With the final ballots yet to be counted, and with the next President still unknown, the polling post-mortem is well under way.
“So much of the experience of watching returns,” remarked frustrated MSNBC anchor Chris Hayes to two leading pollsters, “was anchored by the polling going into it. And it was a bad night for polling—an enormous polling miss.”
Far from learning from the mistakes of 2016, the polling industry seemed to have got things worse. Whether conducted by private or public firms, at the national or local, presidential or senatorial, levels, polls were off by wide margins. The Five Thirty-Eight final poll of polls put Biden ahead by 8.4 points, but the actual difference in popular vote is likely to be closer to 3-4 points. In some close state races, the error was even greater.
Why did they get it so wrong? Pollsters typically receive low response rates to calls, which leads them to undercount key demographics. To get around this, they typically weight for key categories like race, education or gender. If they get too few Latinos or whites without degrees, they adjust their numbers to match the actual electorate.
But most attitudes vary far more within a group like university graduates, than between graduates and non-graduates. So even if you have the correct share of graduates and non-graduates, you might be selecting the more liberal-minded among them. For example, in the 2019 American National Election Study pilot survey, education level predicts less than 1% of the variation in whether a white person voted for Trump in 2016. By contrast, their feelings towards illegal immigrants on a 0-100 thermometer predicts over 30% of the variation. Moreover, immigration views pick out Trump from Clinton voters better within the university-educated white population than among high school-educated whites. Unless pollsters weight for attitudes and psychology – which is tricky because these positions can be caused by candidate support – they miss much of the action.
Looking at this election’s errors — which seems to have been concentrated among white college graduates — I wonder if political correctness lies at the heart of the problem.
Political correctness refers to the policing of speech so that it conforms to cultural taboos, especially the ones concerning race, gender and sexuality. Those who wield taboos gain rhetorical power, encouraging them to stretch the meaning of concepts such as racism to encompass non-racist actions such as voting for Donald Trump.
Across all racial groups, 80% of Americans say “political correctness is a problem in our country”. Only the small “Progressive Activist” 8% of the US population largely thinks it’s not. In practice, the burden of political correctness arguably falls most heavily on university-educated Republican supporters. Data from a recent Cato Institute survey shows that 88% of Trump-voting graduates compared to just 44% of Clinton-supporting graduates agreed that “The political climate these days prevents me from saying things I believe because others might find them offensive.”
Republican supporters with degrees tend to work in graduate-dominated environments, where organisations and peers are more likely to enforce norms of political correctness. As a result, it is highly-educated Republican supporters who are most shy about revealing their beliefs at work.
As figure 1 illustrates, 45% of Republicans with degrees, compared to 23% of Democrats with degrees, said they feared that their careers could be at risk if their views became known.
How does this affect polling? Republican pollster Frank Luntz told Emily Maitlis that Trump voters were over twice as likely as Biden voters — by a 19 to 9 margin — to conceal their intended vote from others. I would expect this ratio to be considerably higher among university graduates, which would, accordingly, skew predictions the most among graduates.
Pollsters claim to have overcome this problem by comparing telephone and online surveys and finding no difference. Since online surveys are anonymous, they reason, a ‘shy Trump’ effect should reveal itself by comparing these two methods, and they find none.
However, we also know that people who internalise social norms often conceal their views in online surveys. The psychologist George Herbert Mead referred to people’s ‘generalized other’, a kind of mental peer group we carry around in our heads that sits in judgement upon us is even if no one is actually watching. For instance, in a recent survey of North American academics, I found that just 23% of academics were willing to state they would discriminate against a Trump voter for a job, but the actual share when using a concealed technique called a ‘list experiment’ was 42%. Likewise, a 2010 study found that the share of white Americans willing to endorse zero immigration jumped from 39% to 60% when the question was concealed in a list, rather than asked openly.
There is also a problem of blowback among elite Republicans. Frank Luntz has also said that feedback from Trump-supporting respondents revealed considerable resentment towards pollsters, who were perceived as part of a media establishment out to misrepresent them. Indeed, studies show that using words like ‘racist’ to describe Trump or his policies tends to increase support for them among conservative respondents by provoking what is termed a ‘reactance’ effect. Knowing you are perceived as racist by elites for supporting Trump may make you less likely to answer a call from a survey firm you associate with that chastising elite.
Again, this perception is likely to be stronger among Trump-supporting graduates than Trump voters with lower education levels, who are less likely to circulate in politically-correct social environments. Research confirms that highly-educated white liberals have the most skewed perceptions of the actual views of Trump supporters, in part because their social circles tend to be politically homogeneous. The problem is worst among those most attentive to politics.
This may explain why the polls didn’t do badly in predicting the white non-graduate vote but failed miserably among white graduates. According to a Pew survey on October 9, Trump was leading Biden by 21 points among white non-graduates but trailing him by 26 points among white graduates. Likewise, a Politico/ABC poll on October 11 found that ‘Trump leads by 26 points among white voters without four-year college degrees, but Biden holds a 31-point lead with white college graduates.’
The exit polls, however, show that Trump ran even among white college graduates 49-49, and even had an edge among white female graduates of 50-49! This puts pre-election surveys out by a whopping 26-31 points among white graduates. By contrast, among whites without degrees, the actual tilt in the election was 64-35, a 29-point gap, which the polls basically got right.
Even with reasonable caveats about the accuracy of exit polls, the size of the polling error among white graduates is the 800-pound gorilla in the room, and is arguably linked to problems of concealment and reactance which loom largest among white graduate Trump voters. That’s why, in order to fix polling, we first need to fix political correctness.
The most important effects of speech policing are often indirect, shutting down important conversations across value divides that could improve policies and reach greater consensus on hot-button issues such as immigration, education or policing.
Like a referee in a football game, we need to call penalties when real racist or sexist infractions have been committed, but must stand firm against cries to those based on false accusations grounded in hair-trigger suspicion and conceptual stretching. Failing to do so silences voices that need to be heard, such as college-educated Trump voters, while breeding anti-elite resentment that demagogues exploit. If America cannot reform its regime of speech discipline, it has no hope of overcoming its yawning cultural divide.