Trump over performed in Democratic leaning states while Clinton under performed in Republican ones

Exit Polls Reveal a Deeply Divided Nation

The exit polls show that the electorate is very deeply divided in many ways. Men preferred Donald Trump by 12 points while women preferred Hillary Clinton by the same amount. Whites went overwhelmingly (21 points) for Trump, while blacks, Latinos, and Asian Americans were more than two to one for Clinton. Young voters 18-29 supported Clinton by 19 points, while voters 45 and older had a preference for Trump 8-9 points.

Voters with a high school education supported Trump by 6 points but those with some postgraduate study went for Clinton by 21 points. People earning less than $50,000 went for Clinton, but people making more were Trump voters. Clinton won the cities by a huge margin (24 points) but Trump carried small towns and rural areas by 28 points. Protestants and Catholics voted for Trump, but Jews and others went for Clinton. Roughly summarized: Well off, young highly educated urban voters and minorities supported Clinton, and older, poorer, less-educated rural white voters supported Trump. The Republicans probably won’t conduct an autopsy this time because they won, but as its base dies off, winning national elections will get more difficult. Democrats have won the popular vote in six of the past seven presidential elections and 4 years from now, the demographics will be even worse for the GOP.

Third Parties Had a Huge Effect on the Election

It is probably a safe bet that none of Jill Stein’s voters really care much for Donald Trump. Most likely if we had an instant runoff voting so that voters can express their first choice system in which voters can mark a first, second, third, etc. choice, all of them would have put down Hillary Clinton as their second choice. Would it have mattered? Actually, it would have. If the Stein voters had all held their collective noses and voted for Clinton, she would have won Michigan by 0.8% and Wisconsin by 0.1%. That would have changed the Electoral College from Clinton 232 vs. Trump 306 to Clinton 258 vs. Trump 280. Trump would have still won, but by much less.

Polls showed that many young supporters of Sen. Bernie Sanders (I-VT) voted for Gary Johnson, probably because they never read his platform (abolish the minimum wage law, privatize Social Security, get the government out of the healthcare business, etc.). If all the Stein voters and a quarter of Gary Johnson’s supporters had voted for Clinton, she would have carried Pennsylvania, and the score would have been Clinton 278 and Trump 260, leading to President Clinton. If 27% of Johnson’s supporters, rather than 25%, had voted for Clinton, that would have flipped Florida as well, leading to Clinton 307 and Trump 231. At the very least, this result shows that if we are going to have more than two parties running candidates for a single office, we should be using instant runoff voting and also their backup choice if their first choice doesn’t have a majority. In the current system, by voting for their first choice, many voters ended up with their last choice. (V)

The election results brings to mind two questions (at least):

  • Why were the polls so far off?
  • Why did the voters pick someone they themselves said was unfit to be president?

Let’s first tackle the polls. We made an Excel spreadsheet with the relevant data. Here the are:


The first six columns give the state and the actual percentage votes for five candidates from Politico. The D -R votecolumn is the Democratic minus the Republican vote in each state. The D -R polls column is the Democratic polling number minus the Republican polling number in each state. For example, Clinton got 34.9% of the vote and Trump got 61.1% of the vote in Tennessee, so the seventh column is 34.9 – 61.1 = -26.2%. Trump crushed Clinton by 26.2% in Tennessee. But the polls said he would beat her by 9% (D -R) column. She underperformed the polls by 17.2% (last column). The table is sorted on the last column. The rows are colored by who actually won the state.

What stands out is that the top part is mostly red and the bottom part is mostly blue. In other words Clinton did much worse than predicted in the red states, but roughly as expected in the blue states.

That’s the basic data. Now we have to figure out why, which is harder. Several theories come to mind:

  • In the red states, many people lied to the pollsters out of shame or whatever
  • In the red states, there were a lot of hidden Trump voters who decided to vote at the last minute
  • Millions voted between the time FBI Director James Comey announced “more emails” and the time he said “nothing here.”
  • Comey gave Trump momentum and there were not enough good polls in November to measure it

Maybe there are other explanations, but it seems odd that in the 25 states at the top (Tennessee to Pennsylvania), of which 22 are red, the polls were off by 5 points or more, that is, outside the margin of error. In the bottom 26 (Vermont to Hawaii), of which 18 are blue, only three of them were polled badly, that is, off by more than 5 points. Blue staters may jump to the conclusion: “We tell the truth to the pollsters and the red staters lie,” but it is impossible to tell that from the data. Still, the results are striking.

One idea that intrigues us is the lack of quality polls near the end. There were only 31 presidential polls taken entirely in November, and of these, only 12 were by experienced pollsters (defined here as pollsters who also polled in 2012). Only one state (North Carolina) was polled more than once. Thus, one explanation is that the polls were pretty good up to Oct. 28, but after Comey’s announcement, Trump picked up speed and there were too few polls to detect it. The lack of polls may well have been caused by the media’s assumption that Clinton was going to win and didn’t want to spend any money to tell them what they already knew.

The staff at FiveThirtyEight have already talked to several dozen pollsters (who maybe should be spending their time updating their resumes). They offered several theories, some of which echo ours:

  • Some women were embarrassed to admit they were voting for Trump
  • Trump supporters are distrustful of institutions, and pollsters are an institution
  • The blue team blew it on turnout
  • The late-deciding voters, particularly those who abandoned the S.S. Johnson, broke decisively for The Donald
  • Because the media thought it was over, they did not sink money into (expensive) polls in the last two weeks
  • Pollsters pretended to be more certain than they really were, and even tweaked their numbers, so as not to be an “outlier”