in use for November 2026. How each claim is checked · How it looks from every side · Sources
Switch between the formula's plan, the districts in use for the November 2026 elections, and an annotated history of congressional maps since 1789. Hover over a district for its numbers, and click a state for a side-by-side scorecard. Partisan colors use each area's average two-party vote across every statewide race from 2016 to 2020; the formula never saw those numbers while drawing.
The Constitution apportions House seats among the states by population but says nothing about how a state should divide its seats. For most of the nineteenth century some states elected their whole delegation at large. Congress first required single-member districts in 1842, added equal population (1872) and compactness (1901, repeated in 1911), then quietly dropped all three standards in the Reapportionment Act of 1929. For the next three decades, districts could go unredrawn for decades and differ in population many times over.
The word itself comes from an 1812 Massachusetts state-senate map signed by Governor Elbridge Gerry, which included a district so contorted that the Boston Gazette drew it as a salamander. The technique is simple:
Two very different doctrines govern district lines today. On population equality, the courts are strict: since Wesberry v. Sanders (1964) congressional districts must be "as nearly equal as practicable," and Karcher v. Daggett (1983) struck down a New Jersey plan whose largest and smallest districts differed by only 0.7%. On partisan fairness, the federal courts have withdrawn entirely. Rucho v. Common Cause (2019) held 5–4 that partisan-gerrymandering claims raise "political questions" that federal courts cannot resolve, partly because no one had offered a manageable standard for how much partisanship is too much.
The third doctrine, race, has narrowed sharply. Thornburg v. Gingles (1986) and Allen v. Milligan (2023) read Section 2 of the Voting Rights Act as sometimes requiring majority-minority districts. On April 29, 2026, Louisiana v. Callais (6–3) struck down Louisiana's second majority-Black district, saying Section 2 liability now needs a "strong inference" of intentional discrimination and that racially polarized voting must be shown in a way that partisanship cannot explain.
That leaves the field to the states: independent commissions (Arizona, California, Colorado, Michigan and others, upheld in Arizona State Legislature v. AIRC, 2015), state constitutional "fair districts" clauses, and state courts (Moore v. Harper, 2023). It also leaves an opening for the argument this site makes: a rule precise enough that a court, or a voter, can check whether it was followed.
Redistricting used to happen once a decade, after the census. In July 2025 President Trump urged Texas to redraw its map before the 2026 midterms, and other states followed, some to add Republican seats and some, in retaliation, to add Democratic ones. Callais then set off a second round across the South. According to AP and NCSL, nine states will use new House maps in November 2026: Alabama, California, Florida, Louisiana, North Carolina, Ohio, Tennessee, Texas and Utah. Sabato's Crystal Ball (Sept. 15, 2026) put the net effect at R+2 to R+6 seats. AP estimated that about one in ten potential voters is now in a different district than in 2024.
Sources: NCSL mid-decade tracker (Sept. 11, 2026), AP (Sept. 28, 2026), Sabato's Crystal Ball (Sept. 15, 2026), CRS LSB11431, and court records. The "Enacted (2026)" layer on the map is the Census Bureau's official block equivalency file for the 120th Congress, which includes every redraw above that is in use this November.
"Fair" has no single definition, and there is a theorem to prove it. Alexeev and Mixon (2018) showed that no districting rule can always guarantee equal population, a minimum compactness and a bounded efficiency gap at the same time. And any rule that chooses among maps by a fairness score can be steered, because there are many reasonable scores and they disagree (see Can it be tuned?). So these principles promise a fair procedure: the simplest rules that produce one answer, with no step where a person, or a party's data, decides.
Each state's plan is defined by k centers c₁…cₖ (one per seat) and k weights w₁…wₖ. A place x belongs to the district that minimizes
Each district is therefore a convex polygon (a "power cell"), and each center is the population center of its district. The weights grow or shrink the districts until every one holds the same number of people. This is the population-weighted "moment of inertia" objective first used for nonpartisan districting by Weaver and Hess (1963–65), in the balanced power-diagram form of Cohen-Addad, Klein and Young (2018). In plain terms, it puts people as close as possible to the center of their own district.
The plan is tied to the census, not to whoever holds power. These steps were added after checking the rule against every idea of fairness we could find (the full scorecard); each one closes a gap without adding a setting anyone could tune.
Why no partisan correction? An earlier version of this project picked, among compact plans, the one with the least partisan bias. That step looked symmetric, but the choice of fairness measure, election years and tolerances moved the national result by about six seats. A rule with a dial can be turned. The simplest rule has no dial: the lean that remains comes from where people live (Chen & Rodden 2013), not from anyone's choice.
A districting plan can fail in court, or fail in practice at the county election office. We collected the arguments that have actually been used against plans, in federal and state law, election administration and the critiques of computer-drawn maps. Each one became an automated test that runs on every state, for the formula's plan and for the enacted 2026 map alike. Each state's results are in its map panel and in the table's Checks column.
Click a column to sort, or a row to open the state on the map. "D seats" count districts whose 2016–20 statewide composite is above 50% Democratic, which estimates lean, not who actually won. County splits are counted at block-group resolution for both plans.
A better map changes nothing unless someone with the power to adopt it does. The rule is one national standard, so its natural home is federal law. Article I §4 lets Congress "make or alter" state regulations for House elections, and Congress has used that power to require single-member districts (2 U.S.C. §2c) and, from 1842 to 1929, contiguous, compact and equal districts. A Neutral Lines Act would replace each state's separate congressional criteria with the same rule everywhere; state legislative maps are untouched. Until Congress acts, states can adopt the same rule themselves through initiatives, commissions and courts, as set out below.
Recommended name: Neutral Lines. The proposal would be the Neutral Lines Standard (the rule), the Neutral Lines Act (a state or federal statute adopting it) and the Neutral Lines Compact (an interstate agreement, below).
Public opinion is a tailwind, not a guarantee. In August 2025, California voters preferred keeping their independent commission by nearly two to one; three months later they approved Prop 50, setting it aside to counter Texas. Voters favor neutral processes in principle, and they will also back a partisan counter-gerrymander when it is framed as self-defense.
Transparent heuristic, not a forecast. Path: citizen-initiated constitutional amendment available (3), initiated statute only (2), an existing redistricting commission (2, counted only when no initiative exists), legislature only (0). Seat shock: how many seats the formula would move relative to the 2026 map, on the 2016–20 composite (0 seats: 3, 1: 2, 2: 1, 3 or more: 0). Checklist: +1 if the formula has no Review or Fail item. Outlook: High ≥ 6, Medium 4–5, Low ≤ 3. Single-district states are omitted.
An interstate compact in which each member adopts the Neutral Lines Standard, effective only once the members' combined seat shift relative to their current maps nets out within one seat. A Republican-leaning state's change is matched by a Democratic-leaning state's, so the partisan balance of the House is unchanged by joining, and the gerrymanders are retired in matched pairs. Compacts that affect federal elections likely need congressional consent (Art. I §10). That consent is itself a vehicle for the federal path below.
Sources: initiative availability, Ballotpedia (18 states allow initiated constitutional amendments, with Mississippi's process de facto suspended; 21 allow initiated statutes). Illinois's amendment initiative is limited to the legislative article, and courts blocked a 2016 redistricting initiative there. Polling: POLITICO–Citrin Center–Possibility Lab (Aug. 2025), via Election Law Blog.
pipeline/): build.py searches and selects the plans, refine.py finishes them on census blocks, score.py checks them, and export.py writes the files this page reads. The per-state certificates and block-group assignment files are in docs/data/plans/.Numbers in brackets throughout the page link here. Historical district stories in the map's History mode carry their own source links.