Commentary: Survey says?

This is where I reassure panicky Democrats and assorted anti-Trumpers that it’s all going to be okay. With under a month to go, the polls paint a reasonably optimistic picture of 2020’s electoral landscape. At this point in the column, readers are urged to refrain from having a glance at my then-seemingly-spot-on August 2004 piece for the Nashville Scene titled “Ten Reasons Why Bush Will Lose.”

Granted, it is sensibly mandatory doctrine for Dems everywhere to avoid the slightest tinge of October confidence lest they sentence themselves to reenacting the electorally collegiate horror of 2016. “Ignore the polls and get out the vote,” that sort of thing. 

Many assume that the story of Trump’s 2016 win is in good measure a tale of polling error. Yes, some polls got it wrong, but it’s worth reminding ourselves that the polling miss wasn’t crushing because it was huge, but rather because it was surgically devastating. As had been well-documented elsewhere, in 2016 the national polls essentially got Hillary Clinton’s popular vote win right, and state polls were within the margin of sanity in all but a few states. 

Alas, the few states that were the exception were the crucial ones—Michigan, Pennsylvania, and Wisconsin—whose flip from blue in 2012 to red in 2016 flipped the election. In the big dog state of the three, Pennsylvania, the polling “miss” (gap between the final Real Clear Politics polling average and the actual result) was actually quite modest at 2.6%, and smaller than gaps in several states the polls got “right.” But many treat Pennsylvania as a polling “wrong” because the gap straddled the zero point of a tie (a Clinton 1.9% polling lead preceding a Trump 0.7% electoral win ). Wisconsin is the one of the crucial three where the miss was substantial: Clinton’s polling lead of more than six percentage points shortly before the election yielded to Trump’s 0.7% winning margin. 

In all three states the final pre-election polls had Clinton’s vote pretty accurately pegged; the shifts that led to the gaps overwhelmingly took the form of unexpected Trump vote. The many post-mortems of 2016 converged on a now-familiar set of likely reasons for this: (1) too few non college educated voters polled in key states, (2) a failure to capture shifts by very late deciders who broke bigly for Trump, and (3) faulty estimates of who was likely to vote (overestimating Black turnout and underestimating rural whites). It’s fair to believe that the first two of these things are less likely to be contorting the polls this year. Most good polls now “weight” for education—making statistical adjustments to their samples to make sure non college voters aren’t undercounted. And while capturing late-breakers is always hard, more polls closer to the election combined with a more mind-made-up electorate with many fewer undecideds this year lessens the risk. 

The last of the three is the thornier likely-voter problem—the eternal challenge figuring out who will actually turn out to vote. You can’t merely ask when you poll (though polls do ask) because many people fib about that. Different polls model turnout in different ways, and many aren’t transparent about their methods. Pollsters may well have learned from their miscalculations in 2016 but each cycle is different, especially when you throw in a pandemic changing the voting plans of large segments of the electorate. 

An explanation for 2016 worth dismissing is the “shy Trump voter” theory, which has been pretty persuasively debunked. Polls that undercount Trump support are missing likely voters, not surfacing lying voters. 

The point of this semi-painful rehash of misbegotten 2016 polling, and its message for 2020’s fibrillating Democrats, is twofold. First, the polling four years ago may have missed the mark, but not by as much as many think they recall: if you’ll pardon a lazy weapon metaphor, aim was off a bit but the gun didn’t misfire. Second, while changes in best practices in the polling game don’t guarantee they will get it entirely right this time (because getting it entirely right is not what polls are meant to do), judgments that polls remain legitimate and can improve qualify as a non crazy belief system.  

But it’s more than just any crazy belief system; it’s also an essential one. Polls actually do matter—not merely as handy utensils for biennial political consumption (and obsession), but as significant instruments of democracy. It matters that polls work well as measures of public opinion, and it matters that we trust them to do so. They provide the crucial evidence that our political system is failing us by disregarding overwhelming public support for everything from gun laws to reproductive choice to expanding Medicaid. It is vital that thoughtful voters not let polling frustrations turn them into polling denialists.  

I closed my astonishingly not-so-prescient piece on Bush in 2004 with a declaration that voter ignorance has run its course: “The voting public is not as dim as many assume, and its tolerance for presidential vapidness has limits….Bush will lose as swing voters come around to the British poet Edith Sitwell’s view: “I am patient with stupidity but not with those who are proud of it.”

Sixteen years later we can all agree that Sitwell’s quotation and my use of it haven’t aged well. Reliable measures of public opinion will never cease to matter, but perhaps there is something to be said for attending more to the ballot then to the polls for these last few excruciating weeks.