Skip to main content
nateessay-03-4x3

This is the third article in a series that reviews news coverage of the 2016 general election, explores how Donald Trump won and why his chances were underrated by the most of the American media.

In 2012, President Obama’s advantage over Mitt Romney, although often paper-thin in national polls, was stronger than it appeared for two big reasons. One was that Obama, in stark contrast to Hillary Clinton, was outperforming his national polls in swing states, largely as a result of his popularity in the Midwest. The other is that 2012 featured remarkably few undecided voters: Only about 4 percent of voters went into Election Day not already committed to Obama or Romney. That reduced the chance of a potential last-minute swing. Even if most of the undecideds turned out for Romney, it probably wouldn’t have been enough to vault him past Obama in the swing states.

Just the opposite was true in 2016, and Clinton’s lead was considerably more fragile than it appeared from national polls. Not only was she underperforming in the Electoral College because of the way her demographic coalition was configured (see the first article in this series for more about that) but a much larger number of voters — about 13 percent on Election Day and as many as 20 percent at earlier stages of the campaign — were either undecided or said they planned to vote for third-party candidates Gary Johnson and Jill Stein. Those undecided voters made Clinton’s lead much less safe and they broke strongly toward Donald Trump at the end of the race. Trump won voters who decided in the last week of the campaign by a 59-30 margin in Wisconsin, 55-38 in Florida, 54-37 in Pennsylvania and 50-39 in Michigan, according to exit polls, which was enough to flip the outcome of those four states and their 75 combined electoral votes.

The late shift toward Trump, like other periods of polling instability throughout the campaign, was consistent with a long-term pattern. Historically, the more undecided and third-party voters there are, the more volatile and less accurate the polling has tended to be. The relationship ought to be fairly intuitive: There’s not much a pollster can do when a voter hasn’t yet made up her mind. In 1980, for instance, final polls showed Ronald Reagan leading Jimmy Carter roughly 43-40, with 17 percent of voters undecided or saying they planned to vote for independent John Anderson. Reagan wound up winning in a landslide, 51-41, making for a seemingly massive polling error. But Carter technically didn’t underperform his polls: it’s just that Reagan hugely outperformed his as undecideds and Anderson voters broke his way. A milder example came in 2000, when a relatively high number of voters said they were undecided or planned to vote for Ralph Nader, presaging an upset win in the popular vote by Al Gore (George W. Bush had led by 3 to 4 percentage points in the final national polls). By contrast, the 2004 and 2008 cycles had very few undecided voters and highly accurate polling.

COMBINED UNDECIDED + THIRD PARTY VOTE
CYCLE WITH 100 DAYS TO GO ON ELECTION DAY
1972 9.3%
8.1%
1976 10.0
8.7
1980 25.5
17.3
1984 7.4
4.7
1988 10.8
5.7
1992 19.1
22.1
1996 19.7
13.3
2000 17.8
8.9
2004 6.4
3.4
2008 13.1
3.6
2012 7.6
4.3
2016 18.5
12.5
2016 featured far more undecided voters than recent past elections

2016 estimates are based on FiveThirtyEight “polls-only” national average. Estimates for years prior to 2016 are based on retroactively running the FiveThirtyEight model, except the estimates in 1972 and 1976 with 100 days to go, which are based on a simple average of national polls in July 1972 and July 1976, respectively. (There would not enough polls in these years to properly run the FiveThirtyEight model 100 days in advance.)

FiveThirtyEight’s model accounted for the high number of undecideds, which is part of the reason it gave Trump better odds than other forecasts.1 I’m somewhat perplexed as to why these voters didn’t draw more attention from other modelers or reporters, however, since they were often key to understanding the progression of the campaign. With a large fraction of voters not firmly committed to either candidate — no doubt in part because of the historic unpopularity of both Clinton and Trump — it didn’t take much to move them from one candidate to the other, and so news events had more impact on the polls in 2016 than they did in 2012.

To the extent reporters considered the high number of undecideds, it was often in focusing on how Trump’s share of the vote was low, without noticing that the same was also true for Clinton. For instance, on Aug. 12, The New York Times wrote that Trump had “been waiting for months for a poll in which he cracks 50 percent of the vote against Hillary Clinton in any of his top battleground states” and suggested he had a ceiling on his support:

For a candidate who once seemed like an electoral phenomenon, with an unshakable following and a celebrity appeal that crossed party lines, Mr. Trump now faces the possibility that his missteps have erected a ceiling over his support among some demographic groups and in several swing states. He has been stuck under 45 percent of the vote in Ohio and Pennsylvania for weeks, polls show, while Mrs. Clinton has gained support.

The flaw in the analysis is that by this logic, Clinton also had a ceiling. Although this article was written at one of the high points of the campaign for Clinton in the midst of her convention bounce, she nonetheless had only 44 to 45 percent of the vote in national polls. A significant fraction of voters disliked both candidates, and Clinton never quite drew enough voters into her column to make her lead truly safe, as it might have been if she’d reached closer to 50 percent of the vote.

Was it predictable that those late-deciding voters would break toward Trump? I tend to think mostly not and that the behavior of the late-deciders was instead mainly attributable to an unfavorable news environment for Clinton in the shadow of the James B. Comey letter to Congress and the Wikileaks dumps. But you could argue the point. A once-popular idea called the “incumbent rule” held that undecideds tended to break toward the challenger (as they did toward Reagan in 1980). There are some big flaws with this idea as applied to 2016: It hadn’t hadn’t held up well in recent years (late-deciding voters broke slightly to President Obama in 2012, for instance) and although Clinton was running as a de facto successor to Obama, she wasn’t technically an incumbent.

Still, if the incumbent rule was iffy, so was the idea that Trump had a ceiling imposed by demographics. All it took was for him to win two-thirds of white voters without college degrees to overcome huge problems with the rest of the electorate. (The Electoral College also helped Trump, of course, allowing him to be elected with only 46 percent of the national vote.)

The undecideds were a warning sign that Clinton hadn’t sealed the deal with quite a wide enough coalition of voters, conversely — especially in the Midwest where undecideds were plentiful. In the states that were the biggest upsets relative to the polls — Michigan, Wisconsin and Pennsylvania — Clinton met or slightly exceeded her share of the vote in polls, but Trump beat his by more.


VIDEO: Yea for peaceful transitions of power

Footnotes

  1. Specifically, our model assumed more undecideds meant a higher chance of a polling error, producing either a Trump win or a Clinton landslide.

Nate Silver is the founder and editor in chief of FiveThirtyEight.

Comments