How Do You Poll With a Candidate Like The Donald?

Within the political polling business, the greatest challenge is getting a representative sample of the people who will turn out to vote. Different demographics have different voting patterns which can be difficult to predict.

In this year’s presidential race, that challenge is still in place but an even greater one is emerging: survey respondents might be lying when asked to give their opinions.

Lying or confused respondents is a common challenge that social scientists have consistently had to deal with over the decades. That’s why the best studies actually don’t rely on self-reported analysis and instead try to rely on direct observation in order to overcome “social desirability bias,” or the tendency of people in to avoid telling others about unpopular opinions.

Within the political realm, social desirability bias was first noted by political science researchers in the 1982 California governor’s race. Days before the election, most polls reported that black candidate Tom Bradley was ahead of his white opponent, George Deukmejian. Surveys conducted on the day of the vote also indicated that Bradley would win. That proved untrue, however, because a least some white respondents appear to have falsely told pollsters that they would be supporting Bradley. Thereafter, the idea that white support for a black candidate might be lower than surveys indicated was colloquially called the “Bradley Effect.”

In the UK, political scientists have identified a “Shy Tory Factor” because supporters of the Conservative Party have been observed for many decades to be less willing to tell pollsters that they would be voting that way. In 10 of the last 12 British elections, polls underrated the Tory vote. A similar accuracy flaw was observed in the recent Brexit vote when the average of the final surveys indicated that “Leave” support stood at 44.3 percent. In the actual tally, Leave won with 51.9 percent. In France, a social desirability effect has been observed as well where polling has overstated the support of left-leaning parties and understated the support of nationalist parties.

Closer to home, a 2013 Pew Research Center study found a similar dynamic at work around the issue of same-sex marriage; surveys consistently understated opposition to legalization of these unions by around 6 points. According to the organization that conducts exit polls for the American television news operations, these surveys end up “more often overstating the [vote for a] Democrat, but occasionally overstating the Republican.”

The one common factor behind all of these countries is that their national media organizations are overwhelmingly dominated by people with left-of-center views. As a result, people who lean rightward sometimes tend to feel alone in their views and thus less willing to express them.

Fans of the Rush Limbaugh program have provided a perfect example of this mentality in their different usage of the word ditto. The conservative radio host formally defined the term in 2007:

Back when this show started August 1st, 1988, it took the nation by storm because there was nothing like it in the national media. The national media was all liberal. Here was this conservative program that reflected the views of millions of people. As people would call in, the first couple minutes of their call, literally, they’d spend thanking me and talking about how great it was to have something like this on the radio, finally, it was so great, and I of course loved hearing it. After awhile, after about six months, it finally just grew old. It was delaying getting to the discussion of the issues. A woman called from I think it was like New Hampshire, and after just one of those calls, said, “Ditto to what they guy just said.” So ditto means, “I love the program. Don’t ever go away.” It doesn’t mean, “I agree with you.” It doesn’t mean, “You’re always right.” It means, “I love the program.”

All of this is very relevant to GOP nominee Donald Trump. His frequent outrageous statements and nationalist positions have made him a constant target of journalists. Because of this frequent criticism, people supporting the real estate mogul sometimes avoided telling pollsters that they intend to vote for him. In one study, Republican respondents answering an online survey were found to be 6 points more likely to say they supported Trump than those who were surveyed by a person on the phone.

Throughout the GOP primaries, online and automated surveys—which are generally less respected within the industry—indicated higher levels of Trump support than live ones. The discrepancy may have been a large reason why his Republican elite opponents refused to take the former reality star’s campaign seriously before it was too late.

A similar dynamic appears to be taking shape in the general election. Instead of taking false comfort, Democratic elites are sounding the alarm about the Shy Trump Supporter Effect.

Liberal pollster Celinda Lake is warning of a “secret Trump vote” which she discerned from differences between automated and live surveys conducted for the general election match-up.

“The pattern is in the online surveys, even if you control for demographics, Trump does three to nine points better than in telephone surveys,” Lake said in an interview with the Washington Examiner. She also believes that there are many blue-collar white voters who typically don’t vote in elections who might be interested in supporting Trump’s unconventional Republican campaign.

Lake’s concerns are important in the light of an extensive analysis by the New York Times which found that exit polling techniques significantly underestimate the number of lower-middle class white Americans and that they are much more important to Democrats than less comprehensive research had indicated. According to the Times report, these “missing white voters” tend to be younger, less religious, and lean Democratic.

Worry about this voting bloc may have played a role in Hillary Clinton’s decision to pick Virginia senator Tim Kaine as her running mate despite her difficulty fending off a surprisingly strong primary challenge from progressive Vermont senator Bernie Sanders. While he has some liberal stances, Kaine has consistently shown an ability to get at least some white conservatives to vote for him in his swing state. Many party stalwarts had favored someone well to the left such as Massachusetts senator Elizabeth Warren or perhaps someone of Hispanic or African descent.

Only time will tell what’s in store come November 9, but things are looking much better for Trump now that he’s seemingly managed to consolidate Republican voters behind him after the party’s convention last week in Cleveland. Currently, he leads Clinton in five of the six most recent surveys. If Lake is right, Trump’s numbers might be even higher than that.

Or perhaps not. After all, the 2012 election featured a huge amount of incorrect and inexpert opinion about how polls showing Mitt Romney trailing Barack Obama were “skewed.” This time around, however, things are different since actual polling experts, including Democrats, are joining in the discussion. The other possibility is that Trump could lose but by less than polls indicate.

The difference between how Democratic and Republican strategists are dealing with this uncertainty is quite notable though. Democrats are taking it as a serious possibility. During their primaries and despite much evidence, the Republican “experts” didn’t. Given their abysmal record of incompetence, that’s not exactly a surprise.

Photo by Gage Skidmore