With just fourteen days left until Election Day, the opinion polls indicate that Joe Biden has a substantial lead over Donald Trump, and some of them are pointing to a blowout. But with memories of Trump’s surprise victory in 2016 still painfully fresh, many Democrats won’t believe he has been defeated until they see Marine One ferrying him from the South Lawn for the final time. In these circumstances, making predictions is a game best left to those hardy enough to make a profession of it. But if the polls can’t be entirely relied upon, they do provide us with the most complete snapshot available of the electorate. To get a read on what the latest polling data are telling us, I spoke on Monday with two veteran polling experts, Ruy Teixeira, a senior fellow at the Center for American Progress, and Charles Franklin, a professor at Marquette University Law School.
I asked Teixeira, who is the co-author of the influential book “The Emerging Democratic Majority,” from 2002, about the national polls, which show Biden with a big lead: on Tuesday morning, he was ahead by 8.6 percentage points in the Real Clear Politics poll average, and by 10.3 points in the FiveThirtyEight poll average. In breaking down this lead, there are many factors to consider. Back in June, I pointed out that, as the coronavirus continued to spread, Trump was losing older voters, a key demographic that he carried handily in 2016. The latest polling confirms this trend. In a New York Times/Siena College poll that was released on Tuesday, likely voters aged sixty-five and up favor Biden over Trump by ten points: fifty-one per cent to forty-one per cent.
Teixeira has been closely following the racial and educational demographics. From this perspective, he said, the biggest difference this year, compared with 2016, is that Trump appears to have lost a good deal of support among white voters, a group he carried by twenty points four years ago, according to exit polls. “It’s very clear what’s going on,” Teixeira said. “On the one hand, based on current trends it appears Biden is probably going to do about as well as [Hillary] Clinton did among Hispanics and Black voters—Asians are a little harder to get ahold of because they are such a small group. The big shift is among white voters, who’ve gone from the big margin they gave Trump in 2016 to being almost closely contested in some polls. That’s ridiculous for a Republican. You can’t win that way.”
The polls indicate that a good deal of Trump’s shrinking lead among white voters is a result of his alienating college-educated whites, particularly college-educated white women. Teixeira readily conceded that’s an important development, but he also pointed to a less ballyhooed development: Trump’s declining support among non-college-educated white voters—the group usually thought of as his base. In 2016, according to the exit polls, Trump carried non-college-educated whites by a whopping thirty-seven percentage points. In the latest national poll from NPR/PBS NewsHour/Marist College, his lead over Biden in this group was only eleven percentage points. Other surveys exhibit a similar shrinkage. Teixeira shared with me some results from the latest Nationscape survey, an ongoing large-scale study of the electorate that researchers at U.C.L.A. and the Democracy Fund, a charitable foundation. Taken last week, the survey estimated Trump’s lead among white non-college-educated voters at fifteen per cent. That’s a fall of eighteen points compared with the survey’s estimates from November, 2016.
Some polls don’t show such a dramatic drop. For example, the new Times/Siena College survey puts Trump’s lead among white respondents without a college degree at twenty-three points, but even that represents a substantial drop from 2016. The reason these declines are so important, Teixeira said, is that the white non-college-educated demographic is so large: according to his estimates, it will constitute about forty-one per cent of the 2020 electorate. The share of white voters with a college degree will be considerably smaller: thirty-one per cent. Moreover, as many college-educated suburban voters have abandoned Trump, he has become ever more reliant on his working-class base. “This is potentially the death knell for him,” Teixeira said. “It was always true that unless he did as well or better among these voters, he was very likely going to lose the election. He isn’t doing well. He’s doing worse, and he’s doing worse by large margins.”
Ultimately, of course, the election will be decided in the Electoral College. Four years ago, Trump overcame a roughly two-point deficit in the popular vote by gaining narrow majorities in Michigan, Pennsylvania, and Wisconsin. That enabled him to win three hundred and four votes in the Electoral College, and his best chance of getting reëlected lies in following a similar path. To pull off a repeat, he would need to turn around polling deficits in a number of Sun Belt states that he won in 2016, including Arizona, Florida, and North Carolina, and then win at least one of the three Rust Belt states. If he carries two of the Rust Belt states, he could conceivably afford to lose Arizona or North Carolina.
Over this past weekend, Jen O’Malley Dillon, Biden’s campaign manager, warned Democrats not to be complacent. In the battleground states, she said, the “race is far closer than some of the punditry we’re seeing on Twitter and on TV would suggest.” That is certainly true. On Tuesday morning, the R.C.P. poll averages from the battleground states indicated that the contests in Arizona, Florida, and North Carolina were all within three points or so. Biden’s lead in Pennsylvania was 3.8 points. In Wisconsin, it was 6.2 points.
Charles Franklin runs the Marquette Law School poll, which has shown Biden leading among likely Wisconsin voters by four to six points since May. The most recent Marquette poll, which was carried out after the first Presidential debate, showed Trump trailing by five points. When I asked Franklin what he attributed this stability to, he pointed to the “overwhelming polarization” that Wisconsin experienced during the Obama years, and which “only got stronger in the Trump years.” In the Marquette poll, Trump’s approval rating has hardly ever fallen below eighty-five to ninety per cent among self-described Republicans, while among Democrats it has rarely reached ten per cent. With most Wisconsin residents already holding strong opinions, there are fewer potential swing voters. So the polls don’t move around as much as they once did.
That isn’t to say there hasn’t been any movement. Compared with 2016, Franklin said, two groups in particular have moved away from Trump: white women without a college degree and white men with a college degree. “Interestingly, those were also the two groups that in 2018 shifted to pretty strong support for the Democratic candidate for governor, and more so for the Democratic candidate for the Senate,” Franklin said. “If we are looking for where swing voters are these days, it’s not the two polar opposites that get so much attention”—white men without college degrees and white women with college degrees—“it’s these two groups in the middle.” The Marquette polling provides a more granular picture of the shift away from Trump among non-college-educated voters that Teixeira highlighted. In Wisconsin, at least, this seems to be overwhelmingly a female phenomenon. Franklin said his data indicate that white men without college degrees are still supporting Trump by more than thirty points. But white women without college degrees are now leaning slightly to Biden.
Surprisingly, the Marquette poll didn’t show a big shift in the senior vote after the onset of the coronavirus pandemic, or during the past month and a half, when the state has experienced a big spike in COVID-19 cases. Franklin said the likely reason is that many elderly Wisconsin residents had already turned against Trump. “You get a lot of commentary attributing Trump’s loss of support among seniors to COVID,” he said. “That’s not the case with us. That loss of support predated COVID in our data.” Among other voting groups, though, Trump’s handling of the pandemic may well be hurting his standing. Franklin uses a statistical model to gauge the impact that various factors are having on voter choices. He said the model indicates that partisanship and Trump’s over-all approval rating have had the biggest effects, but his handling of the coronavirus is the third-most significant factor—ahead of his handling of the economy and his response to the Black Lives Matter demonstrations. “The statistical model suggests that the coronavirus may be having a small but significant effect on voter preferences,” Franklin said.
One of the reasons I was interested in speaking with Franklin is that the Marquette poll, although well-respected in political circles, is one of the state surveys that showed Clinton running well ahead of Trump in Wisconsin on the eve of the 2016 election. (The margin was six points. Trump ended up winning the state, albeit by fewer than twenty-three thousand votes.) This year, the Trump campaign insists that the pollsters are underestimating his support again. Franklin conceded that this possibility can’t be ruled out, but he also pointed to a number of steps that he and his colleagues have taken to minimize the chances of another shocker. “Having gone through an experience like that, you certainly don’t take things for granted or be cavalier about it,” he said.
Unlike some polls, the Marquette survey has weighted its samples by education level since 2012—which eliminates one big potential source of error. To account for the possibility of a surge in new voters, which the Trump campaign is relying on, the Marquette poll includes people who aren’t currently registered but say that they intend to register by the election. (Wisconsin allows same-day registration at the polls.) In another effort to pick up hidden Trump support, Franklin compares the intentions of previously registered voters with new registrants, and checks if they are more pro-Trump. (At the moment, they aren’t, he said.) Also, he and his colleagues eliminate some people as likely voters that other pollsters include. Right now, though, this group actually leans more to Biden—nine points—than the over-all electorate does, Franklin said.
