Live model · Gerrymandering
How Gerrymandering Actually Works
Gerrymandering is drawing district lines to pick the winners before anyone votes. Every district must hold the same number of people, but who goes where is the map drawer's choice. Pack the other party's voters into a few districts it wins by landslides, crack the rest across districts it narrowly loses, and fewer votes can win most seats.
The same 81 voters, district by district
- CH1 Gold
- CH2 Magenta
How to read the stage
The grid is 81 voters: 49 Gold (60%) and 32 Magenta, who cluster in a city. Thick lines are district borders: 9 districts of 9 voters, each one connected piece. Magenta's map packs Gold into three districts Gold wins 9 to 0, 8 to 1 and 8 to 1, then wins the other six 5 to 4. Magenta takes 6 of 9 seats with 40% of the votes. Under Squares Gold wins 6; under Gold's map, all 9.
Wasted votes and the efficiency gap
Stephanopoulos and McGhee call a vote wasted if it goes to a loser, or to a winner past the half needed to win. Cracking wastes losing votes; packing wastes surplus ones. Under Magenta's map Gold wastes 35.5 votes and Magenta 5. The difference over all 81 votes, 38%, is the efficiency gap. Their 2015 paper proposed treating a congressional map whose gap is worth two or more seats as presumptively unconstitutional.
Gerrymander or geography?
To tell them apart, draw many legal maps with no partisan aim and see where a map falls. ReCom, a sampler published by DeFord, Duchin and Solomon, merges two neighboring districts and splits them again along a random spanning tree. Here 1,000 such maps give Gold 4 to 8 seats, 6 most often. None gives Gold 9, as Gold's map does, or 3, as Magenta's does: both maps are outliers.
1,000 legal maps drawn blind
- CH1 Gold
- CH2 Magenta
Why 2026 has new maps
Maps are usually redrawn once a decade, after the census. Before the 2026 midterms, Texas, California (by Proposition 50), North Carolina, Ohio, Florida, Tennessee and Louisiana redrew theirs mid-decade, and a Utah court imposed a new one. Missouri's new map is on hold until a referendum, and Virginia's was struck down. 33 states require congressional districts to be contiguous, the rule the flood fill checks here.
What the model leaves out
Real districts average about 761,000 people, built from census blocks, with tiny population differences allowed. Map drawers also weigh county lines, compactness and the Voting Rights Act, and real voters split tickets and stay home. The two parties' maps here come from a short search that keeps any change that doesn't cost its party a seat: a stand-in for a map drawer's choices, not a real one.
Questions
- Is gerrymandering illegal?
- Not in federal court, for partisan maps. In Rucho v. Common Cause (2019) the Supreme Court held that partisan gerrymandering claims present political questions beyond the reach of the federal courts. State courts can still strike maps down under state law, as a Utah judge did in November 2025. Districts must still have equal populations, and drawing them by race is unconstitutional without a compelling reason.
- What are packing and cracking?
- Packing crams the other party's voters into a few districts it wins by landslides, wasting their surplus votes. Cracking spreads the rest across many districts where they fall just short, wasting every one of their votes. Magenta's map above does both: Gold wins three districts with 25 of its 49 voters, and its other 24 lose six districts 5 to 4.
- What did the Supreme Court decide in Louisiana v. Callais?
- On April 29, 2026, the Court ruled 6 to 3 that Louisiana's map with a second majority-Black district was an unconstitutional racial gerrymander, because the Voting Rights Act didn't require that district. It read Section 2 to impose liability only when the evidence supports a strong inference that a state intentionally drew districts to give minority voters less opportunity because of their race. Louisiana passed a new map in May.
- Can a computer draw fair districts?
- A computer can draw thousands of legal maps in seconds, which is how ensembles like the one above are built. It can't decide what fair means: compact shapes, whole counties, competitive races or seats in proportion to votes. Random maps show what's typical for a state's geography, and that isn't always proportional: here the typical map gives 60% of the votes 6 of 9 seats.
Drill the mechanism
Checking that a district is one piece is a flood fill: start at one voter, visit every neighbor in the same district, and see whether the fill reaches all nine. Number of Islands runs the same search over a grid of land and water, counting how many separate pieces there are.