UNF’s Florida poll is the real deal — with just a half-grain to spare

Reading the methodology of this poll was truly a breath of fresh air.

Every cycle produces a handful of publicly released polls worth trusting and a landfill of polls worth the salt shaker. The new University of North Florida Public Opinion Research Lab (PORL) survey of 848 likely Florida Midterm voters belongs firmly in the first pile.

This is a shop that shows its work, with a full methodology, transparent weighting, no self-identification (they use actual voter history from the voter file and confirm respondents!), and a well-balanced sample (well, except party, but we will address that further on).

As for the methodology and transparency, this is where PORL earns its keep: a mixed-mode design (phone and text-to-web), stratification across 18 geographic strata with Miami-Dade carved out on its own, and a two-step raking weight on education and then party, age, race, sex and geography, benchmarked to the actual 2022 electorate.

Way to go!

This is graduate-seminar-grade transparency, and it shows in a lot of the internal consistency — the Donald Trump 2024 vote recall (50%-40%) lines up almost exactly with the certified result, which is not a bad gut check.

So, when this pollster finds Byron Donalds leading David Jolly 46%-41% in the Governor’s race, and Ashley Moody leading both potential Democratic Senate opponents by high single digits to low double digits (10 over Alex Vindman, 8 over Angie Nixon), save the salt for your morning eggs.

Of course, nothing about these numbers should shock anyone who’s been watching Florida drift red over the past three cycles or has reviewed credible polls, but the margins are notably tighter than 2022 or 2024 — which tracks with the read from PORL faculty director Sean Freeder that the party in the White House usually suffers in Midterms. No salt needed here; these are plausible, internally consistent numbers from a poll built the right way.

A half-grain: The raw sample skews red before weighting.

I know! I know! Post-sampling weighting can widen your margin of error, and this poll needed a good dose of that, as the original sample had a +20 GOP model. Even GOP-leaning consultants don’t see that as a realistic turnout balance; it’s more likely closer to +12. For this writer, post-call weighting is not optimal, but it doesn’t harm the findings or impact the analysis, so I’ll offer just a half-grain here.

To further clarify the half-grain: Any time your raw sample needs a real lean corrected out of it, there’s a little more machinery standing between the interviews and the topline than there would be in a sample that arrived (or was derived) pre-balanced. It’s not a reason to distrust the numbers. It’s a reason to note that the weighting does have a modest effect.

Bottom line: solid shop, solid numbers, tight-but-plausible horse races, and one honest methodological note-to-self about how much lifting the party weight is doing on the raw sample.