Fool me once, shame on you. Fool me twice, shame on me.
That’s how a lot of Democrats are feeling about pollsters and election forecasters. In 2016, polls suggested that Hillary Clinton would defeat Donald Trump — albeit narrowly. In 2020, polls pointed towards a Democratic victory. This time, they were right — but Trump again outperformed his numbers, and due to the increase in mail-in ballots Democrats had to wait until Saturday to celebrate Joe Biden’s victory. Additionally, polls favored Democrats’ chances in both the Senate and the House, but as election night wore on liberals saw their chances at the Senate slip away, and they’ll end up losing seats in the House (though they’ll maintain a majority).
What gives? Why have the pollsters been wrong? And why is it a problem?
First, we should withhold extreme judgment until all the votes are counted and we have all the data. Right now, Biden leads Trump nationally by around 3 points; given that many of the outstanding votes are in California and New York, we can expect that margin to increase, perhaps to as many as a 4 or 5 points.
That still leaves us with a polling error in Trump’s favor — FiveThirtyEight’s final polling average had Biden ahead 8.4 points. Additionally, as Nate Silver wrote on Twitter, only one swing state had a polling error that favored Biden. All others favored Trump — Wisconsin favored Trump by whopping eight points.
The situation is worse in the Senate. Polls in South Carolina suggested a close race, even if the fundamentals favored incumbent Lindsey Graham. Far from being a tight race, the contest was one of the first to be called on election night as Graham trounced Democratic challenger Jaime Harrison by ten points. In North Carolina, nearly every poll conducted gave Democrat Cal Cunningham an advantage over incumbent Thom Tillis — the usually reliable Marist College poll even had him up ten points. Tillis topped Cunningham by over 90,000 votes; Cunningham finally conceded yesterday.
Perhaps the most egregious error came in Maine. The last poll showing Susan Collins ahead of Sara Gideon came all the way back in July, though polls suggested that the race would be close. The race was not close; Collins blasted Gideon by nine points.
There’s a saying that the only poll that really matters takes place on Election Day. And yes, voters saying they’ll vote for Biden matters far less than voters actually voting for Biden. But polls measure more than the status of the election. Opinion polls are a valuable means by which to measure the sentiment of the American people. Presidents including Franklin D. Roosevelt, Richard Nixon, and Bill Clinton all used opinion polls when weighing policy decisions — no one wants to enact a policy hated by a majority of Americans.
Progressives, for example, are quick to point out that their keystone policies including Medicare for All and a Green New Deal are broadly popular among Americans, according to opinion polls. What if these opinon polls, like the polls in the 2016 and 2020 elections, are undercounting conservatives? Then these policies are perhaps more of a 50/50 bargain rather than a popular policy initiative.
Election polls also dictate spending patterns. National committees, PACs, and small-dollar donors alike want to put their money where it can make the most impact; they look for winnable races where their influx of cash can tip the race. Two polls in August showed Democrat Amy McGrath just 3 points down on Senate Majority Leader Mitch McConnell. Maybe, Democrats thought, they had a chance to oust the second-most hated man in Washington. McGrath raised $88 million. She lost by 19 points.
South Carolina offers a similar story. Throughout October, polls showed Harrison neck-and-neck with Graham. Democrats nationwide jumped at the opportunity to win back a deep-red Senate seat and stick it to one of Trump’s top toadies. Harrison smashed fundraising records, raising over $100 million. He lost by ten points.
Of the top seven Democratic Senate candidates by fundraising, just one won their election. The six losers collectively burned through $400 million and have nothing to show for it. Perhaps this is a case of overeager Democrats desperate to retake the Senate, but their eagerness was fueled by poll after poll offering an imperfect view on the state of the race.
How did the polls get it this wrong, again? Perhaps some conservative voters came back to Trump at the end, but this doesn’t seem likely given the relatively stagnant nature of the polls throughout the race. The myth of the shy Trump voter doesn’t make sense either — there would have to be shy GOP Senate voters as well to explain the widescale polling error.
One reason pollsters might be missing Trump voters–and right-leaning voters in general–is the distrust those voters tend to have for institutions. GOP voters tend to be more skeptical of big government and big media. A small segment even believe in conspiracies like Pizzagate and QAnon. Voters that are so distrustful of “mainstream media” might be less likely to speak to pollsters, and would thus be underrepresented in the final tallies.
Yet aren’t we all less likely to speak to pollsters? Calling voters has long been the gold standard in data collection, yet more and more voters are eschewing landlines for cellphones, and fewer individuals are willing to pick up a call from an out-of-state number. Pollsters may need to consider revolutionizing their collection methods — though more research needs to be done to understand the mode effects changing media will have on survey results.
While pollsters take a good, hard look at what went wrong with their methodologies in 2020, it would be wise for us to reflect on our own errors as well. Polls cannot give us certainty — no matter how accurate the polls may be, they cannot guarantee the outcome of an election. We treat pollsters as oracles when they should be treated as meteorologists. If the weatherman says there’s a 70% chance of sun, and it rains, do we discard meteorology? No — we accept the uncertainty of an unpredictable world, and move on with our day.
The parlorization of politics has vaunted pollwatching into the realm of sport. We are all gambling addicts, watching the polls like degenerates chasing 20-1 odds on the filly at Saratoga. Polls are not scoreboards. We need to stop treating them as such.