Truth Is Freedom

THE 15% BAND

Look at this chart.
59 races. Senate and governor.
States ordered from reddest to bluest in 2024.

The solid lines are 2024.
The dashed lines are 2026.
The shaded bands are what moved.

One formula fits all of it:
Republican 2026 = 0.85 × Trump 2024
Democrat 2026 = Harris 2024 + 0.15 × Trump 2024

For every 1,000 Trump points, 850 stayed Republican. 150 landed on the Democratic side.

The numbers:
Senate: Republicans kept ~85% of Trump’s share
Governor: ~86%

Variation explained: 88%

What makes it strange:
It didn’t matter how red the state was. Wyoming, Ohio, Illinois, Massachusetts: same 15% (p = 0.69). The Democratic gain was sized by Trump’s vote, not Harris’s. The Harris part is zero.

The 0.85 beat both standard election-swing models in every set of races.
Drop any single state and the multiplier stays between 0.846 and 0.853.

Why seats flipped:
At 0.85, any state where Trump got under about 58.5% falls below 50%. Texas, Kansas, Iowa, Ohio, South Carolina, North Carolina: all under that line. All flipped.

What this means:
Midterms are messy. This one isn’t. Two sets of races, different candidates, different campaigns, and the same 15% decrease almost everywhere.

A pattern this even demands answers at the ballot level: paper ballot hand counts, poll-book check-ins against votes counted, and voter files showing who actually turned out.

The numbers raise the question.
The ballots can answer it.

A single formula, fixed before any votes were cast, lined up with the results in county after county so closely that the outcome was basically known in advance.

THE 15% BAND, PART 2

I previously shared that the variance above 0.85 meant proportionately more Republicans turned out than in 2024. Below, proportionately more Democrats turned out.

Over the weekend, I took the next step: residual analysis. In the six states where Republicans lost a Senate seat, I looked at the Republican side and the Democratic side of that variance in the 692 counties.

Here's what I found:
On the Republican side, the redder the county, the further above 0.85 it landed. On the Democratic side, the more Hispanic the county, the further below 0.85 it landed.

Adding those two variables to the foundation predicts 97% of these counties within five points.

Different states.
Different races.
Different candidates.
Nearly 700 different counties.
And one simple formula predicts 97% of them.

The 85% foundation holds.
The counties confirm it.

Normally in a midterm, the national mood pushes everyone in one direction, but individual candidates can swim against it. A popular incumbent or a hometown favorite can run well ahead of their party.

In 2026, nobody did. No Republican Senate candidate ran meaningfully ahead of Trump’s share. Strong candidates, weak candidates, incumbents, open seats, red states, blue states, they all landed in a narrow band at or below him.

Whatever was driving the vote was national, and strong enough to wash out the usual local differences.

Updated sign test.

I added four more midterms (1994–2006), so the baseline now covers every midterm since 1994. I also added a beta-binomial test.

How unusual was the 2026 near-sweep?
Binomial: 1 in 5,510
Beta-binomial: 1 in 1,670