Carmena’s Overton & Electoral Research: research

Nearly half of the ‘party’ in early-vote trackers is a model

Fri, 23 Oct 2026 00:00:00 GMT

In 21 states, NBC’s party split is modelled by TargetSmart. Those states hold 48% of this year’s early votes.

In 21 states, the NBC News early-vote tracker labels its party column “modeled by TargetSmart”. Those states hold 48% of the early votes cast in 2026 as of Oct 23: 10.4 million of 21.7 million ballots. So when the tracker says the national early electorate is 49% Democratic, close to half of the ballots behind that figure carry a modelled party, not one copied from a state record.

The 21 states are Alabama, Arkansas, Georgia, Hawaii, Illinois, Indiana, Michigan, Minnesota, Missouri, Mississippi, Montana, North Dakota, Ohio, South Carolina, Tennessee, Texas, Utah, Virginia, Vermont, Washington and Wisconsin.

That blend sits under one of the most shared lines of the month: this early electorate is far more Democratic than 2024’s, D 49 to R 34 against D 42 to R 40 at the same point. The comparison may well hold. But it compares two mixes of registration and modelled party, and the headline does not say how much of each.

What it means

The national split is a blend of two kinds of number. In some states it comes from registration files. In others it is a model’s best guess. They deserve different levels of trust, and the headline figure hides which is which.

It matters most in the Senate races. Three of the ten battlegrounds we track have modelled party only: Michigan (D+17), Minnesota (D+27) and South Carolina (D+2). Ohio and Texas are modelled in the tracker too, so for those two we use the states’ own primary records instead: D+5.9 and D+3.4.

The obvious alternative is worse. Weighting each county’s early votes by how that county voted in 2024 understated the Democratic margin by 7 to 41 points, with a median of about 16, in the 8 states where we could test it against known party. A vendor model that is close is more useful than a county shortcut that is far off.

Every party figure on this site carries a source tag: reg. for registration files, primary for primary history, model for vendor estimates.

What it doesn’t mean

It does not mean the model is wrong. Where we can check it against registration files, as in Iowa, Maine and North Carolina, the tracker matched the official counts within about one point in both 2022 and 2024.

It does not tell you which way any error runs. In Ohio the model and the state file agree on the margin to a tenth of a point. In Texas the tracker read D+7 on Oct 22 and R+1 on Oct 23, while the state’s records, through Oct 21, give D+3.4: the model was more Democratic one day and more Republican the next.

The larger disagreement is about independents, not the margin. Ohio’s file puts 39.4% of early voters in Other; the tracker puts 8%. Texas records put 34.0% in Other; the tracker puts 9%. A tracker can match the D−R gap and still describe a very different electorate.

And registered party is still not vote choice. Democrats have voted early in larger numbers since 2020, so a Democratic early electorate says more about who votes early than about who wins.

How we checked

Registration states. For Iowa, Maine and North Carolina, which record party, we compared the tracker’s 2022 and 2024 splits with the official files. They matched within about one point.

North Carolina. We tested the usual shortcut, party of the last primary voted, on 1,096,068 early voters. It put the margin at D+21.2 against a true D+20.5, but sorted 24% into Other against a true 34%.

Ohio. The Secretary of State’s absentee file, with party from the last primary, gives D 33.3, R 27.4, Other 39.4: D+5.9. The tracker gives D 49, R 43, Other 8: D+6.

Texas. We joined state early-vote rosters to 2026 primary rosters: D 34.7, R 31.3, Other 34.0, or D+3.4, across 2.16M early voters. The tracker gave D 49, R 42, Other 9 (D+7) on Oct 22 and D 45, R 46, Other 9 (R+1) on Oct 23. Our state-records figure has not been refreshed since Oct 21. Adding 2022 and 2024 primaries in four counties, 25% of votes, moved our margin by only about 0.6 points.

What we publish. Until Election Day we report party from state records wherever they exist, show the tracker figure beside it when the two differ, and tag every number with its source.

Democrats lead the generic ballot by 9 points, ahead of 2018’s wave

Fri, 23 Oct 2026 00:00:00 GMT

The Silver Bulletin average is D+9.0 at 11 days out. At the same point, 2018 stood at D+8.2 and 2022 at R+0.6.

On Oct 23, 2026, 11 days before Election Day, the Silver Bulletin average of generic ballot polls had Democrats ahead by 9.0 points. At the same point in 2018, the 538 average had Democrats ahead by 8.2 points, and they went on to win the national House vote by 8.6. At the same point in 2022, the 538 average had Republicans ahead by 0.6 points, and they won the House vote by 2.7.

The generic ballot asks people which party they would vote for in their House race, without naming candidates. It is the closest thing to a national poll of a midterm, and it is the indicator most often compared across cycles.

What it means

A 9-point Democratic lead 11 days out is larger than the lead Democrats held at the same point in 2018, the last midterm that produced a Democratic wave. In 2018 the average barely moved in the final stretch: from D+8.2 at 11 days out to D+8.6 on Election Day, which matched the final House vote of D+8.6.

The 2026 lead is also far from 2022’s position. Four years ago the average sat near a tie (R+0.6) at this point and drifted to R+1.2 by Election Day. On those two benchmarks, the 2026 environment looks closer to 2018 than to 2022.

Recent individual polls sit around the average. In the last 30 days the published polls include YouGov likely-voter samples at D+7 and D+9 and a New York Times/Siena College likely-voter poll at D+11, all ending Oct 19–20.

What it doesn’t mean

It is not a seat count. The national House vote and the number of seats each party wins are related but not the same, because seats depend on where votes fall, which seats are contested and how districts are drawn. A national lead of 9 points does not translate into a fixed number of seats.

It is not error-free. In 2022 the average was R+0.6 at 11 days out and R+1.2 on Election Day, while the House vote was R+2.7. Polls understated Republicans by about 2 points from this point in the campaign. An error of that size today would still leave Democrats ahead, but it shows the average can miss in one direction.

It is not one consistent series. The 2026 figure comes from the Silver Bulletin average; the 2018 and 2022 figures come from 538’s average. Both combine many polls and adjust for pollster differences, but they are built by different teams with different rules, so a gap of under a point between 2026 and 2018 is within the difference the methods alone could make.

It does not say who will turn out. Many polls in the average are of likely voters, but a likely-voter screen is an estimate of who will vote, and in a midterm turnout is lower and less predictable than in a presidential year.

How we checked

We took the Silver Bulletin generic ballot average as published on Oct 23, 2026, and 538’s historical generic ballot averages for 2018 and 2022. We aligned the three cycles by days before Election Day: “11 days out” is Oct 23, 2026, Oct 28, 2022 and Oct 26, 2018.

For the final House vote we used the national popular vote as published: D+8.6 in 2018 (Cook Political Report) and R+2.7 in 2022. Our own county-level build from MIT Election Data and Science Lab returns gives D+8.4 for 2018 and R+3.1 for 2022; the small gaps come from uncontested and missing districts in the precinct files, so we report the published figures.

Margins are Democratic share minus Republican share, in points. D+ means a Democratic lead and R+ a Republican one.

Democrats’ early-vote edge grew from 2 points to 15

Fri, 23 Oct 2026 00:00:00 GMT

Early voters so far are D 49, R 34, Other 17. At this point in 2024 they were D 42, R 40, Other 18.

Of the 21.7 million Americans who had voted early by Oct 23, 2026, 49% were Democrats, 34% Republicans and 17% neither, according to the NBC News early-vote tracker, which uses TargetSmart data. At the same point in 2024, early voters were 42% Democratic, 40% Republican and 18% Other. The Democratic lead among early voters grew from 2 points to 15.

What it means

Democrats make up a larger share of the early electorate than they did two years ago, and Republicans a smaller one. The Other share is almost unchanged, so the shift is between the two parties: Democrats up 7 points, Republicans down 6.

Part of the shift rests on firm ground. In 29 states and DC, which hold 52% of early votes, NBC does not mark party as modelled. In the registration states we could check, Iowa, Maine and North Carolina, the tracker matched official records within about one point in both 2022 and 2024.

It is also a sign of who is motivated to vote early this year. A 15-point Democratic edge means Democratic voters are banking their ballots at a higher rate than Republican voters, at least so far.

What it doesn’t mean

Registered party is not vote choice. Some registered Democrats vote Republican and some registered Republicans vote Democratic, and the 17% Other group splits both ways. The tracker counts party labels, not votes.

The 2024 comparison flatters the shift. Since 2020, Democrats have voted early in larger numbers than Republicans, and 2024 was a presidential year with a different electorate. Comparing a midterm early vote with a presidential one mixes the change in mood with the change in who turns out.

The better yardstick is 2022, the last midterm. At the same point in 2022 the tracker read D 47, R 35, Other 18: a 12-point Democratic edge. Republicans still won the 2022 national House vote by 2.7 points. So a double-digit Democratic lead among early voters is normal for a midterm and, on its own, says little about the result. The 2026 edge is 3 points wider than 2022’s, not 13.

Part of the party column is a model. In 21 states, which hold 48% of this year’s early votes, the tracker’s party is estimated by TargetSmart rather than taken from a state record. In Texas, the state’s own records through Oct 21 give D+3.4, while the tracker moved from D+7 on Oct 22 to R+1 on Oct 23. In Ohio they agree on the margin (D+5.9 against D+6) but not on how many voters are independents.

How we checked

We read the national party split from the NBC News tracker on Oct 23, 2026, and from archived copies of the 2024 and 2022 trackers at the same point before Election Day: Oct 25, 2024 and Oct 28, 2022. NBC publishes the split as whole percentages, so each figure is precise to about half a point.

We compared the tracker with official files in the states that record party. In Iowa, Maine and North Carolina, the tracker’s 2022 and 2024 splits matched the state files within about one point. For the 21 modelled states we have no national check; in Ohio and Texas we built our own figures from state primary records, described in separate posts.

The margins here are Democratic share minus Republican share, in points. A 15-point edge is 49 minus 34.

One in five of 2022’s voters have already voted

Fri, 23 Oct 2026 00:00:00 GMT

21.7M early ballots equal 19.7% of 2022’s total turnout. At the same point in 2024 the figure was 22.7%.

As of Oct 23, 2026, 11 days before Election Day, about 21.7 million Americans had cast an early ballot, by mail or in person. That equals 19.7% of everyone who voted in the 2022 midterm. At the same point in 2024, early ballots equalled 22.7% of that year’s final turnout. In raw numbers, 2026 is at 61% of 2024’s count for the same day.

The exact count was 21,654,855 ballots. Of those, about 12.2 million were returned by mail and about 9.5 million were cast in person.

What it means

A large share of the 2026 electorate has already voted. If total turnout this year matches 2022’s, roughly one ballot in five is already in. That makes the early vote a real part of the result, not a sample of it.

The comparison with 2024 is the fairest one available for the share of turnout. In 2024, 35.5 million early ballots had been cast at the same point, which was 22.7% of that year’s 156.8 million total. This year’s 19.7% is measured against 2022’s turnout of about 110 million, so the two shares answer the same question: how much of a typical electorate for that kind of year has already voted.

The two shares are close. A midterm electorate that is banking its votes at nearly the rate of a presidential one, relative to its own size, suggests early voting has become a settled habit rather than a 2020 exception.

What it doesn’t mean

It is not a turnout forecast. The share uses 2022’s turnout as a yardstick. If more people vote in 2026 than in 2022, the true share already cast is lower than 19.7%; if fewer vote, it is higher.

It does not mean one in five of the same people. The 21.7 million early voters this year are not all 2022 voters. Some voted in 2024 only and some have not voted in a general election in either year. The figure compares counts, not individuals.

The 61% figure is not a sign of low interest. 2024 was a presidential election, with about 47 million more voters than 2022. A midterm running at 61% of a presidential year’s raw count is consistent with a normal midterm turnout gap.

It says nothing about party on its own. Who these early voters are, and how far that tells you anything about the result, is a separate question with its own caveats.

How we checked

We took the national early ballot count from the NBC News early-vote tracker on Oct 23, 2026, which compiles state reports, and the same tracker’s 2024 count at the same point before Election Day. We divided each by total ballots cast in the comparison year: 109,979,856 in 2022 and 156,766,239 in 2024.

“11 days out” is Oct 23, 2026; the 2024 comparison point is Oct 25, 2024. Counts can lag states by a day or more, and some states report only some ballot types, so the national total is a floor rather than an exact figure.

State-level counts vary widely around the national pace. Our Senate battleground table shows each state against its own 2024 level, with the date of each state’s party figures marked.

Senate battlegrounds are voting early at 50–98% of their 2024 pace

Fri, 23 Oct 2026 00:00:00 GMT

None of the ten states is ahead of 2024, a presidential year. Against 2022, the last midterm, nine of the ten are ahead.

As of Oct 23, 2026, 11 days before Election Day, every one of the ten Senate battlegrounds we track had cast fewer early ballots than at the same point in 2024, a presidential year. Alaska is closest, at 98% of its 2024 count. Minnesota is furthest behind, at 50%. Against 2022, the last midterm, nine of the ten are ahead.

In full, against 2024: Alaska 98%, Kansas 88%, Texas 78%, Michigan 69%, South Carolina 68%, Ohio 66%, Iowa 65%, Maine 63%, New Hampshire 51%, Minnesota 50%.

What it means

This is a midterm voting early at a midterm scale. 2024 had far higher turnout than any midterm, so a state at 65% of its 2024 early vote is not necessarily slow.

Against 2022 the picture reverses. South Carolina is at 239% of its 2022 count at this point, Minnesota 149%, Texas 137%, Ohio 117%, Maine 116%, Alaska 110%, Michigan 107% and Iowa 105%. Only New Hampshire is behind, at 90%. Kansas shows 210%, but its 2022 figure is a single snapshot we could not check against the state.

Early voters lean Democratic in nine of the ten states, by the best source each state allows: Maine D+32.0, Minnesota D+27 (model), Iowa D+18.9, Michigan D+17 (model), New Hampshire D+12, Kansas D+8.6, Ohio D+5.9, Texas D+3.4 and South Carolina D+2 (model). Alaska’s early voters are R+8.

What it doesn’t mean

Pace is not turnout. Voters can move between early voting and Election Day without any change in how many vote. A state ahead of its 2022 pace may be seeing more early voting, more voting overall, or both.

It does not say who is winning. A D+19 early electorate in Iowa is a registration count, not a vote margin. Registered party is not vote choice, and Democrats vote early more than Republicans. The three model-tagged states, Michigan, Minnesota and South Carolina, carry the most uncertainty about party.

The party figures are not all from the same day. Iowa and Kansas come from state data of Oct 23; Maine from the state’s file of Oct 20, the latest posted. Ohio’s D+5.9 is from the state file of Oct 22 and Texas’s D+3.4 from state rosters through Oct 21; neither was refreshed on Oct 23. On Oct 23 the NBC News tracker, which models party in Texas, had Texas at R+1.

Some states are small. New Hampshire has only about 19,000 early votes so far.

How we checked

For the 2026 counts we used UF Election Lab’s Oct 23 files, which mirror state data, for Iowa, Kansas, Michigan, Minnesota, Ohio, South Carolina and Texas, and the NBC News tracker for Alaska, Maine and New Hampshire. The 2022 and 2024 counts at the same point come from the NBC News tracker in each year. Minnesota’s official count, which had not updated since Oct 16 earlier in the week, appears in UF’s Oct 23 file at 419,189 ballots.

Party comes from the best available source per state, tagged in the table: reg. for state registration (Alaska, Iowa, Kansas, Maine, New Hampshire), primary for party of last primary from state records (Ohio, Texas), and model for TargetSmart’s estimate via NBC (Michigan, Minnesota, South Carolina). Alaska and New Hampshire publish no official party split, so their figures are NBC’s, which carry state registration.

Texas Democrats are bringing new voters; Republicans are bringing regulars

Fri, 23 Oct 2026 00:00:00 GMT

8% of Democratic early voters in Texas are new, against 4% of Republicans. 84% of Republicans voted in both 2022 and 2024.

Among Texans who had voted early in the 2026 general by Oct 23, 8% of Democrats had voted in neither the 2022 nor the 2024 general election, against 4% of Republicans. Another 16% of Democrats had voted in 2024 but not 2022, against 11% of Republicans. 84% of Republican early voters had voted in both elections, against 74% of Democrats. The figures come from the NBC News / TargetSmart Texas page, where party is modelled.

What it means

The Democratic early vote in Texas is drawing more heavily on people who do not usually vote in midterms. About one Democratic early voter in four (24%) either skipped 2022 or skipped both elections. For Republicans it is about one in seven (15%).

The Republican early vote is mostly habitual voters. Five in six voted in both of the last two general elections. That pattern fits a party whose base votes in every cycle.

For both parties, the mix still leans heavily on regulars. Across all Texas early voters, 77% voted in both 2022 and 2024, 15% in 2024 only, 1% in 2022 only and 7% in neither. Nationally the share who voted in both is 80%.

The group TargetSmart does not assign to either party is the least habitual: 54% voted in both, 28% in 2024 only and 16% in neither.

What it doesn’t mean

The party labels are a model. Texas has no party registration, and NBC marks its Texas party column as modelled by TargetSmart. New voters are the hardest group to model, because they have little history to model from. A modelled “new Democrat” is a guess about someone with few records.

The state’s own records cannot check this. When we sorted Texas early voters by 2026 primary history, a third had no primary ballot at all. In Tarrant County, which posts past general-election rosters, 94% of the 180,830 people who voted in 2024 without having voted in the county’s 2020 or 2022 general had no primary history either. Primary records describe habitual voters well and new voters hardly at all.

“New” here means no vote in the 2022 or 2024 general. Some of these voters will have voted before 2022, or in another state.

It does not mean new voters decide the result. 8% of Democratic early voters is a small group, and the early vote is not the whole electorate.

How we checked

We read the vote-history breakdown from the NBC News / TargetSmart Texas page in two snapshots on Oct 23, 2026, at 03:30 and 17:30 GMT. Both gave the same shares. The later snapshot covered 1,039,431 Democratic, 1,046,185 Republican and 208,668 Other early voters, out of 2,294,284. NBC publishes shares as whole percentages, so each is precise to about half a point.

The national 80% is our weighted sum, from the 17:30 GMT snapshot, of the same breakdown across states on NBC’s pages.

For the Tarrant County comparison we used the county’s posted general-election and primary rosters for 2020 to 2024, kept only voter IDs and counted voters. “New” there means no vote in the county’s 2020 or 2022 general rosters, because earlier rosters are not posted.

Texas early voters: D+3 on the state’s own records, R+1 in the tracker

Fri, 23 Oct 2026 00:00:00 GMT

Joining state early-vote rosters to 2026 primary rosters gives D 34.7, R 31.3, Other 34.0 across 2.16M early voters.

Texas has no party registration, so trackers estimate the party of its early voters. Using the state’s own records instead, we matched 2,159,376 Texans who voted early through Oct 21, 2026, to the 2026 primary and runoff rosters. 34.7% last took a Democratic primary ballot, 31.3% a Republican one, and 34.0% voted in neither: a Democratic edge of 3.4 points. On Oct 23, two days later, the NBC News / TargetSmart tracker put Texas early voters at D 45, R 46, Other 9: R+1. On Oct 22 the same tracker had read D 49, R 42, Other 9, or D+7.

What it means

On the state’s records through Oct 21, the Texas early vote leans Democratic. In raw counts, 749,324 early voters last took a Democratic ballot and 676,293 a Republican one. The tracker has swung from D+7 to R+1 in a day, so the two sources now disagree on the sign, although they cover different dates.

Much of the gap is about who counts as Other. A third of Texas early voters did not vote in a 2026 primary. TargetSmart’s model assigns most of them to a party, leaving 9% Other. On a two-party basis, primary history gives Democrats 52.6%; the tracker gave 53.8% on Oct 22 and 49.5% on Oct 23.

The Democratic share is falling by the day, which is consistent with the tracker’s move. On Oct 19, the first day of in-person early voting, voters with a Democratic primary history outnumbered Republicans 307,000 to 243,000. On Oct 21, Republicans led 187,000 to 181,000, and voters with no 2026 primary were the largest group.

What it doesn’t mean

Primary history is not vote choice. In March 2026 more Texans voted in the Democratic primary (2.30 million) than the Republican one (2.15 million), which is unusual for the state. So “last primary Democratic” is partly a measure of who turned out in March, and overstates the Democratic lean of the electorate.

It does not settle the Other third. Voters with no 2026 primary could lean either way. Vendor models disagree on them: on Oct 22, L2’s public dashboard had Texas at D+6, VoteHub’s primary-history measure at D+4, and On Point’s county-based model at R+6. The sources do not agree on the sign.

It is not complete. The state rosters we joined cover 95.1% of early ballots reported by the Secretary of State, including 97.1% of in-person votes but only 56% of mail ballots. The Oct 22 roster was partial when we pulled it.

How we checked

We downloaded the Texas Secretary of State’s daily Early Voting Roster for the Nov 3, 2026 general, Oct 14 to Oct 22, and removed duplicates by voter ID. We joined it to the statewide rosters for the March 3, 2026 primaries and May 26 runoffs. The most recent 2026 primary ballot wins: the runoff first, then March. 124 voters on both parties’ rosters were counted as Other.

To test whether a one-year window distorts the result, we used 2022 and 2024 primary rosters that four counties post publicly: Tarrant and Hidalgo for both years, Dallas and Bexar for 2024. Those counties hold 25.2% of Texas early voters. Adding the older primaries cut Other by 4 to 7 points per county but moved the margin only about 0.6 points toward Republicans: Tarrant −1.8, Hidalgo +2.8, Dallas −0.2, Bexar −1.1.

The state-records figures were not refreshed on Oct 23, because the rosters need a browser download. Names and precincts were dropped as the files were read. Only voter IDs and aggregate counts were kept.

In Ohio, every source agrees on the margin and disagrees on independents

Thu, 22 Oct 2026 00:00:00 GMT

The state file says D+5.9 with 39.4% Other. The tracker says D+6 with 8% Other.

Ohio’s Secretary of State publishes a free file of every absentee and early ballot, with each voter’s party taken from the last partisan primary they voted in. As of Oct 22, 2026, it listed 838,018 returned ballots: 33.3% from voters whose last primary was Democratic, 27.4% Republican and 39.4% with no partisan primary. That is D+5.9. The NBC News / TargetSmart tracker, counting the same ballots, says D 49, R 43, Other 8: D+6.

What it means

The state file and the tracker agree on the margin almost exactly, from very different methods. One reads a field in the state’s records; the other is TargetSmart’s model. VoteHub’s primary-history measure is close too, at D+6.6.

On a two-party basis the sources sit in a narrow band. The state file gives Democrats 54.9% of the two-party early vote. The tracker gives 53.3%, L2’s public dashboard 55.7%, On Point 56.0% and VoteHub 56.3%.

The Ohio early vote leans Democratic on every source we found. The agreement on the margin is not new: at a similar point in 2024, 13 days out, the 2024 state file with party as of 2024 gave R+10.0 and the tracker gave R+8.

What it doesn’t mean

The sources do not agree on independents. In the state file, 39.4% of early voters have no partisan primary on record. The tracker has 8% Other. Those numbers measure different things. The state’s Other is everyone who skipped recent primaries, including many habitual partisans. The tracker’s 49–43 is a modelled partisanship score that puts most of them into a party.

“Every source” has limits. L2 and On Point put the margin at D+10 and D+9, three to four points more Democratic than the state file. Their methods are partly opaque, and both assign party to most voters.

The primary record has its own tilt. The May 2026 primary drew mainly committed Democrats and Republicans, and a contested primary on one side can inflate that side for a cycle. In North Carolina, where we could test the method, it got the margin within a point but understated the unaffiliated share by 10 points.

Party is not vote choice. Neither a primary ballot nor a modelled score says how anyone will vote on Nov 3.

How we checked

We read the Secretary of State’s export as of 11:00 ET on Oct 22, 2026: 1,197,163 requests for the Nov 3 general, with no duplicate voter IDs. A ballot counts as returned when it has a return date. The file contains names and addresses, so it was reduced to counts by county, party and status in the browser and never saved.

Returned ballots split 348,877 by mail and 489,141 in person, using the same-day rule UF Election Lab uses. By mode, mail voters were 28.0% D, 22.5% R and 49.5% Other; in-person voters 37.0% D, 30.9% R and 32.2% Other. Only 1,452 returned ballots came from Libertarian primary voters.

The 358,145 requested ballots not yet returned are 16.8% D, 17.2% R and 66.4% Other.

Our total matches UF Election Lab’s exactly, and NBC’s is 557 higher. We have not refreshed the state file since Oct 22. By Oct 23, UF counted 922,179 Ohio early ballots, and the tracker’s split was unchanged at D 49, R 43, Other 8.

North Carolina’s unaffiliated primary voters lean Democratic, 61–39

Tue, 20 Oct 2026 00:00:00 GMT

Of unaffiliated early voters with a primary history, 160K last took a Democratic ballot and 102K a Republican one.

Of the 377,501 unaffiliated voters who had cast an accepted early ballot in North Carolina by Oct 22, 2026, 262,596 have voted in at least one party primary since 2017. Among them, 160,307 last took a Democratic ballot and 102,289 a Republican ballot. That is 61% Democratic and 39% Republican.

The remaining 114,905 unaffiliated early voters, about 30%, have no primary ballot on record in that period.

What it means

Unaffiliated voters are the second-largest group in North Carolina’s early electorate: 34.4% of early voters, against 43.0% Democrats and 22.5% Republicans. How they lean matters for any reading of the early vote, and the primary record is one of the few direct traces of it in state data.

The trace points toward Democrats. Among unaffiliated early voters who have chosen a party’s primary ballot, Democratic ballots outnumber Republican ones by about 58,000.

This is also why the last-primary shortcut overstates both parties in North Carolina. It moves these 262,596 unaffiliated voters into a party, and the larger part of them into the Democratic column.

What it doesn’t mean

A primary ballot is not a general-election vote. Voters pick a primary for many reasons: the more competitive race, a local contest, a candidate they want to stop. An unaffiliated voter who last took a Democratic ballot may still vote Republican in November.

It does not describe all unaffiliated voters. The 30% with no primary history are left out, and they may lean differently from those who vote in primaries. In 2024, unaffiliated North Carolinians as a whole split very differently between the races for President and Governor, which shows how much their vote depends on the contest.

It does not measure the 2026 electorate as a whole. These are early voters only, as of Oct 22. Election Day voters, and voters who have yet to vote early, may differ.

It is not a trend. We did not compute the same split for earlier cycles, so we cannot say whether 61–39 is high or low for this group.

How we checked

We used the North Carolina State Board of Elections absentee file for the Nov 3, 2026 general, keeping only accepted ballots (1,096,068 in all, as of 05:19 ET on Oct 22, 2026). Each record carries the voter’s registered party.

We joined it on voter ID to the statewide voter history file dated Oct 18, 2026, which records the party ballot each voter chose in every primary since 2017. For each unaffiliated early voter, we took the most recent Democratic or Republican primary ballot.

The window changes the split a little. Counting only the March 2026 primary, 108,702 unaffiliated early voters took a Democratic ballot and 59,182 a Republican one, about 65–35, and more than half had no 2026 primary ballot. We report the full 2017–2026 window because it covers the most voters.

The counts come from the same join used in our test of the last-primary shortcut. No voter-level data was kept; only the aggregate counts were stored.

Your last primary is a good guess at your party, until it isn’t

Tue, 20 Oct 2026 00:00:00 GMT

Tested on 1,096,068 North Carolina early voters, the last-primary shortcut gets the margin right and the Other share badly wrong.

In states without party registration, such as Ohio and Texas, a common way to give early voters a party is to use the party of the last primary they voted in. We tested that shortcut in North Carolina, where the true registered party is on file. Across 1,096,068 accepted early ballots as of Oct 22, 2026, the shortcut put the Democratic margin at D+21.2 against a true D+20.5. But it put 24% of early voters in Other, against a true 34%.

What it means

For the margin, the shortcut works. Using every primary since 2017, it lands within 0.7 points of the true D−R gap. That is why we use it in Ohio and Texas, where no registration exists, and why we report the margin from it with some confidence.

For the shares, it does not. The proxy gives D 48.4, R 27.2, Other 24.4. The true registration split is D 43.0, R 22.5, Other 34.4. Both parties look larger than they are and the unaffiliated group looks smaller.

Two errors cause this, and they pull in opposite directions:

Unaffiliated voters who vote in primaries get a party. Of 377,501 unaffiliated early voters, 160,307 last took a Democratic primary ballot and 102,289 a Republican one. The proxy counts all 262,596 as partisans.
Partisans who skip primaries become Other. 101,759 registered Democrats and 50,582 registered Republicans have no primary on record since 2017. That is about 21% of registered partisans, and the proxy puts them all in Other.

The two errors almost cancel on the margin. They do not cancel on the size of each group.

What it doesn’t mean

It does not mean the shortcut works everywhere. North Carolina has party registration and lets us check; Ohio and Texas do not. Primary turnout patterns differ by state, and a contested primary on one side can tilt the proxy for a whole cycle.

The time window matters. Using only primaries since 2022, the proxy gives D+16.0. Using only primaries since 2024, D+15.3. Using only the 2026 primary, D+17.4, with almost half of early voters (49.2%) in Other. The more history the proxy can see, the closer it gets, and the full-history version is the only one within a point of the truth.

It does not tell you how anyone voted. Registration is not vote choice, and neither is a primary ballot. An unaffiliated voter who took a Democratic primary ballot may have done so because that primary was the more competitive one.

How we checked

We used the North Carolina State Board of Elections absentee file for the Nov 3, 2026 general, keeping accepted ballots only: 1,050,226 cast early in person and 45,846 by mail. Each record carries the voter’s registered party.

We joined it on voter ID to the statewide voter history file dated Oct 18, 2026, which lists every primary each voter took part in, with the party ballot chosen. For each early voter we took the most recent Democratic or Republican primary ballot within the window and called that the voter’s proxy party. Voters with no such ballot were Other.

We then compared the proxy with registration for every voter and counted where each registered group landed. No voter-level data was kept; only the counts above were stored.

Unaffiliated voters don’t split the same way twice

Mon, 19 Oct 2026 00:00:00 GMT

In North Carolina in 2024, unaffiliated voters were about 47% Democratic for President and about 77% for Governor.

In North Carolina’s 2024 general election, unaffiliated voters made up 35.6% of everyone who voted. On the same ballot, they appear to have given the Democratic candidate about 47% of the two-party vote for President and about 77% for Governor, which Stein won for the Democrats. That is a gap of about 30 points between two races decided by the same voters on the same day.

What it means

Unaffiliated is not a lean. It is a group that moves with the candidates. In the presidential race the estimate has unaffiliated voters at 45.2% Democratic and 50.6% Republican, a slight Republican tilt. In the governor’s race it has them at 65.7% Democratic and 20.2% Republican.

Surveys show the same direction and a similar size. In the 2024 exit poll, North Carolina independents were 47% Democratic for President and 58% for Governor. In AP VoteCast they were 51% and 64%.

The same pattern appears elsewhere. In Iowa in 2022, the same method puts unaffiliated voters at about 40% Democratic for Senate and about 22% for Governor. Across the races we checked, unaffiliated voters moved by 20 to 30 points between contests on one ballot.

For reading the early vote, this matters. An early electorate that is one-third unaffiliated, as North Carolina’s is this year, cannot be turned into a vote count by assuming independents split evenly, or the way they split last time.

What it doesn’t mean

The numbers are estimates, not measurements. No official record shows how any group voted. We assumed every registered Democrat voted for the Democrat and every registered Republican for the Republican, then attributed the rest of each candidate’s vote to unaffiliated voters.

That assumption overstates how far unaffiliated voters swing. Any registered Republicans who voted for Stein are counted here as unaffiliated Stein voters. In the 2022 Senate race, allowing for 10% of registered Democrats crossing over changes the unaffiliated split from about 42–51 to about 52–46. The real gap between President and Governor is probably smaller than 30 points, though the surveys still put it at 11 to 13 points.

It does not mean unaffiliated voters are Democrats in disguise. In top-of-ticket races in North Carolina and Iowa in 2024, the estimate has them leaning slightly Republican.

How we checked

Turnout by party comes from the North Carolina State Board of Elections voter history file, which records the party each voter was registered with at the time of voting. In 2024, 31.5% of North Carolina voters were registered Democrats, 32.9% Republicans and 35.6% unaffiliated or minor-party.

Vote totals come from the certified 2024 results. For each race, the unaffiliated vote for a candidate is that candidate’s total minus the votes of their own party’s registered voters. Dividing by the number of unaffiliated voters gives the split.

We compared the results with exit poll and AP VoteCast figures for self-described independents. The surveys use self-identification, not registration, so they measure a related but different group.

County maps undercount Democratic early voters by about 16 points

Sun, 18 Oct 2026 00:00:00 GMT

Weighting each county’s early votes by its 2024 result understated the Democratic margin by 7 to 41 points in 8 states with known party.

A quick way to guess the party of early voters in a state without party data is to weight each county’s early votes by how that county voted in 2024. We tested it in 8 states where the party of early voters is known. As of Oct 22, 2026, the county method understated the Democratic margin in every one of them, by 7 to 41 points, with a median of about 16.

By state, the understatement in the D−R margin was: Arizona −7.3, Nevada −7.5, Florida −14.1, Kansas −15.6, Maine −17.4, North Carolina −19.9, Iowa −31.0 and Pennsylvania −41.0.

What it means

Geography misses most of what makes early voters different. Most of the Democratic tilt among early voters happens within counties: in a given county, the people who vote early are more Democratic than the county as a whole. A county average cannot see that, because it treats every early voter in a county as if they voted like the county overall.

Measured another way, geography explains only a small part of the gap. Taking each state’s 2024 result as the baseline, the county method captures roughly 0 to 40% of the Democratic tilt that registration shows among early voters, and in most states well under a fifth. The rest of the tilt happens inside counties.

The bias is smallest where early voting is old and universal. In Arizona and Nevada, where most voters vote early, the county method was off by about 7 points. It was largest where the early electorate is mail-heavy and skewed toward Democrats, as in Pennsylvania and Iowa.

For the Senate races without party data, this sets a floor. In Michigan, Minnesota and South Carolina, a county estimate is a lower bound on how Democratic the early vote is, not an estimate of it.

What it doesn’t mean

The registration benchmark is not the truth about votes. Registration is not vote choice, and in states with many long-registered Democrats who vote Republican, registration overstates the Democratic margin. Pennsylvania’s −41 is partly that: its early voters are 67% registered Democrats.

The two measures answer different questions. The county method estimates vote preference and is almost two-party, with about 1 to 2% Other. Registration splits voters into Democrats, Republicans and unaffiliated, and unaffiliated voters were 9% to 34% of early voters in these states. That is why we compare margins and two-party shares, not raw Democratic percentages. On a two-party basis the bias is −4 to −23 points, median about −11.

It does not mean county maps are useless. They show where the early vote is coming from. They cannot show who within each county is voting.

How we checked

For each county (each town in Maine), we took accepted 2026 early ballots, by mail and in person, and multiplied them by that area’s 2024 presidential shares. Summing across a state gives the county-method estimate. Areas with no 2024 match took the statewide shares.

Early-vote counts come from state files: the North Carolina State Board of Elections file, Maine’s absentee voter file of Oct 20, the Iowa Secretary of State’s daily statistics, and the UF Election Lab county files, which we checked against official sources where we could. 2024 results are county presidential returns; Maine uses town-level results from the state tabulation.

We compared each estimate with the registered party of the same early voters, as recorded by the state, and report the difference in the D−R margin.

Only four places hand out voter files for free

Sat, 17 Oct 2026 00:00:00 GMT

North Carolina, Ohio, Clark County NV and some Texas counties. Elsewhere you pay, file a form, open an account or prove who you are.

A voter file is a state’s list of registered voters, usually with a record of which elections each person voted in. We checked how to get one in all 50 states and DC. Only four places let anyone download voter-level files for free, with no account, form or payment: North Carolina, Ohio, Clark County in Nevada, and Texas, through the state’s early-vote rosters and some county files. Everywhere else you pay, file a request, open an account or show that you qualify.

What it means

Most independent election analysis depends on a handful of open states. North Carolina publishes its full statewide registration file and full vote history for anyone to download. Ohio publishes its statewide voter file with every voter’s primary history. Clark County, Nevada posts its voter list with each voter’s 20 most recent elections, plus daily early-vote files.

Texas is partly open. The Secretary of State’s Early Voting Roster has a public interface with no login that lists who has voted early, by county and day, but with no party. The statewide list with vote history needs a public information request, a signed affidavit that it will not be used commercially, and a fee. Some counties post past rosters: Tarrant for 2022 and 2024 primaries and recent generals, Hidalgo for 2022 and 2024, and Dallas and Bexar for 2024.

That is why so many of our posts use North Carolina, Ohio and Texas. They are where we can check the numbers ourselves.

What it doesn’t mean

It does not mean other states are closed. Many sell files cheaply. Pennsylvania’s full voter export is $20 for any county or all of them. Alaska’s list sells for about $21, according to news reports. Colorado charges $50 per report. Arkansas offers a free history file, but only through a request form.

Some are expensive or restricted:

Iowa charges $1,500 a year for the statewide list with updates and $450 a year for the absentee list, and limits use to political, research or official purposes.
South Carolina sells only to registered South Carolina voters, up to $2,500.
Minnesota sells only to registered Minnesota voters, $46 statewide.
Illinois releases its statewide file only to registered political committees.
Michigan requires a public records request.

“Free” does not mean frictionless. Ohio’s files sit behind a browser check that blocks scripted downloads, though a normal browser passes it. Some Texas county sites do the same.

It does not mean the data is free to use for anything. Several states limit commercial use of voter lists, and some make it a crime. Pennsylvania’s terms bar commercial use; Iowa calls commercial use a misdemeanor; Illinois a felony.

How we checked

Between Oct 22 and 23, 2026, we checked each state’s election office for a voter file, a vote-history file and an absentee or early-vote file. For each we recorded whether it was free to download now, what it costs, who may buy it and what it contains, with links to the office’s own pages.

We did not solve CAPTCHAs, create accounts, submit request forms or make purchases. A state counts as free only if a file could be downloaded directly. Prices are those published by the states, except Alaska’s, which comes from news reports.

New voters pulled counties toward Biden in 2020, and did nothing in 2024

Fri, 16 Oct 2026 00:00:00 GMT

County shares of new voters tracked the presidential swing at 0.42 in North Carolina and 0.50 in Ohio in 2020. In 2024 the link was about zero.

In 2020, counties with a larger share of new voters swung further toward Joe Biden. The county-level correlation between the share of new voters and the presidential swing was 0.42 across North Carolina’s 100 counties and 0.50 across Ohio’s 88. For new unaffiliated voters in North Carolina it was 0.71. In 2024 the relationship disappeared: −0.10 in North Carolina and −0.03 in Ohio.

A new voter here is someone who voted in that general election with no earlier general-election vote on their state record.

What it means

New voters mattered to the 2020 swing in a way they did not in 2024. In 2020, where more people were voting for the first time on record, the county moved further toward the Democratic candidate. In 2024, the counties with many new voters swung about as much as the counties with few.

The 2020 effect was strongest among new unaffiliated voters in North Carolina. Unaffiliated voters were the largest group of new voters that year: 487,207 of 1,246,188, against 380,488 Democrats and 378,493 Republicans.

New voters were also a large part of both electorates. In North Carolina they were 22.5% of 2020 voters and 16.1% of 2024 voters. In Ohio, 11.0% and 7.7%.

For 2026, this is a warning against assuming that a surge of new voters favours one side. It did in 2020. It did not in 2024.

What it doesn’t mean

It does not show how new voters voted. A correlation across counties says that places with more new voters swung more. It cannot say whether the new voters themselves caused the swing, or whether something else about those counties did. No official record shows individual votes.

It does not extend to midterms. We measured presidential swings only. New voters are a much smaller share of midterm electorates: 7.5% of North Carolina’s 2022 voters, and 3.1% of Ohio’s.

“New” is limited by the records. North Carolina’s history file starts in November 2016, so a 2020 “new” voter may have voted before 2016. Ohio’s file holds only voters registered today, so people who have since moved, died or been removed are missing, and the gap grows the further back the year is.

Party means different things in the two states. North Carolina records party registration at the time of voting. Ohio has no registration, so party there is the last primary ballot, and most voters have none.

How we checked

For North Carolina we used the State Board of Elections statewide voter history and registration files, both dated Oct 18, 2026. Each voter was matched by ID across general elections from 2016 to 2024 and counted by county, party and whether they had voted in an earlier general.

For Ohio we used the Secretary of State’s statewide voter file, last updated Oct 17, 2026: 8,015,003 voter records with general-election columns back to 2000. A voter is new if no earlier general-election column is filled.

For each state and year we correlated each county’s share of new voters with the county’s change in the Democratic presidential margin from the previous election. Only county counts were kept; no voter-level data was stored.

Bellwether counties barely exist in midterms

Thu, 15 Oct 2026 00:00:00 GMT

Only 13–16% of counties moved with the nation, within 4 points, in most House midterms. A perfect record adds 0–4 points of edge.

A bellwether county is one that moves the way the nation moves. In U.S. House midterms, there are very few of them. Across five pairs of midterms from 2002 to 2022, only 13% to 16% of counties swung in the same direction as the national House vote and landed within 4 points of it, except in 2022, when 33% to 38% did. Counties with a perfect record so far were barely more likely to pass the next time: 0 to 4 points, where the groups are large enough to measure.

What it means

Direction is easy in a wave year; size is hard. In 2006, 2010 and 2022, between 80% and 91% of counties moved the same way as the nation. Far fewer moved by about the same amount. County House swings depend on candidates, incumbents and open seats, which the national vote averages away.

Midterm counties are noisier than presidential ones. In presidential elections, 19% to 66% of counties pass the same test each cycle. In midterms, the pass rate is 13% to 16% in every cycle except 2022.

A past record predicts little. In the 2010–2022 panel, which has the largest groups, counties that had passed every swing so far did 0.6 and 3.5 points better on the next one than counties that had already failed. In the contested-only sample the edge was 0.3 and 1.4 points.

The handful of perfect counties look like luck. Two counties passed all five swings from 2002 to 2022: Brevard, Florida and Spokane, Washington. That is more than chance would give, but in the larger 2010–2022 panel, 18 counties passed every swing against 15.7 expected by chance.

What it doesn’t mean

It does not mean counties tell you nothing. They are useful for seeing where a swing is coming from. 2018 is the example: the nation moved 14.3 points toward Democrats, but only 59% of counties did. Weighted by votes, 83% moved toward Democrats. The 2018 wave came from large and suburban counties.

It does not apply to every office. Governor results track their own state more closely: in most pairs of consecutive midterms, 77% to 85% of counties moved the same way as their state, and a perfect record so far gave a 4 to 14 point edge, slightly better than House races.

It does not say 2026 will be like 2022. The 2022 pass rate was high because 91% of counties moved toward Republicans, many by similar amounts. That is one cycle, not a trend.

How we checked

We built county results for U.S. House, Senate and Governor for every midterm from 2002 to 2022 from free precinct and county returns. Our national House vote is within 0.3 points of the published figures for 2006 to 2022. Coverage is thin in 2002 (56% of counties), so some tests start in 2006 or 2010.

A county’s House margin sums every district that touches it. The swing test asks whether a county moved in the same direction as the national House swing and within 4 points of it. We ran it on all counties and on a contested-only sample that drops counties with uncontested districts or suspect totals.

For the survivorship test we compared counties with a perfect record so far against those that had already failed, and checked counts against a shuffle of 2,000 random reorderings.