Every turnout chart you've seen is slightly wrong. Here's how we fixed it.
California's 2024 turnout wasn't 53%. It was 61%. The correction explains a lot.
The wrong denominator
Open any political data site — 270toWin, Ballotpedia, your state's secretary of
state — and look up California's 2024 presidential turnout. You'll get something
around 53%. That number is produced by dividing votes cast by California's
voting-age population (VAP): every resident 18 and older, from the 2020 Census.
But California has 3.7 million non-citizens age 18 and older. They can't vote. If
you include them in the denominator, you're measuring turnout against people who
were never eligible to turn out. The honest question is: of California's
citizen voting-age population, how many voted?
That's a 7.5-point correction. California goes from looking disengaged to looking
about average. This isn't a small methodological quibble — it changes the story
you tell about American democracy.
CVAP: citizen voting-age population
The Census Bureau publishes a Citizen Voting-Age Population Special Tabulation
every year. It's the right denominator for turnout. It draws from the American
Community Survey and breaks out citizenship by age for every state, county,
congressional district, state legislative district, and city in America.
Until last week, Akashic Edge was using 2020 VAP like everyone else. We just
loaded CVAP 2020–2024 into the product. Now every turnout number on the site —
on the cartographer, the historian, every county and state page — uses CVAP as
the default denominator. The VAP number stays available via the picker in the
corner of the turnout panel, but it's no longer the headline.
Some of the corrections are dramatic:
Miami's old 39.7% turnout made it look catastrophically disengaged. Actual CVAP
turnout in Miami is 56.0% — 16 percentage points higher. The old number wasn't
wrong because Miamians didn't show up; it was wrong because the denominator
counted Miami's large non-citizen population as if they could have voted.
San Jose jumps from 46.6% to 60.6%. San Francisco from 53% to 65%. Los Angeles
from 44% to 56%. Every high-immigrant geography in America was being systemically
understated, which in turn has fed a decade of "Latinos don't vote" narratives
built on a denominator that included non-Latinos and non-citizens.
VEP: voting-eligible population
Florida's 2024 turnout looks different from California's for a different reason.
CVAP gives you 67.3% — that's already a 5-point correction from the 62% VAP
number — but Florida has another subtraction that California doesn't:
felony disenfranchisement.
Before you measure turnout in Florida, you should subtract the 1.15 million
Florida citizens who are legally barred from voting due to a felony conviction.
That's 7.1% of Florida's CVAP. Do the subtraction, and Florida's 2024 turnout
jumps to 72.5% — among the highest in America, not the mediocre middle-pack
number you'd get from VAP.
This is the VEP — Voting Eligible Population — denominator that Michael McDonald
at the University of Florida has been computing for decades. Akashic Edge now
ships it at the state level for 50 states (Mississippi excluded pending The
Sentencing Project's 2024 revision). VEP is the default denominator on state
pages. Sub-state VEP is a future phase.
The non-voter pool
Once you have CVAP, a new question opens up: of the eligible citizens who
didn't vote, who are they? And what partisan lean does that pool have?
Texas has 8.4 million eligible non-voters in 2024. That's the largest
non-voter pool in any state. Who they are matters:
Hispanic citizens are 36.3% of Texas's non-voter pool — the largest single
demographic bloc. Historically, that fact generated a reliable story: "if only
they voted, Texas would be blue." The Texas Republican Party would be in ruins.
Look at the partisan lean of the Texas non-voter pool in Akashic's Turnout Gap
panel and you'll see something different:
Near wash. Not D+15 the way Texas voters lean (actually they lean R; voters
were R+14 in 2024). Non-voters in 2024 aren't a silver bullet for Democrats
anymore. The "drop-off" electorate shifted sharply more Republican in 2024,
most dramatically among Hispanic and Black men.
We didn't apply 2024 exit-poll voter lean to non-voters — that's a common
methodological error. We apply 2024 exit-poll voter lean plus an adjustment
for the published non-voter R-shift. Sources: Pew 2024 non-voter analysis, AP
VoteCast non-voter subsample, Harry Enten and Nate Cohn post-election
breakdowns. The adjustment is biggest for Hispanic (−16 pp) and Black (−18 pp)
because those are the groups where the 2024 drop-off electorate diverged most
from the engaged voter pool.
California's non-voter pool in 2024: R +1.4. Not D +13. Mobilizing California
non-voters under the 2024 lean structure would slightly shift the state
right, not left. Same pattern in New York, Illinois, and Los Angeles County —
deep-blue geographies where non-voters are less partisan than active voters.
2024 is not forever
These non-voter numbers are a 2024 snapshot. Post-2024 special elections —
Wisconsin Supreme Court, federal House specials, gubernatorials — show
Democrats winning back ground with low-propensity voters. The 2026 lean
constants will differ. The model has the year hardcoded in every constant
name precisely to force an annual refresh.
Every denominator decision is documented at /methodology/turnout.
Every turnout number is also available via the public API at
/api/v1/turnout?geo_id=state:2020:48. Prophet tier ($99/mo) and above.
The short version: if you've been reading turnout numbers off any political data
site in the last decade, they were computed against a denominator that included
millions of people who couldn't vote. The correction isn't a small one. And the
non-voter pool you inherit after the correction doesn't lean the way decades of
commentary assumed.