They’re celebrating in the streets of Washington DC and New York, and President-Elect Joe Biden opened his victory speech with the words “The people of this nation have spoken and delivered us a clear victory…” It’s starting to feel like the promised Democratic landslide happened after all.
The pollsters, after three or four days of intense anxiety and abuse, are basking in this last-minute reprieve. If, as is now likely, Biden ends up with 306 Electoral College votes to Donald Trump’s 232, that looks a lot like the comfortable win they predicted — and the final 4-5% popular vote advantage will be only a few points off. OK, they got Florida wrong, but basically the polls were right, right?
Like what you’re reading? Get the free UnHerd daily email
Already registered? Sign in
Of all people, I should be pushing this latest version of events. I predicted on election day that even if the polls were as wrong as they were in 2016, Biden would still win comfortably. So am I now vindicated too?
Sadly not. The word ‘comfortably’ is not just a detail, it’s the whole story. And there has been nothing comfortable about the past week.
The overall story that the polls told was that this election was not even going to be close — as YouGov’s Doug Rivers put it with admirable candour in our pre-election interview, “either we’re badly off or this is not going to be close.” Trump obviously didn’t win, so the praise being heaped on Robert Cahaly, the pollster who forecast a Trump victory, is equally misplaced — but the President came a lot closer than he was supposed to, and that’s important.
One of the many facts Democrats have deployed for years when discussing the shock 2016 result is that as few as 70,000 voters across crucial swing states handed Donald Trump his victory. Poll watcher Nate Silver’s explanation for why he gave a Trump victory such a low probability at the time is that it was a genuine “low probability event” — a number of states that were incredibly close just all happened to flip into Trump’s column. It was kind of a fluke, in other words.
But contrary to the forecasts, we’ve just been through another nail-biter. As results stand, if just 25,000 voters across Wisconsin, Arizona and Georgia had chosen Trump instead, Biden could never have reached the magic 270 electoral college votes. Commentators were right to observe that 2016 was weirdly close, but then so was 2020.
In 2012, the last “normal” presidential election, only 4 out of 50 states were even close, with final victory margins of under 5%; in 2016 that number more than doubled to 10, and in 2020 it was still 8. In 2012, there was only one state that had a final margin of under 1% (FL), but there were 4 such states in both 2016 (MI-NH-PA-WI) and 2020 (AZ-GA-PA-WI). If these four states had edged in the other direction, in both elections the other candidate would have won. No polling company can realistically measure a state-wide difference that small — in both elections those four states could easily have gone either way.
Once again, the polls systematically underestimated the Trump vote across the country: states such as Ohio, Iowa and Texas were widely predicted to be close when they weren’t, and Wisconsin was predicted to be in the bag for Biden when it should have been too close to call. Looking only at states where 100% of the votes have been counted, the YouGov 50-state model has an average of over 5 points margin error in the Democratic direction, and in deep Red states such as Arkansas and Mississippi the error increases to over 8 points.
The reason this matters is not to give pollsters another shellacking — they have been doing their best to get it right — but to reckon with the profoundly important fact that, once again, there was a slice of the electorate that simply wasn’t showing up in the official measures. It is analogous to the wider political situation: just as these voters aren’t properly represented in opinion polls, they still aren’t well understood in the mainstream conversation. The fact that the 2020 result will probably end up with an identical Electoral College margin to the 2016 result, with just a few states flipped narrowly the other way, is a sign of how little we have moved on in the past four years in terms of understanding this hidden part of the population.
The New York Times’ graphics showing whether each county moved more Republican (red) or Democrat (blue) shows near-stasis this year compared to the strong red shift in 2016.
Pollsters will soon start coming out with technical postmortems which, while no doubt useful, must not be allowed to confuse the bigger picture. Sure, the demographics were different this time — this time it wasn’t poor white men that they miscounted so much as certain sections of the Latino and Black populations, alongside the college-educated and suburban white women.
But the fact that their blind spot has shifted does not make it less concerning. The two most common explanations — “shy Trump” voters not admitting the truth to pollsters (or perhaps even themselves), and the (in my view more salient) fact that less engaged members of each of these demographic groups are less likely to end up taking a survey in the first place — both point to the same bigger problem. There is a section of the population that is almost invisible to the mainstream and incomprehensible to polite society. They are as “unheard” as ever, and it must be our mission to understand them better.
The opposite instinct, dangerously tempting for some American liberals, is to sit in moral condemnation of the over 70m of their compatriots who voted for Donald Trump. This attitude requires us to believe that 70% of the citizens of Wyoming, and 40% of the population of New York, are bad people — the idea is itself morally repugnant and will only lead to future electoral failure.
A more useful objective for Democrats would now be to focus, in a way they failed to do in the past four years, on what about Trump’s offer made it so attractive to so many good people, despite his glaringly obvious personality flaws. This is their best bet to taking their party out of margin-of-error territory next time round. Happily, in his remarks so far, Joe Biden has emphasised the need for healing and speaking for all Americans, but there will be many in his administration who take a sharper view and it will be a daily struggle to resist them.
I concluded my pre-election piece by saying that, if Trump ended up surprising everyone and actually winning re-election, “it would mean we have to rethink the basic epistemic foundations of our public conversation”. We didn’t reach that rather hyperbolic-sounding moment, but those foundations took another knock last week. The close result means that Donald Trump and his millions of supporters were not entirely wrong, even though they weren’t right either, to dismiss the opinion polls as part of a wider mainstream stitch-up; it means that his low approval ratings, endlessly referred to throughout his presidency, were probably also off by a similar margin, and that the whole framing of the race was misconceived.
It is a powerful reminder that we should be careful before writing off those who reject the establishment way of thinking alongside the sophisticated tools used to reinforce that reality. It also serves as a warning to the many wielders-of-charts to be more humble in their presentation of them (note to Nate Silver: “Fuck You, we did a good job” is not quite the right tone).
At points during results night, I had the eerily familiar feeling that Trump had gone and done it again. At those times it almost felt uncanny, as if he had been able to reach otherwise hidden voters using intuitions that the pollsters and pundits simply didn’t possess. In the cool light of day, with Biden’s electoral college victory now clear, that sensation is receding with all the speed of a bad dream. A few more months and perhaps the whole Trump era will start to feel the same way. But we mustn’t let it — the issues and people he spoke to remain all too real, and must now be taken seriously, not forgotten.