Eleven months before Election Day 2024, early polling suggests third-party and independent candidates will roil the presidential race. A much-discussed poll of battleground states from The New York Times found remarkable support for Robert F. Kennedy, Jr. in a three-way race against President Joe Biden and former President Donald Trump, with the scion of the famous political family scooping up votes from roughly a quarter of respondents.
In the Real Clear Politics average of a possible five-way race, the combined total of Kennedy, Cornel West, the leftist intellectual, and Green Party candidate Jill Stein, a perennial source of annoyance for Democrats, hit an unsettling 19 percent. Polls that test a two-way and a five-way race show Trump improving his margin in these expanded fields by about two points.
And these are just the polls, including the announced candidates. We still don’t know if the moneyed centrist operation No Labels will go forward with its announced plans for a bipartisan ticket, and if so, who their candidates will be.
But do these polls mean anything? Can surveying methodology accurately assess a minor candidate’s support, let alone their durability?
Professional pollsters say that the short answer is no. Accurate polling of third-party candidates “is a minefield,” says Joe Lenski, co-founder and executive vice president of Edison Research, which conducts the exit polling for media outlets like CNN and NBC.
Lenski and several other polling experts I spoke with roundly dismissed horse-race polling some 11 months out, with or without third-party candidates. If polls this far from Election Day meant anything, we’d have had a 2008 Election Day major party presidential choice between Hillary Rodham Clinton and Rudy Giuliani instead of Barrack Obama and John McCain.
But multiply the inaccuracy of year-out polling with the mercurial draw of third-party candidates, said Lenski, and the picture gets even murkier. The biggest issue facing pollsters is that voters can be poor predictors of their behavior. Saying you’ll vote for a longshot candidate weeks or months before Election Day is much different than holding a ballot and voting for a candidate you know can’t win.
That conundrum has only grown in the era of increasingly disliked major party candidates. Trump and Rodham Clinton hit record levels of unfavorability in 2016 election polling. The combined support of the two most prominent third-party candidates, Stein and Libertarian Gary Johnson, often cleared 10 percent in mid-election year polling. In one July 2016 poll, just four months before the election, the two combined for 18 percent.
But in the end, Johnson and other third-party candidates collectively came in under 6 percent of the popular vote, which is unsurprising. Third-party candidates routinely fade in the stretch. A June 2000 Gallup survey found Ralph Nader, the consumer advocate and Green Party nominee, and Reform Party nominee Pat Buchanan combining for roughly 8 percent. Still, on Election Day, they only won 3 percent of the national vote, albeit enough to tip Florida, and thus the presidency, to George W. Bush. In 1980, Republican congressman-turned-independent presidential candidate John Anderson scored around 20 percent in Gallup polling for most of the spring and summer but wound up with under 7 percent of the popular vote.
Why do third-party candidates almost always shed their early support? One reason is what looks like support in early polling often isn’t genuine. When voters are underwhelmed by the major party nominees and want to express their frustration to a pollster, they may claim to back a third option. With the average favorability of Trump and Biden well underwater, according to FiveThirtyEight, the double-digit polling numbers for RFK Jr. are essentially a “cry for help” from the voting public, Lenski says.
A second reason independent candidates tend to fade is that the more voters learn about them, the less they like them. Early polling can’t capture how well these potential spoilers will wear over time. For example, a July 2016 Associated Press poll found Johnson’s favorable rating was 10 percent, and his unfavorable was 12 percent—indicative of his status as a relative unknown. But in September, following one of the former New Mexico governor’s major campaign gaffes—when he asked, “What is Aleppo?” on MSNBC’s “Morning Joe” in response to a question about how he would handle the besieged Syrian city—the AP found Johnson’s unfavorable rating more than doubled to 23 percent. By contrast, his favorable rating had also increased to 18 percent at the time, following the first wave of his paid TV ads, making measuring his strength all the more challenging.
Today, Kennedy’s familiar name likely buoys his poll numbers. Less well-known by average voters is his heterodox set of policy positions and penchant for baseless conspiracy theories.
“What people actually know about RFK Jr. and his views on specific issues is probably much less than what people know about Donald Trump and Joe Biden,” Lenski says. Once voters learn more, Lenski adds, “the less they might agree with him as a vehicle for expressing their distaste for having this Biden-Trump choice foisted upon them again.”
RFK Jr’s polling numbers are now being compared to Ross Perot, the billionaire independent presidential candidate, who cracked 20 percent in national polls by April 1992 and briefly led the three-way race presidential contest by June. But flimsy poll data can be misleading. Lenski points out that, unlike many third-party candidates, Perot was “generating actual positive support” after a series of well-received television appearances that displayed his trademark folksy wit and ideological running room as a protectionist fighting two free-trade candidates, Bill Clinton and George H.W. Bush. The independent insurgent would win 19 percent of the popular vote, the best showing by an independent since former Republican president Theodore Roosevelt ran as the Progressive or Bull Moose Party’s standard bearer in 1912. No one ever accused Robert F. Kennedy, Jr. of being folksy.
Another challenge for pollsters when assessing third-party support is figuring out who will show up. A registered voter who dislikes the two major parties may convey support for an independent to a pollster but, come Election Day, may not bother to vote. Determining who is a “likely” voter can confound pollsters.
Polls typically frame questions in ways that assume the respondents will be voting. According to Dan Cassino, executive director of the FDU Poll, if pollsters instead try to ask if a respondent is going to vote, they’ll all likely get a “yes” because people lie.
Including dishonest abstainers skews the results because voting history can be highly correlated with candidate preference. That dynamic is at play this year. The recent New York Times swing state poll found of those who didn’t vote in 2020, a plurality, 37 percent, said they would vote for RFK Jr.
Instead of asking respondents if they’ll vote, pollsters can rely on voter rolls, treating those who voted previously as more likely to vote in the upcoming election. But this poses a related problem: Predicting a respondent’s likelihood of voting and weeding out non-voters is only possible if the candidate field is similar to past elections. A strong third-party candidate, however, “by definition will turn out voters who didn’t turn out last time,” Cassino says. This predicament makes third-party polling, while not impossible, “really, really difficult,” he adds.
A case in point is Reform Party candidate and former professional wrestler Jesse Ventura, who, in his own bit of braggadocio, “shocked the world” when he won the 1998 Minnesota gubernatorial election. It was particularly shocking because Ventura was trailing in polls. But at the time, Minnesota was one of few states with same-day voter registration, and nearly one-third of Ventura’s vote came from people who were not even registered to vote before Election Day. Polls failed to capture Ventura’s base.
The best solution to zeroing in on faithful supporters of a third-party candidate, Cassino argues, involves polling two sets of voters. A “control group” can choose a candidate from a highly obscure operation, such as the Prohibition Party or the Constitution Party, and a second group can choose a more prominent option, such as Kennedy or West. The control group result provides a baseline. In the second group, the support registered above the baseline for the third-party candidate should be considered genuine. (Cassino ran this experiment in 2016 to show the actual support for Johnson and Stein in New Jersey, finding “true” support for Johnson at under 7 percent and a negligible amount of support for Stein, in line with their election day results.)
Another problem: On top of the difficulties of accurately assessing levels of support for third-party and independent candidates, pollsters struggle to understand the demographic drivers of such support. “If you have a sample size of 500, and you have RFK Jr. at 10 percent, you’re only going to have 50 RFK Jr. voters to analyze,” Lenski warns. “Anything under 100 [for a] sample size is going to have a ton of noise and a ton of sampling error.”
Weighting data to match the electorate’s demographics, which pollsters must do, can wreak havoc with small subsamples. One extreme example comes from the USC Dornsife/Los Angeles Times Daybreak tracking poll of the 2016 presidential race (in which the same people were repeatedly surveyed throughout the election cycle). The poll included a lone 19-year-old Black male Trump supporter. In trying to ensure the overall sample adequately represented young Black men, according to a New York Times analysis, one idiosyncratic voter was “weighted as much as 30 times more than the average respondent.” The final poll overstated Trump’s national popular vote share by five percentage points.
In theory, surveying larger groups of people would address the problem of misleading subsample data, but it’s costly. You have to gather a lot of data to get that 5 percent or 10 percent,” says Cassino,” So it’s expensive, you spend tens of thousands of dollars on this, and it’s very inefficient.” The same goes for oversampling, a survey technique that allows pollsters to generate larger samples of specific subgroups and is made more difficult if done for a fluid segment like third-party voters, compared to static demographic characteristics.
Despite the myriad challenges, Anna Greenberg, senior partner of GQR, a Democratic polling and research firm, must constantly monitor potential defectors that can swing an election. “Anyone who’s going to be undecided, be a soft supporter, be flirting with a third party, are going to be people we look at, and say, ‘Who are these folks?’” Greenberg told me. But ultimately, she says, understanding people who are at least considering third-party candidates is more art than science. To get a bead on their motivations requires augmenting imperfect quantitative data from polls with qualitative data derived from focus group discussions with likely and unlikely voters.
Greenberg noted that she hasn’t regularly expended the resources on deep dives of third-party supporters because we haven’t had “many situations where third-party candidates were going have a definitive effect.” Are we in one of those situations now, with Kennedy polling in double-digits?
A large amount of minor candidate support doesn’t automatically mean that candidate will be a spoiler. We haven’t seen an independent candidate consistently crack double-digits in polls since Perot in 1992 and 1996 (when he ended up with 8 percent). But the Texas computer mogul didn’t tip either election to Clinton. Exit polling found that Perot voters were split evenly between Bush and Clinton. (Some recent polling suggests that Kennedy also draws from the likely Democratic and Republican nominees in roughly equal proportion.)
But a minor candidate in the low single digits can potentially scramble a tight election, as Nader did in 2000. George W. Bush edged Al Gore by just 537 votes in the pivotal state of Florida. The Green Party candidate’s share of the Florida vote was just 1.6 percent, which amounted to 97,488 votes, vastly more than the margin of victory.
For the 2016 election, the winning margins in Michigan, Pennsylvania, and Wisconsin were within 1 percentage point. While the case that Stein’s Green Party campaign threw the election to Trump isn’t as airtight as the case that Nader’s 2000 campaign threw the election to Bush, it’s mathematically plausible. In her post-2016 campaign memoir What Happened, Rodham Clinton included the leftist physician on her list of culprits: “In each state, there were more than enough Stein voters to swing the result.”
But if there’s a time to panic, that time is not now. As Greenberg says, “There’s just no predictive value of these polls whatsoever.”
