carmenaresearch/early-vote-tracker-methods

Early-Vote Tracker Methods: Bias & Checks

Reference tables on how early-vote trackers are built: county-lean estimates understated Democratic margins by a median 16 points, and bellwether counties barely exist in midterms.

Published by Carmena’s Overton & Electoral Research, an independent, non-partisan US election research firm. These are the public tables behind our charts. Party means voter registration, not vote choice, unless a column says otherwise. Figures are compared at the same number of days before Election Day in each year.

Files

• chart-county-bias.csv (8 rows): How far the county method understated the Democratic early-vote margin. Columns: state State; bias Bias of county method (points). Source page: https://carmenaresearch.com/data/#chart-county-bias
• chart-bellwethers.csv (2 rows): Share of counties that moved with the nation, within 4 points. Columns: midterms Midterms; low Low (%); high High (%). Source page: https://carmenaresearch.com/data/#chart-bellwethers
• chart-voterfile-access.csv (7 rows): Where voter files are free, with no account or form. Columns: place Place; access Access. Source page: https://carmenaresearch.com/data/#chart-voterfile-access
• chart-oh-agree.csv (3 rows): Ohio early voters: state file against the tracker. Columns: party Party; stateFile State file; tracker Tracker. Source page: https://carmenaresearch.com/data/#chart-oh-agree
• chart-tx-records.csv (3 rows): Texas early voters: state records against the tracker. Columns: party Party; stateRecords State records; tracker Tracker. Source page: https://carmenaresearch.com/data/#chart-tx-records
• chart-tx-new-voters.csv (3 rows): Texas 2026 early voters by vote history and party. Columns: history Vote history; democrats Democrats (%); republicans Republicans (%). Source page: https://carmenaresearch.com/data/#chart-tx-new-voters
• chart-new-voters-swing.csv (5 rows): Correlation of county new-voter share with the presidential swing. Columns: group Group; year Year; r Correlation (r). Source page: https://carmenaresearch.com/data/#chart-new-voters-swing

Provenance and method

Carmena’s Overton compiles cumulative early-ballot counts from state and county election records and joins them to voter registration and vote history. Calculations are ours; the method is described on our methodology page. Last updated 2026-10-23. Updated as our tracking runs, typically weekly during the early-voting period.

Caveats

• Tested only in the 8 states where party is known from records.
• A perfect past record adds 0–4 points of predictive edge.
• Cheapest paid files: Pennsylvania about $20, Alaska about $21.
• The two sources agree on the margin and disagree on independents.
• Adding 2022 and 2024 primaries in 4 counties (25% of votes) moved the margin about 0.6 pt. The tracker read D 49, R 42, Other 9 (D+7) on Oct 22.
• Nationally, 80% of early voters voted in both 2022 and 2024. “New” = voted in neither 2022 nor 2024.
• Correlation across counties, not a measure of individual voters. In 2024 the correlation was about zero.

Also on

• GitHub: election-data/early-vote-tracker-methods
• Kaggle: Early-Vote Tracker Methods: Bias & Checks
• Our data page: carmenaresearch.com/data

Licence and citation

CC BY 4.0. Credit “Carmena’s Overton & Electoral Research” and link to the file or page you used. Citation: Carmena’s Overton & Electoral Research. “Early-Vote Tracker Methods: Bias & Checks” (CSV). Last updated 2026-10-23. https://carmenaresearch.com/data/. Licensed CC BY 4.0.

More public data, with JSON copies and a changelog, is on our data page. In-depth data and custom research are available for commercial use on request. New analysis goes out by email first: subscribe.