What the New York Times Pollsters Got Wrong About Swing State Polls

Democrats shouldn’t freak out—just yet

Michael Greiner
Nov 14, 2019 · 5 min read

Since last week, Democrats I have been speaking with are freaking out. After 2016, polling shows that Democrats care more about beating Trump than anything else. His poor approval rating and the growing support for impeachment have given his detractors hope.

But then, on October 25th, the New York Times completed a poll that showed Trump statistically tied with or leading Democratic candidates in the key swing states of Michigan, Pennsylvania, Wisconsin, Florida, Arizona, and North Carolina. These six states, of course, were the swing states Trump won by the narrowest margin in 2016. If Clinton had won two or three of them rather than Trump, she would be president now and things would be a lot different for sure.

But we live in a world where that did not happen. I live in the swing state of Michigan, in the famous Macomb County. This is the county where pollster Stan Greenberg conducted his famous research, designating my county as the home of the Reagan Democrats, working class voters who abandoned the party of the New Deal for Ronald Reagan’s conservative revolution. This county has been called a “bellwether,” and it voted overwhelmingly for Trump in 2016. It was one of three counties in the country that the Cook Political Report designated as being most responsible for electing Trump.

All I can say is that I’m sorry. I did my best, but I was overwhelmed. The result is what we all know happened. Trump won Michigan by 0.2 percentage points, just 10,704 votes in a state where third party candidates received 250,902 votes.

And here we are.

I must say that the polling results referenced by the New York Times really doesn’t match what I am seeing on the ground. While there is surprisingly strong Trump support remaining, it is clearly not what it was three years ago. I was cautiously optimistic, until I saw this poll.

The New York Times poll is a bit of an outlier. At the same time they were conducting their poll, Emerson College, a highly rated pollster according to FiveThirtyEight, was also polling. While the New York Times interviewed 501 voters, Emerson’s poll interviewed 1051.

Their results are strikingly different. The New York Times found that Trump was tied with Joe Biden, that he led Elizabeth Warren by six points, and that only Bernie Sanders narrowly led Trump. In contrast, Emerson found all the Democrats leading Trump by significant margins: Sanders by 14, Biden by 12, and Warren by 8.

The polling website Real Clear Politics’s average for the key midwestern states of Michigan, Pennsylvania, and Wisconsin projects Trump losing to any one of the three leading Democratic candidates. And in North Carolina, Biden and Sanders are leading Trump, and Warren is within 0.2 percentage points.

So what’s going on here? Should Democrats be freaking out?

First, let me say that Democrats can take nothing for granted. Trump does have an Electoral College advantage. In fact, some analysts have predicted that Trump could win re-election while losing the popular vote by an even larger margin. And the betting markets are still giving Trump a 42% chance of re-election. So it’s close.

The problem is the unequal distribution of Trump opposition. Trump opponents tend to be concentrated in a few states, including California, New York, Massachusetts, Maryland, Illinois, and Hawaii. But remember, it doesn’t matter whether a presidential candidate wins a state by one vote or a million, he or she still gets the same number of Electoral College votes: all of that state’s allocation. So Trump’s route to victory involves winning a bunch of lightly populated Republican states and winning a couple of swing states like Michigan, Pennsylvania, Wisconsin, Arizona and Florida by small margins. That’s how he did it in 2016, and it could happen again.

G. Elliott Morris, the polling expert at The Economist has looked closely at this poll, and is less worked up about it than the New York Times’s Nate Cohn. First, he pointed out that all the results are within the margin of error. Even though the result may show Trump or the Democrat slightly ahead, the actual sentiment the Times was trying to measure is just as likely to be exactly the opposite.

Furthermore, Morris points out that Elizabeth Warren is significantly less well-known than the other two candidates. This means that as she increases her name recognition, she has plenty of room to go up in the polls. In fact, the more surprising result is how well she is doing against far better-known candidates.

But there is one problem that has been virtually ignored about the Times polling: it interviewed all registered voters. As a result, it included in its results people who will vote, and people who won’t.

Out of their sample, the New York Times found that about 15 percent of its respondents have not made up their minds. That diverse group, although heavily male, is designated by the Times as “persuadable voters.”

That characterization is incorrect. As I pointed out in another article, in a race like this, there are virtually no persuadable voters. People who turn out to vote are the people who have already chosen one candidate or another. Being undecided means one thing: you don’t vote. At the end of his analysis of the results, Nate Cohn correctly points out that “a disproportionate number of persuadable voters tend to be low-turnout voters as well.” Thus, the people he is spending so much time focusing on are unlikely to vote at all.

Indeed, Nate Cohn incorrectly defines and overstated the importance of “persuadable voters” when he wrote that “ . . . their votes effectively count twice: a voter who flips from one party to the other not only adds a vote to one side, but also subtracts one from the other side’s tally.” To set the record straight, undecided voters only count for one vote, because they were not on either candidate’s side from the beginning. By definition, they are “persuadable voters.” It is only when you convince one of your opponent’s supporters to defect to your side that you can subtract one vote from their total and add one to yours, effectively counting the vote twice.

Finally, Cohn attempts to analyze these voters’ attitudes from the results. But as I pointed out, the Times only sampled 501 voters in Michigan. Of those, only 15 percent are “persuadable.” In other words, we are talking about 75 survey responses. With such a small number of responses, the margin of error will be so large that to draw any conclusions from this data will be meaningless.

Do Democrats have reason to worry about 2020? The answer is clearly yes. But this is no time for panic. Headlines like, “One Year From Election, Trump Trails Biden but Leads Warren in Battlegrounds: signs that the president’s advantage in the Electoral College has persisted or even increased since 2016” may sell newspapers, but they overstate the importance of what they are reporting.

After the overwrought reporting of 2016, you would think our journalists might have learned a thing or two. Instead, it looks like all of us need to maintain a healthy level of skepticism about everything we read, even if it’s in the New York Times.

Our Human Family

Conversations on achieving equality

Thanks to Clay Rivers

Michael Greiner

Written by

Mike is an Assistant Professor of Management for Legal and Ethical Studies at Oakland U. Mike combines his scholarship with practical experience in politics.

Our Human Family

Conversations on achieving equality

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade