Every poll that drops into a Florida political inbox gets the same first question from anyone who knows what they’re doing: How did they build this thing?
Change Research’s new “Deep Dive into Florida” from Freedom Project USA has some genuinely interesting topline findings — Donald Trump underwater with no-party voters (no surprise); Democratic voter motivation up (no surprise); David Jolly leading Byron Donalds (surprise); and Alex Vindman and José Javier Rodríguez both leading (shock).
Before anyone runs to the betting markets, let’s shake some salt on the methodology.
Two things stand out.
Problem 1: “Will you vote?” is a notoriously bad question
The survey splits its 2,070 respondents into those who “say they will likely vote” — and that self-reported screen becomes the foundation for the most consequential numbers in the memo, including the Jolly-Donalds head-to-head.
Here’s the problem: asking people whether they plan to vote is one of the least reliable questions in survey research. It is a textbook example of what social scientists call socially desirable response patterning.
The concept is straightforward. When a pollster asks, “Will you be voting in the upcoming election?”, most respondents know the “correct” answer. Voting is civic virtue. Saying no is like telling your doctor you haven’t exercised in three months — technically honest, socially uncomfortable. So people say yes. Consistently. Across every demographic. In volumes that bear no resemblance to who actually shows up.
And at a time when the President’s numbers are underwater, of course Democrats are more likely to say they will vote. Many of them probably will. But intention isn’t the best yardstick for turnout — the voter file is.
This is not a small distortion. In a typical midterm, 35% to 45% of registered voters don’t cast a ballot. But self-reported “likely voters” in pre-election surveys routinely run 70%, 80%, even higher. The people inflating that number aren’t lying. They intend to vote. They just don’t. Life intervenes. Motivation fades. The enthusiasm they felt when they answered the survey doesn’t survive a Tuesday in November.
Problem 2: The voter file solves this — and it wasn’t fully used
Here’s the good news: Florida has one of the richest voter files in the country, and it contains something far more valuable than self-reported intentions — actual voting history.
Did this person vote in 2022? In 2020? In the last three Primaries? That behavioral record is the gold standard for modeling who will actually turn out. A registered voter with a consistent midterm history is categorically different from one who showed up once in 2020 and hasn’t voted since. The file knows the difference. Self-reported enthusiasm does not.
Change Research does note in its methodology that it recruited some respondents through text messages sent to cell phone numbers on the voter file — so the file was in the room. That, in our view, is a missed opportunity. When your likely voter screen is built on “I think I’ll vote” rather than “this person has voted in four of the last five comparable elections,” your likely voter universe is softer than it looks.
Partisan breakdown
Kudos to Change Research for a +12 GOP model, but two concerns.
Respondents were recruited using “targeted advertisements on Facebook and Instagram, as well as across the web via Facebook’s ad platform” and “text messages sent, via the Switchboard platform, to cell phone numbers listed on the voter file for individuals who qualified for the survey’s sample universe, based on their voter file data.”
The second method is great. The first is the kind of methodology that rewards enthusiasts for raising their hands. The gold standard is the pollster reaching out to the respondent, not the other way around. Ad-based recruitment lets respondents self-select, and self-selection skews.
The bottom line
None of this means the findings are wrong. The issue environment Change Research describes — cost-of-living stress, insurance anxiety, no-party voters hostile to Trump — matches what other data sources show. The directional story is in line with other current polling.
But the specific margins, and especially the enthusiasm gap numbers, deserve scrutiny. A survey that built its likely voter screen on behavioral voter file history rather than self-reported intention would give those findings a far more durable foundation. These methodology choices are likely the reason this poll is the outlier showing a top-of-the-ticket Democratic sweep.
About the pollster
It should come as no surprise that Change Research, while well-regarded, is a left-leaning shop that reps Democrats and progressive causes. This isn’t an attack — but it may shed some light on the findings.
Read this one with a shaker in hand.