Data
This atlas compares congressional district plans using presidential election baselines from 2008, 2012, 2016, 2020, and 2024. Districts are colored by Democratic two-party vote share: Democratic votes divided by Democratic plus Republican votes.
Elections
The atlas scores each plan against a selected presidential baseline. For the current atlas plans, election returns are aligned to 2020 Voter Tabulation Districts (VTDs), census tracts, counties, or statewide units where those are the simulation input geographies. These units are then aggregated to congressional districts.
Older presidential data come from established precinct-level election sources. For 2008, the atlas uses data from Dave’s Redistricting App and the Harvard Election Data Archive (Dave’s Redistricting App 2025; Ansolabehere and Rodden 2025). For 2012, the atlas uses Ryne Rohla’s data available through Dave’s Redistricting App block files (Rohla 2012). For 2016 and 2020, the atlas uses Voting and Election Science Team data aligned by the ALARM Project (Voting and Election Science Team 2018, 2020; Kenny and McCartan 2021).
The 2024 presidential baseline comes from Kenny (2026). That dataset draws on several precinct-level sources. Redistricting Data Hub provides returns for AL, AZ, GA, IL, KS, LA, MA, MD, MO, MS, NC, NE, NH, OH, SC, TN, TX, UT, VA, WA, and WI (Redistricting Data Hub 2025). New York returns come from Benjamin J. Rosenblatt and Twitter’s @cincy9 (Rosenblatt 2025). California and Michigan data were collected directly from state and local election offices by Christopher T. Kenny. Returns for the remaining states were collected by Josh Metcalf (Metcalf 2024).
Precinct returns are projected to the census block level using the geomander R package, allocating votes proportionally by voting-age population (Kenny 2025). Block-level totals are then aggregated to VTDs, tracts, counties, statewide units, or districts. For AK, DE, ND, SD, VT, and WY in 2012, precinct-level results are imputed from surrounding presidential baselines and county-level returns from the MIT Election Data and Science Lab using a logit-shift procedure (MIT Election Data and Science Lab 2018; Rosenman et al. 2023).
Enacted plans
The enacted maps show congressional district plans for 2022, 2024, and 2026. The 2026 enacted map uses updated assignments for states with mid-decade congressional redistricting and otherwise carries forward the 2024 map. The current 2026 layer includes new assignments for CA, FL, LA, MO, NC, OH, TN, TX, and UT.
Simulated plans
The simulated maps come from the ALARM Project’s 50-state congressional simulation ensemble (McCartan et al. 2022). These simulations are alternative congressional maps generated under standard redistricting criteria, including contiguity, population equality, compactness, and state-specific rules.
The atlas shows five state-by-state summaries from that ensemble: most Democratic, 75th percentile, median, 25th percentile, and most Republican. Each simulated plan can be scored against any of the presidential baselines available in the atlas.
Optimization plans
The Safe Max and Max plans were generated by Christopher T. Kenny using redist (Kenny et al. 2026). They are prototypes and will likely be updated in the future.
The county-splits optimization plan comes from Shahmizad and Buchanan (2025). The source files are available on Dave’s Redistricting Atlas via links on GitHub. At-large states are omitted. Hawaii uses the median 2020 simulated plan from McCartan et al. (2022) because each island is a county and all legal plans have exactly one county split.