In many contexts, these errors don’t matter that much. When we see a poll indicating that 57 percent of Americans think that the empty Supreme Court seat should be filled by the winner of the presidential election, it doesn’t matter that much if the percentage is 54 percent or 60 percent, a swing of 3 percentage points in either direction. But when we’re looking at a close election in a swing state? Those errors matter.
They certainly mattered in 2016. Coming into Election Day that year, FiveThirtyEight’s average of polls in Michigan, Pennsylvania and Wisconsin had Hillary Clinton leading by 3.1, 3.1 and 5 percentage points, respectively. But Donald Trump won those states by 0.2, 0.7 and 0.8 points — meaning that the polls were off by between 3 and 6 points. Such a shift was always possible, given how polling works, but the big Clinton leads in key states meant that estimates of the likelihood of different outcomes of the presidential contest consistently indicated that Clinton was more likely than not to be the next president.
One additional effect of that change was that skepticism about polling skyrocketed. Yes, the national polling in 2016 was quite accurate, predicting a 3-point Clinton win in the popular vote, just off the actual 2-point margin. The misses in swing states, though, led to an awful lot of people throwing their arms up: Sure, the polls say former vice president Joe Biden is winning, but who can trust polls?
On Tuesday, we looked at how the national polling landscape is more favorable to Biden now than it was to Clinton in 2016. But it’s worth noting that the state landscape is similarly stronger for the Democratic candidate this year. In fact, as of writing, if the polls are as wrong in 2020 as they were in 2016, Biden would win the presidency, anyway.
We can visualize this. The interactive below uses current polling averages from FiveThirtyEight to show who’s expected to win each state. The darker blue or red, the bigger the lead for Biden or Trump, respectively.
What happens if the polls are off? To see what happens if they’re off in each state as much as they were in 2016, click the “same as 2016” button. At the moment, that still yields a Biden win: He picks up six states and wins the electoral college by nearly 100 votes. If the polls are as wrong as they were in 2016 but in the other direction — meaning that the polls understate Biden’s support, the swing to the Democrat is (predictably) even bigger.
Or uniform error of 0 points.
(One important note: This interactive doesn’t account for states that allocate electoral votes based on congressional voting, such as Maine and Nebraska.)
That’s the thing about polling errors. If pollsters are generally making similar, incorrect assumptions about the electorate, you can see swings in one direction or another. But that the swings in 2016 favored Trump isn’t necessarily how it always goes. In 2012, for example, polls often underestimated President Barack Obama’s state-level support by even bigger margins than the errors in 2016. People expected a close race that year, but Obama won fairly easily. The polls missed in a way that underestimated his support.
If you want to see how much a uniform shift in one direction or the other affects the results, feel free. We’ve included a slider letting you do that. You can also hover over a state to see how the shift in the poll numbers affects the outcome in the state.
Just because a shift similar to 2016 still gives Biden a victory at this point doesn’t mean that Biden is going to win. If the polls change — something that’s happened only modestly so far this year — so will his chances. (The interactive above is updated regularly with new averages.) If the miss in the polls is not uniform (a near certainty) or bigger than it was in 2016, that, too, could make Biden’s path trickier.
But if you had to pick which candidate you wanted to be at the moment, you’d want to be Joe Biden.