A Florida polling firm has admitted that it moved some voters into James Fishback’s column before publishing a poll that showed the Republican candidate suddenly breathing down Byron Donalds’ neck.

Then real voters showed up.

Fishback did not finish a close second. He took just 10.48 percent of the statewide Republican primary vote in Florida Election Watch’s first set of unofficial returns.

Florida Election Watch showed Donalds winning with 47.79 percent after all 67 counties reported into the statewide system. Lieutenant Governor Jay Collins finished second with 25.16 percent in the 11-candidate Republican field.

Collins’ share was more than double Fishback’s, and Donalds beat Fishback by more than 37 percentage points. The state labels the August 18 figures its first set of unofficial returns under Florida law, meaning final certification still follows.

ADVERTISEMENT

Donalds held a lead of more than 384,000 votes over Collins and more than 633,000 over Fishback in the state’s first set of unofficial returns.

Those numbers make the pre-election polling controversy impossible to shrug off as harmless campaign chatter.

The returns also captured a far larger electorate than one survey sample. Nearly 1.7 million Republican votes were recorded across the 11-candidate field, compared with the 476 registered Republicans contacted for the poll.

Paul Renner finished fourth with 8.55 percent, behind Fishback. No candidate outside the top three reached 11 percent, underscoring how badly the final poll misplaced the order and scale of the field.

Every one of Florida’s 67 counties had reported in that statewide snapshot. The remaining unofficial status could alter raw totals, but it could not turn a 37-point gap into the near-tie TPSI had published five days earlier.

Donalds carried a clear statewide mandate while Fishback landed closer to Renner than to either of the top two finishers. The ballot count left no plausible turnout model that could rescue the survey’s central picture of the race.

The People’s Sentiment Initiative, or TPSI, released its final survey on August 13. It put Donalds at 34 percent, Fishback at 29.4 percent and Collins at 20.7 percent.

The firm described Donalds’ 4.6-point edge as sitting at the boundary of its 4.5-point margin of error. It presented the race as competitive and projected roughly 1.6 million Republican ballots.

TPSI said its “Meridian Model” allocated most undecided voters according to their stated preferences and affinities. Its published methodology said results were weighted by age, gender, race, education and region.

ADVERTISEMENT

The survey also declared that no campaign or outside organization had sponsored, commissioned or paid for that research. That claim became far more consequential when the firm’s earlier financial relationship with Fishback’s campaign surfaced.

That poll did more than miss the final margin. The firm later acknowledged that it had identified respondents who selected Collins but whose broader answers seemed to fit another political coalition, then reassigned some of them to Fishback.

Read that again. People said they supported one candidate, and the published result credited some of them to another.

The Washington Examiner reported that TPSI described the practice in a post-election analysis. The firm said it identified voters who selected Collins but whose wider profile aligned with a competing coalition, then “leaned” those respondents toward Fishback.

TPSI called the move a consequential methodological error. It said the underlying survey data had not been fabricated or altered, retired the adjustment and instituted controls intended to prevent a repeat.

FiftyPlusOne reached a harsher judgment. The polling aggregator said TPSI had fabricated vote-choice data, permanently banned future TPSI results and marked existing surveys as partisan.

Patriot Polling was suspended pending a response, while TPSI’s future data was excluded outright. The firms’ older database entries would carry partisan labels.

Pollsters make adjustments all the time. They weight samples for age, geography, party registration and likelihood of voting.

That is a normal attempt to make a sample resemble the electorate. Changing a respondent’s stated candidate preference is something else entirely.

ADVERTISEMENT

The controversy also raised a disclosure problem. Federal campaign-finance records showed Fishback’s campaign paid TPSI $1,575 for polling and survey work ahead of earlier surveys, according to the Examiner.

The campaign also paid $1,500 to the chief executive of Patriot Polling, another firm that published favorable numbers in the race.

A polling organization can work for a campaign. The public simply deserves to know who paid whom before a poll is packaged as an independent snapshot of the race.

FiftyPlusOne, a polling-ratings and aggregation group, responded by suspending TPSI and Patriot Polling. It said the firms had failed to provide adequate funding disclosures and that TPSI had violated standard polling practice.

The admission and the sharp difference between the adjusted survey and the ballots Floridians cast demand more than a quiet correction.

Polls describe campaigns, and they can change them.

A surprise surge can attract donors, earn television coverage, shape debate invitations and pressure voters to abandon candidates portrayed as fading.

A bad number can become a self-fulfilling prophecy. A flattering one can manufacture momentum that never existed.

Fishback’s campaign benefited from a public narrative that he was closing in on the President Trump-backed leader. The official returns exposed that narrative as fantasy.

ADVERTISEMENT

Donalds received 811,101 votes. Fishback received 177,796.

That was not a polling miss around the edges. It was a political mirage.

Fishback is responsible for his campaign, his spending and his own conduct. The evidence reported so far does not establish that he personally ordered TPSI to rewrite individual answers, and honesty requires saying that plainly.

But the polling firm’s admission stands on its own. So do the campaign payments and the missing disclosure that followed those numbers into public view.

Conservative voters should be the first people demanding clean data. We have spent years watching legacy institutions massage language, bury inconvenient facts and tell Americans to distrust their own eyes.

We cannot condemn that behavior when it helps the Left and excuse it when a Republican candidate gets the benefit.

If a pollster thinks a voter chose the “wrong” candidate, the answer still belongs to the voter. Anything else turns measurement into manipulation.

Donalds now advances to a general election with a decisive mandate from Republican voters, and he wasted no time drawing the contrast ahead.

The election settled who Florida Republicans wanted. The aftermath should settle something else too.

Campaigns may spin, and pollsters may model.

No firm should be allowed to take a voter’s answer, replace it with a more convenient one and call the result public opinion.

 

Join The Conversation. Leave a Comment.