In Depth

US election: why did the polls get it wrong again after the shock of 2016?

Biden denied landslide as Trump outperforms predictions in almost every state

Democrats heading into the presidential election had cause to be confident, with polls giving Joe Biden a commanding lead in all of the country’s key swing states.

But memories of Hilary Clinton’s shock loss in 2016 came flooding back as the first results trickled in on Wednesday, with Donald Trump taking Florida with 51.2% of the vote share.

Unlike in the election four years ago, the Democrats are still on the verge of victory, as Biden picks up vital “blue wall” states in the midwest that Clinton lost. Yet pollsters had predicted a far smoother path to the White House for Biden - so why did they get it wrong again?

How far off target were the predictions?

According to FiveThirtyEight’s poll aggregator, the consensus among major national polls on 1 November was that Biden had an 8.4% lead over Trump across the country’s 50 states. 

If correct, Biden would have enjoyed a landslide victory. But a very different “reality sunk in the minute Florida started reporting its election results”, sociologist Salvatore Babones writes in an article for the Sydney Morning Herald.

Before the election, Emerson College had Biden up by six points in Florida, while Quinnipiac University and Monmouth University gave the Democrat a five-point lead. Those predictions of a Biden win were backed up by Reuters/Ipsos and NBC News/Marist polls, which both forecast that he would take the Sunshine State by four points. 

When the votes came in, however, Trump had won the state and its 29 electoral college votes by 3.5 points.

Elsewhere in the country, “polls downplayed Trump’s support even more flagrantly than they did in Florida”, says Wired. Biden held a seven-point lead in Wisconsin and an eight-point lead in Michigan, according to CNBC/Change Research

But with 99% of votes counted in both Wisconsin and in Michigan by end of play on Wednesday, he was up by just 0.7 and 2.7 points respectively.

In Ohio, which Trump won by just under 450,000 votes in 2016, polls said Biden was “within less than a point”, the magazine adds. “He’s currently down eight.”

In 2016, a saving grace for the pollsters was that while their prediction that Clinton would win was incorrect, their polling on the popular vote ended up being fairly accurate. 

RealClearPolitics called an average Clinton lead of 3.3 percentage points over Trump across the US on the day of the election, and she ended up winning the popular vote by 2.1 points. 

But the size of the polling errors this time round appears to be far greater.  

Biden is currently leading the national popular vote by 2.4 points, which translates into about 3.5 million votes. But FiveThirtyEight’s aggregation of polls on the night before the election gave Biden a lead of more than three times those totals.

However, in FiveThirtyEight’s defence, the site did point out that the polls suggested the election would be “a fine line between a landslide and a nail-biter” - one prediction that has proved correct.

Why did the pollsters get it wrong again?

Opinion polling is a science of data and modelling. But when the data proves faulty, the result may be what The Atlantic describes as a “catastrophe for American democracy”.

While pollsters “claimed that they had learned the lessons from their 2016 mistakes”, they were “actually more wrong this time around”, data journalist Mona Chalabi writes in The Guardian

Speculating on why the data did not reflect reality, Chalabi suggests two explanations: the pollsters may have “found it even harder to track down and speak to 1,000 adults who accurately represented 240 million voters, or Trump voters were even more reluctant this time to tell a stranger their preferred candidate”.

A similar explanation is put forward by sociologist Babones in the Sydney Morning Herald, who points out that one pollster - Rasmussen - got Florida right, after also being the only major company that came close to predicting the outcome in the 2016 election.

“Instead of using human poll-takers like the other major polls, Rasmussen uses a pre-recorded voice,” Babones writes. “Human beings are social animals, and that’s just as true when they’re answering a telephone survey as when they’re arguing on Twitter.

“In an America where Trump supporters are routinely called ‘racist’ (or worse), it’s no surprise that many of them prefer to keep their political leanings to themselves.”

In other words, in a US election that saw the highest turnout in more than a century, shy Trumpists appear to have caught the pollsters flat-footed once again.

Recommended

Has Johnson learnt his lesson on overpromising?
Boris Johnson bumps elbows with a first responder as she gets her vaccination
Behind the scenes

Has Johnson learnt his lesson on overpromising?

Tories plot leadership revolt against Johnson over lockdown
Boris Johnson leaves number 10 Downing Street for PMQs.
Behind the scenes

Tories plot leadership revolt against Johnson over lockdown

Government to ditch EU rules on UK workers’ rights
An employee works on an engine production line at a Ford factory
Behind the scenes

Government to ditch EU rules on UK workers’ rights

Why Italy’s government is on the verge of collapse
Matteo Renzi holds a press conference
Getting to grips with . . .

Why Italy’s government is on the verge of collapse

Popular articles

How scientists developed a Covid-19 vaccine in record time
A nurse prepares a dose of the Oxford/AstraZeneca Covid-19 vaccine
In Focus

How scientists developed a Covid-19 vaccine in record time

Stalin-themed kebab shop closes after one day
Tall Tales

Stalin-themed kebab shop closes after one day

Who are the richest people in the world?
Elon Musk
In Focus

Who are the richest people in the world?

Free 6 issue trial then continue to