One Week Out, Comparing 2020 to 2016
On Nov. 6, 2016, two days before the last presidential election, Politico reporter Katelyn Fossett outlined the 16 worst political predictions of that year. Among them: Bernie Sanders did not stand a chance in the Democratic primary (he nearly took the nomination from Hillary Clinton), Jeb Bush will be the Republican nominee (he was not), and Democrats, finally, will turn Texas blue (they did not).
That list, of course, did not include the absolute worst but most common political prediction of 2016: Donald Trump would lose the general election to Hillary Clinton.
He of course did not.
Four years later, we are all hearing similar certainty about the presidential outcome this fall, including the possibility of a blue wave crashing into Texas. Will those predictions be wrong again? Where exactly does the race for the White House stand today, seven days out from the election, and how does that outlook compare to what was predicted in 2016?
Let’s take a look, first by examining the RealClearPolitics national polling average. On Oct. 31, 2016, one week before Election Day, Hillary Clinton had a 3.1 percentage point advantage over Donald Trump nationally. She was polling at 48.0 percent while he was at 44.9 percent. That advantage had narrowed from seven points just 14 days earlier, on Oct. 17.
As of today, Oct. 27, seven days out from Election Day 2020, Joe Biden leads Donald Trump by 7.3 points in the very same RealClearPolitics average national poll. That advantage has narrowed from 10 points on Oct. 13, 14 days earlier.
RealClearPolitics also calculates an average of battleground state polls – the contests that, because of the Electoral College, will ultimately decide who wins the presidential race. With seven days until Election 2016, Clinton had a 2.8 point advantage over candidate Donald Trump in that reading. With one week to go until the 2020 contest, Biden has a 4.1 point lead over President Trump.
And what did the Electoral College predictions look like in 2016?
There was widespread agreement that Clinton would win the race to the 270 electoral votes needed to clinch the White House. Here is what Time magazine reporters wrote “just hours” before the polls closed on Nov. 8, 2016: “[P]ollsters and predictors have released their final maps of the 2016 election—and most agree that Hillary Clinton will win, but no one agrees by how much.”
Specifically, Time looked at seven different predictions, including:
FiveThirtyEight. The data analytics website concluded Hillary Clinton had a 71.8 percent chance of winning the White House and would likely take 301 electoral votes. Trump’s chances for electoral victory were just 28.2 percent and the website thought the Republican nominee would top out at 235 Electoral College votes.
Princeton Election Commission, which predicted a Clinton win with near certainty. In fact, the commission put her chances of victory at 98 to 99 percent.
Larry Sabato, the election sage from the University of Virginia’s Center for Politics, predicted Clinton would earn a whopping 322 electoral votes from states including Wisconsin, Michigan, Florida, Pennsylvania, and North Carolina. She lost every single one of those states.
Today, FiveThirtyEight puts Biden’s chances of winning at 87 percent; the Princeton Election Commission estimates Biden will win 355 electoral votes; and analysts at the University of Virginia think Biden will win a host of swing states including … Wisconsin, Michigan, Florida, Pennsylvania, and North Carolina.
Of course, these predictions are all based on polling. Going back to Katelyn Fossett’s 2016 pre-election article, which, again, was published two days before the election, her 13th worst prediction was that “polling was dead or dying.” Even before Election Day 2016, analysts were already casting doubt on some polls because they were incorrect during primary season, particularly in Michigan. Fossett suggested that polling mechanisms were fine, however. In fact, she wrote, “Aside from the big Michigan primary upset, this year, polls were fairly accurate throughout the primaries, and news consumers still seem to be as poll-obsessed as ever.”
Of course, as it turned out, 2016 general election polling was not quite on the mark. Donald Trump significantly outperformed the polling in swing states. For example:
The final RealClearPolitics (RCP) average of polls in Wisconsin showed Clinton with a 6.5 point advantage. Trump ended up winning the state by 0.7 points.
Trump was ahead in the final RCP Florida average, but by only 0.4 points. He actually won that state by 1.2 points.
In Michigan, the final polling spread was 3.6 points in favor of Clinton, but Trump won by 0.3 points.
Trump won Pennsylvania by 0.7 points, but the RealClearPolitics average of polls had him 2.1 points behind going into Election Day.
In North Carolina, Trump was up in the final RCP average by 0.8 points, but he won the state by a much larger margin, 3.6 points.
It was not only Donald Trump who overperformed, however. According to the American Association of Public Opinion Research, “[S]o did most Republican candidates in competitive Senate races” and “Republican candidates for the U.S House of Representatives also tended to outperform their poll numbers. Nationally, the actual congressional vote was +1.1 for Republicans, whereas the final polling average from RealClearPolitics was estimated at +0.6 for Democrats.”
Why did analysts get it so wrong in 2016?
Generally, Americans have heard explanations that blame technology. With mobile phones in pockets that tell users who is calling, Americans are simply less likely to take pollsters’ calls. Another commonly-cited reason is that voters aren’t enthusiastic about telling a pollster that they are planning to vote Trump, but in a 2017 New York Times article reporter Nate Cohn said, “[T]here’s not much evidence” for that claim. In fact, “Trump did not fare noticeably better in polls conducted online, where voters wouldn’t have to admit socially undesirable views to a person on the other end of the telephone.”
On Nov. 9, 2016, the day after Election Day, Reuters offered a different analysis. Cliff Young, president of Ipsos Public Affairs US, Reuters’ polling partner, said the errors “came down to the models the pollsters used to predict who would vote – the so-called likely voters.” Most pollsters miscalculated the demographics (voters’ age, race, and education), Young said, “and turnout was lower than expected, a result that generally favors Republican candidates.”
This year, since voting is happening earlier, we might have a better window into who is voting, at least when it comes to party registration.
According to the U.S. Elections Project, as of Oct. 25, nearly 60 million people had already voted, more than the 47.2 million who cast ballots before Election Day 2016. By Oct. 25, Texas had reached 80 percent of its total ballot numbers from 2016; Montana had garnered 70 percent; and Georgia, North Carolina, and Tennessee had each reached 65 percent.
Because of the coronavirus pandemic, more Americans have been expected to cast ballots before Election Day this year. Still, this data seems to indicate that the country could be on track for record turnout. The question is: who is turning out?
The U.S. Elections Project said more than 23.5 million Democrats requested mail-in ballots, and, as of Oct. 25, about 11 million (approximately 47 percent) have been returned. Republicans requested 13.4 million ballots and just over 5.1 million have been returned. That’s a return rate of just over 38 percent.
What about the all-important independent voter? More than 15 million voters with no party affiliation requested ballots, but only 4.8 million have returned them. That makes for a dismal return rate of 31 percent. According to exit polls, Independents went 46 percent to 42 percent in favor of Trump in 2016. Dampened turnout by unaffiliated voters could hurt the incumbent.
In all, only 40 percent of mail-in ballots have been returned, well below the overall 2016 voter turnout rate of 56 percent. Of course, these numbers were calculated with nine days left to go in the election. Voters have time to postmark their ballots. Still, if turnout is low …
Of course tens of millions of Americans also will vote in person on Election Day. According to the U.S. Elections Project, “Republicans need to vote in-person to make up ground on the Democratic mail voters.”
The U.S. Elections Project analysis believes Republicans have a three-point advantage when it comes to the number of Americans who are likely to vote in person. Voters with no party affiliation make up nearly 22 percent of the individuals expected to vote on Election Day, Democrats are expected to make up 41.5 percent, and Republicans just over 40 percent.
Will that be enough? Fortunately, we won’t have to wait much longer to find out.