New UNF Poll Shows Tight Presidential Race In Florida

Oct 20, 2020

A Public Opinion Research Lab (PORL) at the University of North Florida poll of likely Florida voters released Tuesday gives Democratic presidential nominee Joe Biden a razor-thin lead with 48% saying they intend to Biden and 47% saying they intend to vote for President Donald Trump.

UNF says the poll has plus or minus 3.3% margin of error.

“While some polls have shown Biden with a big lead in Florida and other key states, we made an effort to capture hard-to-reach voters, and our results suggest that it might be a long night on November 3rd,” said Michael Binder, PORL faculty director and associate professor of political science, in an email to WJCT News.

Forty-three percent of the 863 registered voters interviewed for the poll said they would vote by mail, with 29% planning to vote early in person, and 26% said they planned to vote on Election Day. Twenty-five percent said they had already voted by mail.

Related: Election 2020 News, Candidates, Precinct Finders And More

In Duval County, almost 17% of eligible voters had cast their ballots by Tuesday morning, with almost 112,000 votes cast by 9 a.m., according to the Duval County Supervisor of Elections.

So far in Duval, Democrats are voting in significantly larger numbers; 60,300 have cast votes, compared to 34,938 Republicans and 15,302 with no party affiliation.

“This is Florida, and elections are never easy here. I expect this race to come down to the wire,” Binder said.

When asked whether they approve of the job President Trump is doing, 47% approved while 49% disapproved.

Tip: Click on the image to enlarge the text, or you can also use your browser's magnification feature to zoom in to a portion of the text. If viewing this on a touch screen, use your device's pinch-to-zoom feature to enlarge the text.
Credit University of North Florida

 

Credit University of North Florida

Poll Methodology

The UNF PORL Florida Statewide Poll is comprised of 863 registered, likely Florida voters and was conducted Monday, Oct. 12, through Saturday, Oct. 16. The sample frame was comprised of registered, likely Florida voters 18 years of age or older. "Likely voters" were determined through vote history and a self-identified likelihood to vote.

The landline and cell phone numbers used for this survey were sourced from the September 2020 update of the Florida voter file. Respondents were contacted by live callers via the telephone from 3 p.m. to 9 p.m. daily, with up to two callbacks attempted. PORL employees conducted interviews in both English and Spanish.

Data collection took place, both at the PORL facility with its 27-station Computer Assisted Telephone Interviewing (CATI) system, as well as remotely, using specialized CATI software. The margin of sampling error for the total sample is +/- 3.3 percentage points.

Through hand dialing, an interviewer upon reaching the individual as specified in the voter file asked that respondent to participate, regardless of landline telephone or cellphone. Data were then weighted by partisan registration, age, race, sex, education, geography, self-identified likelihood to vote and 2016 vote choice. Education weights were created from the Census’ 2019 American Community Survey (ACS) estimate for the percent of college-educated individuals within each strata in the state of Florida, approximately 29% statewide.

Partisan registration, sex, race, and age weights were created from the September update of the Florida Voter File to match the active registered voters in Florida combined to a turnout model similar to 2016. To ensure a representative sample of likely voters, the 10 Florida designated media market areas were stratified into 18 strata. Heavily urban and highly educated counties were separated out from their media markets and given their own stratum.

Quotas were placed on each of these stratified areas to ensure a proportionate number of completed surveys.

All weighted demographic variables were applied using the SPSS version 26 rake weighting function, which will not assign a weight if one of the demographics being weighted on is missing. In this case, respondents missing a response for any of the demographic information, specifically education, were given a weight equal to their weight for the remaining demographics.

There were no statistical adjustments made due to design effects. This study had a 9% response rate.

The American Association for Public Opinion Research (AAPOR) Response Rate 3 (RR3) calculation was used, which consists of an estimate of what proportion of cases of unknown eligibility are truly eligible. The survey was directed by Dr. Michael Binder, UNF associate professor of political science.

Bill Bortzfield can be reached at bbortzfield@wjct.org, 904-358-6349 or on Twitter at @BortzInJax.