‘Electability’ Polls Are Unreliable and Need to Stop
What you should know about the surveys stalking the 2020 election
Kamala Harris faced many challenges in her presidential campaign before ending it this month, including doubts that she was “electable.” She wasn’t alone.
Last month, the news cycle was awash with breathless commentary in the wake of a November 4 New York Times UpShot/Siena College poll, published under this provocative headline: “One Year From Election, Trump Trails Biden but Leads Warren in Battlegrounds.” The poll came out as Warren had been surging in the polls to a position challenging Biden’s front-runner status in the Democratic presidential primary for the first time. Twitter and TV exploded with dire warnings of a Trump second term should Warren, an unabashed “progressive,” become the Democratic Party’s nominee.
The problem with this conclusion is that it’s based on “electability” polls that are unreliable, leading to erroneous narratives that can make or break campaigns, especially for lesser-known candidates who also seek to break through gender or racial glass ceilings like Warren and Harris.
Horserace polling is replete with electability polls because the electability question is central in voters’ minds and, as such, is the type of data heavily prioritized by media outlets. There are significant incentives to produce this type of polling but little scrutiny placed on the practice. Decades of political science scholarship shows that polling helps create narratives that can impact voter behavior, the ability of candidates to raise money, and electability, all of which tie to candidate poll performance in a positive feedback loop. Research shows that voters highly value candidate electability, defined as a candidate’s potential to compete against the opposition party’s nominee, as one of the most important factors driving their vote choice. Even in today’s hyper-ideological environment, two-thirds of likely Democratic primary voters in a FiveThirtyEight/Ipsos survey indicate they’d prefer a candidate who can beat Trump over one who aligns with them on the issues, even immediately following an ideology-priming event such as a debate.
The only candidates for whom head-to-head ballot tests are capable of reliably measuring “electability” are those who enjoy what I call “saturation” name recognition. The test only works when two or more equally well-known candidates are compared to each other. It is really important to illustrate how hard it is to reach saturation-level name recognition among the American electorate.
Even Biden, who served as vice president for eight years, is not universally known by voters. The most recent iteration of the Economist/YouGov tracking poll, which samples 1,500 American adults on a rolling basis, finds 15% of sampled adults unable to offer an opinion as to whether they approve or disapprove of Biden. The latest iteration of the Morning Consult Democratic primary tracking poll finds 8% of potential voters reporting they’ve “heard of, but can’t offer an opinion” on Biden and 1% have never heard of him, for a total of 9% in what I call the “unfamiliar with the candidate” category. However, it must be noted, we are now talking about a far more sophisticated population of voters: potential Democratic primary voters. Participants in presidential primaries are among the most engaged and informed voters in the country. Yet, 8% of these voters appear incapable of offering the most basic of opinions about a man who served as President Obama’s veep.
Name recognition remains a major issue for all three of these Democratic candidates among general election voters, even for Biden.
To the politically engaged, Warren must seem like a household name, but the portion of the electorate that is politically engaged is abysmally small, which is why despite years in national headlines, Warren started off the 2020 cycle with 31% of potential Democratic primary voters in her “unfamiliar with the candidate” categories. Seventeen percent of Democratic primary voters indicated they had heard of her, but were unable to express an opinion, and another 14% said they had never heard of her.
After moving into the top tier and becoming a real contender for the nomination, Warren was able to cut that number in half, down to 18% in the latest Morning Consult data. Yet 18% of the most tuned-in voters in the country — those who profess an intention to participate in the 2020 primary — remain ignorant of who Warren is as of November 2019.
It gets worse for the general election, according to data from Morning Consult published here for the first time. This data reveals that months into the Democratic primary cycle, name recognition remains a major issue for all three of these Democratic candidates among general election voters — even for Biden.
Nationally, 12% of registered voters in the Morning Consult data indicated they had heard of Biden but could not express an opinion about him, and 3% indicated they had “never heard of him.” The numbers were virtually the same for Bernie Sanders, but were significantly higher for Warren. Sixteen percent of registered voters indicated they had “heard of, but had no opinion,” about her and 14% said they had “never heard of” her, for a combined 30% “unfamiliar with the candidate” percentage.
In the battleground states the Times polled, including Wisconsin, Michigan, and Pennsylvania, name recognition is an even bigger problem for Warren compared to Biden and Sanders. However, it should be noted that the percentage of registered voters unfamiliar with the Democrats goes up for all three candidates.
For Biden, 16% of registered voters in Wisconsin and 18% in Michigan are incapable of expressing an opinion about him. In Pennsylvania, 14% of voters say they’ve “heard of but can’t offer an opinion” about him or “have never heard of” him. Interestingly, Sanders performs equally well on name recognition in the Midwest as Biden, which likely explains why, in poll after poll in these states, both Biden and Sanders appear to dominate the electability primary. Sanders, after all, campaigned in all three states in the 2016 primary, winning Wisconsin and Michigan.
In Warren’s case, combined, 34% of Wisconsin’s registered voters, 37% of Michigan’s, and 32% of Pennsylvania’s fail to register an opinion about her — nearly double the numbers for Biden and Sanders. This gap is more than enough to explain why she consistently trails them by a few points on the Trump head-to-heads in the Times/Siena poll, as well as other electability surveys.
Other explanations for the difference, such as ideology, are insufficient. After all, both Sanders and Warren are associated with Medicare for All, an issue that is often cited to “explain” Warren’s electability underperformance. As Sanders is fond of saying, he “wrote the damn bill!” and isn’t suffering the same purported consequences.
And what about Iowa? Is it, as the Times’ Nate Cohn argued, a state where Warren must be well-known due to the concentration of campaign activity there? The Morning Consult survey taken largely over the same time as the Times’ poll finds 31% of registered voters in Iowa saying they have “heard of, no opinion” and 15% who say they have “never heard of” Warren. This is, of course, nowhere as close to the name ID saturation of Trump.
Again, even among the much more tuned-in Democratic primary voters, establishing name ID for the 2020 candidates has been a challenge. Here in December, and five primary debates later, 39% of potential Democratic primary voters still report they’ve “heard of, no opinion” or “have never heard of” Pete Buttigieg, who is the current front-runner in Iowa.
The name recognition nut is tough to crack. Indeed, due to his media empire and status as a billionaire, Michael Bloomberg is already in fifth place in the Morning Consult national tracker in terms of polling support at 5%, vaulting over candidates who’ve been months on the stump, such as Amy Klobuchar and Cory Booker.
The point of all this isn’t to argue that Warren, or any other candidate, is electable or not, only that due to low knowledge in the electorate about most candidates, polls are incapable of telling us the answer. But when polls are released that purport to do so, by quality pollsters representing the nation’s best outfits, consumers of this data take their findings as gospel. Because of the potential impact on the actual performances of the candidates, those conducting presidential primary horserace polling must be sensitive to the narrative-setting applications of this data. The electoral fortunes of candidates may depend upon it.