The Gerrymandering Machine: How Software Ate the Map

Redistricting technology has advanced to the point where mapmakers can predict election outcomes at the precinct level with greater than 95% accuracy. This means the people who draw the maps can determine who wins before a single vote is cast, which raises the question of why we hold the election at all.

WHAT HAPPENED

• Post-2020 redistricting produced maps in 43 states using commercial GIS software
• Maptitude for Redistricting (Caliper Corporation) is used by 38 state legislatures
• The software can generate and evaluate millions of potential maps against partisan criteria in hours
• Academic studies estimate that post-2020 gerrymandering shifted approximately 16–20 House seats from competitive to safe
• The number of truly competitive House districts (Cook PVI between D+5 and R+5) has declined from 164 in 1992 to 82 in 2024

THE MECHANISM

Modern gerrymandering operates at a precision that would be unrecognizable to Elbridge Gerry. The original gerrymander (1812) was drawn by hand on a paper map using county-level population data. The 2021 gerrymander was drawn by algorithm on a computer screen using block-level census data, voter registration files, consumer data profiles, and historical election returns at the precinct level. The software can evaluate millions of potential district configurations against multiple optimization criteria simultaneously: maximizing partisan advantage, minimizing legal vulnerability, and maintaining population equality to within single digits.

The result is maps that produce predetermined outcomes with remarkable consistency. In the 2024 election, 388 of 435 House races (89%) were won by the candidate whose party was favored by the district's partisan lean. Of the 47 "upsets," 34 occurred in districts with a PVI of less than R+3 or D+3 — districts so close to neutral that any wave election would flip them regardless of the map.

POLLERBULL SIGNAL

• What moves odds: Gerrymandering is already baked into our House forecast model through district-level PVI scores. The model treats the current maps as a structural constraint: approximately 190 seats are safe Republican, 180 are safe Democratic, and 65 are genuinely competitive. All electoral volatility occurs in those 65 seats.
• What would falsify this: If independent redistricting commissions (now used in 10 states) produce maps that significantly increase the number of competitive seats, the structural constraint loosens. Current trend: commission-drawn maps have approximately 15% more competitive seats than legislature-drawn maps.