John Hood: Polls underestimate GOP support
RALEIGH — Just before the 2020 elections, the RealClearPolitics.com polling average had Joe Biden leading Donald Trump nationwide by 7.2 percentage points. Biden did end up winning the popular vote, yes, but by 4.5 points. There were even larger gaps between pre-election surveys and vote totals in some battleground states.
Coming on the heels of polling misses in 2016, these results sent survey researchers scrambling for answers — because, contrary to myth, election polling has not always exhibited such a pronounced Democratic skew.
In 2012, for example, Barack Obama’s national margin of victory was 3.2 points bigger than the polls predicted. In 2008, Obama’s polling lead was 7.6 points and his popular-vote margin was 7.3 points, a tiny difference. In 2004, George W. Bush did outperform his polling average, but also by less than a point.
One popular explanation for the 2016 and 2020 gaps has been the “shy Trump vote,” the purported unwillingness of some poll respondents to admit their support for the controversial presidential candidate. Although there are surely voters who fit that profile, I think the theory is formulated too narrowly.
For one thing, the 2020 misses weren’t just about the presidential race. Here in North Carolina, Democrat Cal Cunningham went into Election Day with an average poll lead of 2.6 points in the Senate race. Republican Thom Tillis actually won by 1.8 points. Even starker was the difference between Gov. Roy Cooper’s polling lead (11) and his margin of victory (4.4).
Cato Institute pollster Emily Ekins recently offered two additional explanations for the recent Democratic skew in polling that sound plausible to me. While consistent with “shy Trump,” they suggest a broader phenomenon that could prove lasting.
The first explanation involves distrust. While trust in social institutions has been declining among most Americans, the drop has been greater among Republicans and Republican-leaning independents — who have become especially distrustful of such institutions as universities and the news media, often for good reason.
Who sponsors most pre-election polls? According to Ekins, nearly 80% of the surveys that came out in the month before Election Day 2020 were sponsored by universities or media outlets. “Consider that the sponsors of these polls often explicitly identify themselves when they contact respondents and ask them to participate in a survey,” Ekins wrote. “If most Republicans believe journalists and academics are politicized, it stands to reason they might assume the polls they sponsor are politicized, too.”
Some of these voters, then, will refuse to participate, skewing the sample. And even attempts by pollsters to weight their samples to compensate for this have inherent limitations. The kinds of GOP-leaning voters who agree to be polled may well differ from those who refuse.
Speaking of GOP-leaning voters who say yes, Ekins offers another explanation that isn’t so much about distrust as about dread. Voters may believe that if they talk honestly to pollsters, their political views will somehow be disclosed, subjecting them to scorn or ridicule. She suggests this may be especially true for those who work in corporations, universities, or other professional settings suffused with wokeness.
Here’s a striking finding: among those who didn’t go to college, the shares of Republicans and Democrats who worry their political views will harm them at work are about the same. Among voters with graduate degrees, however, 60% of Republicans express this concern, vs. only 25% of Democrats. That’s a lot of doctors, lawyers, managers, engineers, and other professionals (plus a few conservative professors). Some may refuse to take surveys at all. Others may answer only some of the questions honestly.
Have these feelings of distrust and dread intensified during the Trump era? Almost certainly. But they didn’t originate with him. And they won’t fade quickly. That argues for a dose of healthy skepticism when it comes to reading political polls. Don’t ignore the margin of sampling error. And don’t forget that there are other potential sources of error, as well, including the fact that some of us just don’t like the institutions sponsoring the polls.
