Pollsters likely miscalculated who was going to turn out.
Going into the Florida governor’s primaries on Wednesday, top-line polls had the eventual Democratic winner Andrew Gillum in fourth place, with most showing him getting just 12 percent of voters’ support on average. Gillum — the state’s first African-American gubernatorial nominee — ended up pulling off a major upset and taking the nomination with more than 34 percent of the vote.
The unexpected outcome led to many observers wondering how exactly the polls — which consistently favored a victory by establishment candidate Gwen Graham — could have gotten it so wrong, again. Polling experts say there are likely a few factors at play, including the heightened volatility of polling in primary elections, when it can be more challenging to identify likely voters.
“Only a small percentage of the electorate actually vote and that electorate is not stable from election to election,” said Chris Jackson, a vice president at Ipsos, a market research firm. Because of this, “it’s tougher sometimes to get a representative sample [during primaries],” Quinnipiac’s Peter Brown said. The sample of people polled may not have fully captured what the ultimate electorate ended up looking like.
Young voters and African-American voters — who ended up turning out heavily for Gillum — were potentially among the groups that were underrepresented in these polls, Celinda Lake, a Democratic pollster, said. Undecided voters, who accounted for more than 20 percent of the folks who were surveyed, on average, and whose preferences were likely masked in earlier surveys, appeared to go heavily for Gillum on Election Day as well, according to Florida-based political consultant Doug Kaplan.
Here’s what eight experts had to say about the polling disconnect in the Tuesday primary.
These responses have been lightly edited for length and clarity.
African-American voters and younger voters were among the groups to give Gillum a boost. They may have been underestimated by the polls.
Celinda Lake, Lake Research Partners, president
Polls missed youth turnout, and that happened in other races like Alexandria Ocasio-Cortez’s. The campaign also targeted campuses that just got back [in session]. Polls missed the enthusiasm and solidification of the African-American vote and the base Andrew had there.
Doug Kaplan, Gravis Marketing, president
The race was very close. The undecided voters on Election Day broke toward Gillum. [Philip] Levine and [Jeff] Greene collapsed on Election Day, along with an increased turnout. [Compared to the 2016 primary], the GOP saw increased voter turnout by 13.5 percent; Democrats saw increased voter turnout by 35 percent.
Jay Leve, SurveyUSA, president
SurveyUSA’s poll had Gillum leading among African-American voters, leading among urban voters, leading in northwest Florida, and tied for the lead among voters age 35 to 49. Our published analysis — written two weeks before the primary and well before late liberal cash poured to Gillum — called the contest a “free-for-all,” which it was. Ours was the only poll that did not show Gwen Graham as a clear frontrunner.
That said: there were a total of nine polls released by five different pollsters in the four months leading up to yesterday’s primary, and eight of the nine polls had Gillum in fourth place; 1 poll had Gillum tied for third. No poll gave Gillum more than 16 percent of the vote — less than half the 34 percent he won with.
Gillum pulled together a broad coalition of liberals and progressives, many of whom were white and Hispanic. He consolidated young voters from one end of the state to the other. For a Tallahassee mayor to win Broward County by more than 20,000 votes over a Miami Beach mayor speaks to the depth and breadth of his primary support.
Brandon Finnigan, Decision Desk HQ, director
(Vox live results are provided by DecisionDesk.)
While his better-financed opponents were roaming about the state, Gillum resonated with African-American voters as the first potential black Governor of Florida. He won every county with a significant number of voters and an African-American population that exceeded the national average.
In counties with very large African-American populations, he absolutely destroyed Graham. While all of the Democratic nominees made their rounds with African Americans, the big ones were basically fighting over white and Latino voters, leaving Gillum to dominate among African Americans and pull off the surprise win.
In a state where the Democratic Party is heavily dependent on nonwhite voters, a candidate that connects strongly with a minority bloc can win in a crowded field that spreads its energy across all blocs more evenly. This isn’t to say every Democratic voter that pulled for Gillum was black, but most black voters did so.
Primary polling is volatile and proper methodology is crucial
Chris Jackson, Ipsos Public Affairs, vice president
In the Florida Democratic primary, about 1.5 million votes were cast or about 15 percent of the total Florida population. For a poll to accurately identify the correct 15 percent of the population is a significant undertaking.
The public polls that were conducted in the Florida primary either had small samples — less than 500 interviews — or were conducted by computer interviewing, or both. These methods of polling, while quite affordable, can really struggle with identifying small populations. These two points appear to have combined in Florida with polls understating the support for Gillum.
Jay Leve, SurveyUSA, president
Primaries with more than two candidates on the ballot — there were seven candidates in Florida — can be volatile, with complex dynamics that are too subtle for pollsters to pick up. In Florida, all pollsters missed the fact that liberals who said in May, June, and July that they were flirting with Graham or Levine were in fact just waiting for the real thing to come along. In late August, it became clear that the real progressive was Gillum, and that’s who voters went home with.
Patrick Murray, Monmouth University polling, director
Turnout for both parties was significantly higher than in prior Florida gubernatorial primaries, with the “populist” candidate doing better than projected in both contests. It seems highly likely that the 2018 primary electorate included a large number of voters with a history of only turning out in general elections.
Only Mason-Dixon used a full telephone frame (live calls to landlines and cell phones) drawn from a voter list. The others used an online panel for all interviews (SurveyUSA), or a hybrid of an online panel to “replace” cell phone calls and interactive voice response (IVR) calls to landlines either drawn from a voter list (FAU) or randomly dialed from all phone exchanges in Florida (Gravis).
Obviously, every poll missed the performance of Gillum, and to a lesser extent DeSantis, regardless of their methodological approaches.
Larry Sabato, University of Virginia’s Center for Politics, director
Because there is no party cue to nudge voters a certain way [in a primary] — all candidates have a D next to their name — people depend on other cues to push them to the polls. These cues can kick in late, and not just because the primary is held in brutally hot late August. Voters mainly do not feel urgency to solidify their primary choice, again because in November they’ll vote for any Democrat nominated.
No doubt the [Sen. Bernie] Sanders endorsement did help Gillum, and it came late. Gillum didn’t air many TV ads compared to the others, so primary voters may have learned what they needed to know about him only in the last couple of weeks.
This was a big candidate field, relatively, and with a lot of moving parts in a primary, there can be fluctuation right up to Election Day.