A global pandemic shutting down entire countries for months. Riots scarring cities across America with the worst urban violence in half a century. A sitting American president getting infected with a deadly virus, leading to his hospitalization less than a month before a presidential election. And a Supreme Court justice dies months before that election.
In many ways, 2020 has been an unprecedented, even bizarre, year for American politics.
And yet the horse race of this year’s presidential election has been surprisingly dull.
Looking at the polls, and the narrative they’ve spawned in much of the media, there’s almost a feeling of déjà vu.
That’s been especially true after President Trump’s polling numbers tanked in the late spring, after having received an initial ‘rally around the flag’ boost during the early stage of the coronavirus pandemic.
Since then, Trump’s approval rating numbers have remained deep underwater, and he has consistently trailed Biden by a wide margin in nationwide polling, and by smaller – though significant – margins in key battleground states.
And just as in 2016, media outlets and political analysts have pointed to Trump’s consistent polling deficit as proof that he is the clear underdog, with only a slim chance of winning, while his Democratic opponent could potentially win in a historic landslide.
2016 Redux?
So just how bad are Trump’s polling numbers?
Right now, according to the RealClearPolitics aggregate of polls, Biden leads Trump in nationwide polling by an average of 9.8 points (51.9% to 42.1%) when third party candidates aren’t included, and by an average of 7.6 points (51.2% to 43.6%) when third party candidates are included.
To be clear, that’s not a good place for a candidate to be in, particularly an incumbent president seeking reelection.
It’s also worse than where Trump was polling four years ago - but not catastrophically so.
On this date four years ago Clinton led Trump by an average of 6.0 points in a head-to-head matchup, and by 5.4 points on average in a four-way race.
Trump ended up losing the popular vote to Clinton by about 2.1 points, while winning states in the Electoral College worth 306 votes to Clinton’s 232.
In other words, ceteris paribus, Trump’s deficit at this point in the race would appear to put him on track for a loss in the popular vote of around 4-6 points.
A four point loss in the popular vote would make it harder for Trump to win in the Electoral College, but it certainly doesn’t make it impossible, though a six-point popular vote deficit would be difficult to overcome.
But are all other things equal in the polling this year? Or has something changed?
Are the polls better this year, or worse?
Ask most mainstream political analysts why Trump is yet again considered the underdog this year, after his surprise win four years ago which defied most polling in battleground states, and you’ll likely hear one of two answers.
First, Biden’s lead is wider and more consistent this year than Clinton’s was in 2016.
Second, pollsters have learned from their mistakes four years ago, and have adapted their methods to better draw out shy Trump voters who eluded pollsters in Wisconsin, Pennsylvania, and Michigan last time.
There’s some truth to both points.
Regarding the latter, many pollsters in 2020 are trying new techniques to avoid the misses that dogged polling in 2016, when nearly all polls consistently showed Pennsylvania and the Upper Midwest safely Democratic.
Some pollsters are now weighting their poll samples not just by race, but now also by education, to compensate for the oversampling of college-educated voters.
Others are now using the ‘neighbor question’, developed by the Trafalgar Group, which asks respondents how they think most of their neighbors will vote – a method for allowing reticent voters to hint at who they actually will vote for, without actually saying so over the phone.
Thus, when pollsters show Biden leading by about seven points in Michigan, six points in Wisconsin, nine points in Pennsylvania, and nine points in New Hampshire, it appears that Trump’s chances of eking out a victory this time are slim – even slimmer than last time.
Add to that Biden’s leads in traditional Republican states like Arizona and Georgia, and it seems to confirm the narrative of an impending Biden victory.
Social Desirability Bias in the era of Cancel Culture and Black Lives Matter
But a minority of experts have pushed back on claims that the polling in 2020 is more accurate than it was in 2016.
Some, like the chief pollster for the Trafalgar Group, Robert Cahaly, argue that polls this year are even more likely this year than in 2016 to miss shy Trump voters.
In a recent appearance on The Political Trade podcast, Cahaly argued that many right-of-center voters are disinclined to share their views openly with polls conducted by live callers – even though such polls have long been considered the gold standard of political polling.
This ‘social desirability bias’ – the tendency of respondents to tell pollsters what they think they want to hear, rather than what the respondents actually think – has gotten progressively worse since 2016, Cahaly argues, making it even more difficult to draw out shy Trump voters.
In addition, Cahaly claims that long-form polls, which present a battery of questions – sometimes many dozens – to respondents cause the sample to skew away from people who don’t follow politics closely, with highly-opinionated, strongly partisan respondents being disproportionately likely to actually complete such surveys.
Other pollsters, however, like YouGov's Doug Rivers, have disputed claims of a large hidden Trump vote hiding from live caller polls.
“Hundreds of thousands of people have signed up to take surveys and we interview them repeatedly over time," Rivers told the UnHerd podcast last month. "What that enables us to see is the same people at different points in time and whether they've changed or not. What we've seen is about 7% of the Trump voters now say they will vote for Biden rather than Trump, and there's only a 1% offsetting flow in the other direction. So that's a big deal."
But unlike automated calls, internet-based polling methods, like those used by YouGov, rely on self-selecting groups of participants, who seek out and volunteer to join the pool of possible respondents. While the samples used in internet polling can be weighted by demographic groups like race, age, and education, the participants tend to be far more politically-oriented and active than the general population - or even the voting public as a whole.
State polling in 2016 vs 2020
Keeping in mind the two conflicting narratives on the state of polling in 2020, how do the state polls look right now, in comparison to where they were in October 2016, and what, if anything does that tell us about this year’s election?
RealClearPolitics, which tracks Trump’s performance in battleground states and compares it to his position in 2016, has found that the president is now polling an average of half a point ahead of where he was overall in those key states four years ago.
But looking at averages of polls for multiple states provides a fairly weak metric for measuring the state of this election. Taking each state’s polling gives a better idea of how to compare this season’s polling with that of 2016.
Four years ago today, Clinton led Trump in eight of the top 11 battleground states, according to the RCP aggregate of polls.
She held a 7.5-point lead in Pennsylvania, a 7.3-point lead in Michigan, a 6.8-point lead in Wisconsin, a 6.7-point lead in New Hampshire, a 2.9-point lead in Florida, a 2.6-point lead in North Carolina, a 2.3-point lead in Ohio, and a 1.3-point lead in Nevada.
Trump led in just three states, two of them traditionally Republican strongholds: he had a 4.7-point lead in Georgia, a 3.3-point lead in Iowa, and a razor-thin 0.7-point lead in Arizona.
On election day, Trump carried nine of those eleven states, winning all but Nevada and New Hampshire, which he lost by 2.4 points and 0.3 points respectively.
Notably, Trump outperformed his polling numbers at this point four years ago in every state with the exception of Nevada, where Clinton performed one point better than the polling had indicated in mid-October.
In the latest RCP averages for the 2020 election, Trump again finds himself trailing in most battleground states. Biden leads in ten of the eleven states, narrowly trailing Trump in Georgia.
On the face of it, Trump’s position in the battleground polls is even worse than in 2016, when he led in three states in mid-October.
But when we look at the margins of Biden’s lead compared to Clinton’s in states where both led at this point in the race, we see that Trump is outperforming his October 2016 numbers in five of the eleven states – Michigan, North Carolina, Ohio, Pennsylvania, and Wisconsin.
Poll bias?
State polls, as we saw above, were poor predictors of the eventual outcome in 2016, with Trump trailing in mid-October in eight of the 11 top battleground states, while he ended up winning in all but one.
Putting aside the question of whether the polls are more or less accurate this time than in 2016, what happens if we apply transfer the results of 2016 to the current polls – in other words, what happens if we subtract the difference between the 2016 polls and the actual results from the current polling?
Nate Cohn of The New York Times builds such a model, using the NYT’s aggregate of polling, rather than RCP’s.
If the polls are as wrong in 2020 as they were in 2016, the model shows, Trump still is likely to lose must-win battleground states, though Biden’s lead is drastically reduced.
While the NYT aggregate shows Biden up nine in Michigan, eight in Pennsylvania and Wisconsin, five in Arizona and Florida, and four in North Carolina, with the 2016 polling bias removed, Biden trails in North Carolina, Wisconsin becomes effectively a tossup, and Biden’s lead in the remaining states is cut to between 2-4 points – still a difficult hurdle, but not an impossible one for Trump to overcome.
But Cohn’s model compares current polling – in October 2020 – to the bias in the final polls in 2016, on the eve of the election.
What happens if we apply the same idea to polling from this point in 2016 – comparing the difference between the RCP average of polls in mid-October with the actual results and then subtracting that from the current average of polls?
Now, Trump leads in eight of the 11 battleground states, losing only one state (Pennsylvania) which he won in 2016, with a second state (Arizona) a virtual tossup, with Trump leading by a tenth of a point.
Another tight election
None of this should be construed as meaning that the polling numbers, which currently show Trump trailing almost across the board, actually mean he is likely to win.
Rather, the point is to highlight the level of uncertainty in this election.
With election models giving precise predictions, based on polling, regarding the chances of a Trump victory (slim to extremely slim, depending on the model) about equal to those of a Biden landslide, it’s important to remember the difficulties in accurately polling a diverse country of 330 million people, particularly in such a chaotic period.
Both candidates have a very realistic chance of winning next month’s election, and each clear paths to 270 electoral votes.
Instead of trying to gauge the odds at this point for either candidate, it is more useful to instead consider the hazards facing each campaign.
Danger signs for Trump
One of the most high-profile issues in this election cycle since the outbreak of the pandemic has been mail-in voting.
Putting aside the practical issues now being debated, mail-in voting presents a potentially serious threat to the Trump campaign.
For decades, Republicans have tended to perform better in low-turnout elections. Registered voters generally lean more Democratic as a group than actual voters, with more non-voters favoring Democrats than Republicans.
As distance voting and early voting has increased in recent years, however, turnout has risen.
The push to make mail-in voting easier and more widespread could lead to a significant increase in turnout, fueled primarily by people who in previous years did not bother to vote. As the actual voting population grows, Republicans could find themselves facing a blue wave of new voters.
Perhaps even more worrisome for the Trump campaign is the sharp drop off in support from white voters, and in particular, older white voters.
Exit polls show Trump won voters over 65 by even points, 52% to 45%.
However, recent polls now show Trump trailing Biden among voters over 65 by a wide margin.
A recent IBD/TIPP poll shows Biden leading Trump 51.5% to 44% among voters over 65.
Trump is also performing relatively poorly among whites – a demographic group which backed him 58% to 37% in 2016.
While the president has substantially improved his standing with black and Hispanic voters, recent polls show support from whites slipping to 53%, compared to 42% for Biden.
Elderly white voters, who strongly backed Trump in 2016, appear to be moving towards Biden following the outbreak of the coronavirus pandemic. If that trend continues, it will represent the loss of one of the Republican Party’s key demographic groups.
A third factor is the relative stability of Biden’s lead over Trump.
While Clinton led Trump for nearly all of the 2016 election cycle, the margin of her lead was not nearly as stable as Biden’s.
Clinton’s lead shrank or expanded significantly during the course of the election season, influenced in part by the developments in the investigation into Clinton’s unauthorized use of a private email server.
During the past year, Biden’s lead in head-to-head matchups with Trump ranged between 4.0 points, according to the RCP average, and 10.2 points.
Over the same period in the 2016 election cycle, however, Clinton’s lead varied more, and was less consistent. At its peak, Clinton led Trump by 11.2 points, but trailed him twice, by a maximum of 1.2 points.
More importantly, her lead dipped significantly numerous times. On no less than seven occasions, Clinton’s lead fell below two points during that 12-month period, while Biden’s never fell below four.
The fourth and final factor is the incumbent advantage.
Trump ran in 2016 as an outsider with no record in politics. Today, he has nearly four years in government under his belt.
Typically, incumbents – in particular incumbents who succeed in winning reelection – receive a boost from their incumbency, outperforming their performance in the first election. In the past century, only one president – Barack Obama – has managed to win reelection by a narrower margin than in the previous election.
The fact that there is no apparent incumbency boost for Trump – at least in the polls – should be worrisome to the Trump campaign.
Warning signs for Biden
If Trump should be concerned by his relatively poor polling numbers with whites and elderly voters, Biden should be very concerned by continued erosion of minority support.
Clinton suffered in 2016 both from a significant drop off in black turnout, and from increased support among black and Hispanic voters for Trump in comparison to 2012.
That trend appears to be continuing in 2020, with black support for Trump doubling in polls from about 8% in 2016 according to exit polls, to the mid-to-high teens this election.
Among Hispanics, Trump is now polling in the mid-to-high thirties, compared to the 27% he received in 2016.
Should these trends continue, Biden will not only struggle in battleground states like Florida, Ohio, Georgia, and North Carolina, it could even put seemingly safe Democratic states like Virginia and Colorado into play.
The coronavirus pandemic could also have unforeseen effects on the election, depressing turnout among Democratic-leaning voters, rather than increasing it.
Polls have shown Republicans disproportionately planning to vote in person on election day – by a wide margin – while Democrats are far more likely to say they will vote by mail. According to a Pew Research Center study released last Friday, 63% of registered voters who say they plan to vote in person on election day back Trump, compared to just 31% who say they back Biden. By contrast, 69% of respondents who said they plan to vote by mail back Biden, compared to just 27% who back Trump. In-person early voters favor Biden over Trump by a margin of 55% to 40%.
But if large numbers of those mail-in votes don’t materialize – either because a voter doesn’t actually bother to register for it (in states which require it), fill out the form, print it, and mail it in; or because large numbers of votes are disqualified over technicalities – the effect of the virus on the election could actually be a net negative for Democratic voter turnout.
Finally, Republicans have registered more new voters than Democrats in a number of battleground states, cutting into the traditional Democratic edge in voter registration numbers.
While Democrats have invested far more heavily than Republicans in ad spending, the GOP has significantly bolstered its ground game in 2020.
In North Carolina, for instance, the number of registered Democrats has fallen significantly since 2016, dropping 6.2%, CBS News reported earlier this month, while the number of registered Republicans rose by 3.5%.
In Pennsylvania, the number of registered Republicans increased by 3.7% since 2016, compared to a 1.5% drop in the number of registered Democrats, cutting the Democratic lead in the number of active registered voters by about 27%.
The number of registered Democrats in Florida increased by 5.8% since 2016, while the number of registered Republicans rose by 9.6%.