NetChoice Florida numbers need a little seasoning

Every poll landing in political inboxes in Florida gets the same first question (at least from anyone who knows what they’re doing): “How did they build this thing?”

The NetChoice survey from Echelon Insights this month looks at a wide range of topics, including antitrust views, what matters to consumers, and several 2026 Florida races. The polls show Byron Donalds leading by 6 points over Donalds and by 4 points over Jerry Demings. These results are similar to other recent public polls.

But before anyone makes big decisions based on these results, it’s important to take a closer look at how the survey was done.

Three things stand out.

Problem One: That party model may slightly favor the democrats

The survey uses a turnout model that gives Florida a Republican advantage of about +10. In our view, this might slightly underestimate the possible GOP edge. Right now, national trends are favoring Democrats.

If not for those national trends, we would have estimated a +14 GOP turnout model. But since this is Florida, GOP candidates will have more money to compete, which could balance out those national factors. Also, with GOP registration still growing, we think a +12 GOP model is more accurate.

This would probably change the results by a point or two. It’s not a huge difference, but it’s still worth mentioning.

Problem Two: Self-reported party ID is a known unreliable narrator

The survey asks people to indicate which party they belong to, rather than checking this against the official Florida voter file. This difference is important but often overlooked. In Florida, party registration is public information. Echelon had access to the voter file and used it for screening and weighting, but the final sample still appears to depend on self-identification.

The problem with self-ID is simple: people lie, misremember, or offer an aspirational response.

A registered Democrat who voted Republican for three cycles may call himself an independent — or even a Republican — when asked.

A registered Republican who voted for Kamala Harris last November — OK, it’s only theoretical — might do the same, but in the opposite direction. Checking respondents against the voter file, instead of just asking about party affiliation, removes this uncertainty and gives you a model based on facts you can verify.

When you have the file, use the file.

Editor’s note: We can only go on the published methodology. If the pollster verified the respondent in the voter file, let us know, and we will correct this!

Problem Three: No live outbound calling

This survey was done completely online and through text-to-web. That means people either joined a web panel or got a text message with a link to the survey.

There is no outbound live calling in this methodology. That matters.

Live telephone interviews, especially live calls to cell phones, remain the gold standard for public opinion research. They involve real human interaction, which digital methods can’t match, and they help ensure randomness, which is important to avoid bias.

Online panels and text-to-web surveys can introduce self-selection bias at every step, from who clicks to who completes or skips the survey. Panels usually come from groups of people who have agreed to take surveys for rewards, which introduces a bias that live calling does not have.

Matching the voter file after the fact helps a little, but it can’t fully address who decided to respond in the first place.

None of this makes the NetChoice survey worthless.

The sample size is solid, the field dates are reasonable, the results are consistent with other published polls, and Echelon Insights is a credible shop.

However, because the survey relies on self-reported party ID instead of checking the voter file and doesn’t use live outbound calling, these Florida results should be viewed with some caution.