As we close in on Election Day, polls tracking the horse race between Hillary Clinton and Donald Trump are everywhere. And as results are released, partisans and strategists work hard to discredit those in which their candidate does poorly and to emphasize those in which he or she is doing well. Some make dubious attempts to “unskew” all the polls so that they favor a particular candidate.
At least some people apparently believe that surveys are biased against their side.
That’s part of a general human tendency to reject uncomfortable information or facts that don’t fit one’s view of the world. There’s evidence that political partisans, like sports fans, discount or dismiss information that challenges their favored candidates or ideas, and even accept conspiracy theories that support their views.
In a survey, we found evidence that ordinary citizens are more likely to believe in the accuracy of polls that show their favored candidate winning.
Here’s how we investigated this
We showed 299 respondents a news story comparing two polls that had equivalent methodological quality (in sample size, representativeness, margin of error, response rate, etc.) but with different results. In one, Clinton was ahead 49 to 43 percent; in the other, Trump was ahead by an equivalent margin. Fully 70 percent of respondents recognized that the two polls were equally accurate. But when the others thought one was better than the other, it was almost always the one that favored their preferred party, as shown in the figure below.
Individuals were strongly biased toward evidence that their candidate was winning. The results replicated our earlier results on issue polling.
Don’t confuse me with the facts
In fact, people so completely rejected the idea that the opposing candidate could win that they endorsed polling results that favored their candidate even when that poll was objectively questionable.
Here’s how we know this: We assigned another 320 people to compare two polls with different levels of methodological quality. One of these used a large, nationally representative sample with a high response rate and low margin of error. The other used a low-quality online convenience sample. After we presented the polls’ results, an expert commentator’s remarks were quoted, explaining that the first poll (which had the relatively better methodological quality) was of higher quality and likely more accurate. The expert also debunked the poorly-conducted poll.
When the better quality poll reported that Trump was ahead, Republicans acknowledged that this poll was more accurate. But Democrats declared that the two polls were only equally accurate. However, when the first poll favored Clinton, Democrats reported that it was indeed more accurate while Republicans claimed that both were equally accurate on average.
In other words, our experiments reveal the psychological accuracy of the joke that goes, ‘My mind’s made up, don’t confuse me with facts.’ No matter what the evidence may show or what the experts may claim, most partisans see the world through the lens of their existing beliefs.
Ozan Kuru is a Ph.D. candidate in communication studies at the University of Michigan.
Josh Pasek is an assistant professor of communication studies and Center for Political Studies faculty associate at the University of Michigan.
Michael Traugott is a Center for Political Studies research professor at the Institute for Social Research, University of Michigan.
Data collected by Time-sharing Experiments for the Social Sciences, funded by the NSF, with Jeremy Freese and James Druckman as principal investigators.