The Washington PostDemocracy Dies in Darkness

The other electability problem: Voters aren’t great at determining electability

Democratic presidential candidate and former vice president Joe Biden greets a woman in a crowd on April 29. (Jonathan Ernst/Reuters)

The undercurrent in the swollen Democratic presidential primary field has become a near-audible rumble: Who can defeat President Trump?

Time and again, journalists who have spoken with Democratic voters report that this is a — if not the — significant consideration being used to evaluate candidates. The candidates themselves are aware of it; it’s why Sen. Bernie Sanders (I-Vt.) released an internal poll showing him edging out Trump in the three states central to Trump’s 2016 victory. Every general election poll that comes out, the rumble is there, rumbling away: Who’s got the best shot?

It’s not an irrational way to approach the subject. It’s just . . . problematic. Why? Because electability has been a questionable indicator in the past and because it’s heavily colored by individual prejudices.

In October 2003, as Democrats were jockeying for the opportunity to face George W. Bush in the 2004 presidential election, The Post and our polling partners at ABC News surveyed Democratic voters to gauge their preferences for a nominee. At the time, former Vermont governor Howard Dean had a slight lead. The candidate who the most respondents saw as most electable, though, was Sen. Joe Lieberman of Connecticut.

Well, given the margins of error, Lieberman, Dean and two other candidates were all sort of mushed together as most supported and most electable. That’s one key pattern here: The candidates that are seen as most electable are generally also those who have the most support.

So the eventual nominee, Sen. John F. Kerry of Massachusetts, was in fifth on both metrics. Not much support, not generally seen as electable.

Of course, Kerry lost. Perhaps Lieberman, who’d nearly been elected vice president four years prior, really would have been more electable. That’s another problem with these assessments of electability: It’s nearly impossible to determine how accurate the predictions were since there are no countervailing scenarios in which other candidates ended up facing off in the general election.

In October 2008, the candidate viewed as most electable wasn’t the eventual nominee, Sen. John McCain of Arizona — it was former New York mayor Rudolph W. Giuliani. From October to December, Giuliani lost support for his candidacy as former Massachusetts governor Mitt Romney, now a U.S. senator from Utah, surged. McCain, too, lost both support and a sense that he was most electable. The two move in tandem: A candidate that gains support is also seen as more electable.

A really, really great example of that came in 2012. The Post and ABC polled in September and October of that year, witnessing a massive shift in how Republicans viewed the field. In the October poll, Texas Gov. Rick Perry had a lead and was seen as the most electable.

During a debate in early November, though, Perry suddenly forgot one of the three federal agencies he wanted to shut down. His “oops” response came to define his candidacy, and it collapsed. Businessman Herman Cain and Romney picked up much of the slack.

But consider what that graph shows. Perry was seen as electable — until one moment eroded much of that sense that he could defeat President Barack Obama. That’s another problem with electability: One gaffe, one incident, can torpedo a candidate’s chances, making electability a tricky thing to evaluate in the abstract.

In 2016, Republican voters did a little better. Trump held both a lead in the polls and was considered the most electable. From October to January, our polling tracked the collapse of Ben Carson and the rise of Sen. Ted Cruz (R-Tex.). But neither came close to Trump.

Was Trump really the candidate best positioned to beat Hillary Clinton that year? Maybe. Maybe this is a situation where the electorate got it right.

But we can just as quickly point to a situation where they got it wrong: The 2008 Democratic contest.

Clinton was a favorite in this one, too, until Obama started gobbling up delegates. Obama and former senator John Edwards of North Carolina weren’t seen as being nearly as electable as Clinton. Obama, of course, proved himself to be more than sufficiently electable.

Notice something interesting about his line, though. Obama’s support in polling was higher than the percentage of people who thought he was electable. It was a common question in 2007: Was America ready to elect a black man as president? It was, but it wasn’t clear at the time that voters would agree.

In March, HuffPost and YouGov conducted a poll focused on the question of electability, asking voters if they thought a candidate would be more or less electable depending on their gender, race or age.

Here, for example, are the results of the question asking if a nonwhite candidate would gain or lose support because of their race. White voters mostly thought it wouldn’t make a difference. Black voters were most likely to say that they weren’t sure what the effect would be, though significantly more said that the candidate would be less likely to get support from voters because of their race than said that they’d be more likely to get support.

Interestingly, looking only at Democratic voters, the response was different. Nonwhite Democrats were significantly more likely to say that a nonwhite candidate would see more support because of his or her race than were white Democrats.

On the question of gender, men and women overall were about as likely to say that a female candidate would see less support because of her gender.

Among Democrats, though, there was a pronounced gender difference. Women were 8 points more likely to say that a woman would be less likely to get electoral support . ..

. . . and 10 points more likely to say that a man would get more electoral support because of his gender.

The Post’s Dave Weigel commented on that earlier this week.

That sentiment is understandable after the 2016 election, when Clinton was seen as the more electable candidate over Trump until polls closed in Western states on election night. (Whether another Democratic candidate would have been more electable that Clinton has been a constant refrain in the years since her loss.)

The point is, though, that views of the electability of women among Democrats are linked to the gender of the respondent, as are views of the electability of nonwhite candidates. The other point is that these predictions have a very iffy track record, though they are much more likely to be proven wrong than right.

Trying to predict who was going to win the 2016 election proved elusive for literally the entire duration of the campaign. Trying to predict which Democrat is best suited to beat Trump this time 18 months in advance seems similarly fraught.