Donald Trump’s victory in the US presidential election four years ago brought into question as never before the reliability of opinion polls. Can they be believed this time around?
What do the polls say?
With 16 days to go before the November 3 election, Democrat Joe Biden is ahead of the Republican president by 9.0 percentage points nationally, according to polling averages from the RealClearPolitics website.
But in the United States, candidates win the White House not through the popular vote, but with the Electoral College.
In 2016, Trump lost the popular vote to Hillary Clinton, but won enough states to garner the electoral votes needed to become president.
This year, six states are seen as key to winning the White House: Florida, North Carolina, Arizona, Wisconsin, Pennsylvania and Michigan.
But if the polls are correct, Biden also has the advantage there, although he is at times within the margin of error, ranging from 1.7 percentage points ahead in Florida to 7.2 in Michigan.
Where were the errors in 2016?
The polls on the eve of the vote correctly predicted a slight national lead for Clinton, but “the place where the polls missed were in some of those Midwestern swing states” that Trump eventually won, Chris Jackson of Ipsos Public Affairs told AFP.
He said under-representation within polling samples of white residents without college degrees who voted for Trump was among the causes.
Most polling institutes say they’ve corrected their methodology to preclude such mistakes this time around.
Battleground states under-polled last time have been surveyed much more closely and more often.
Beyond that, pollsters note consistency: Since the spring, Biden has been ahead with an average lead which has never fallen below four percentage points.
As a comparison, the Trump-Clinton polling lines crossed twice, signalling an uncertain race.
Finally, in a country extremely polarized, there are far fewer undecided voters susceptible of altering the contest at the last minute.
Are there reticent Trump voters?
Some feel that there are Trump voters reticent to tell pollsters they prefer him given the controversy that surrounds the president.
“The polls were wrong last time, and they’re more wrong this time,” Trump has said.
Trafalgar Group, a polling institute favored by Republicans that seeks to employ a methodology to account for the possibility of reticence, had been one of the few in 2016 to predict Trump winning Pennsylvania and Michigan.
This time, however, even they give the advantage to Biden in crucial states such as Pennsylvania and Wisconsin.
Four years ago, the businessman and political newcomer was a novelty, and such candidates are always difficult for pollsters to assess.
“Everyone sort of has an opinion about him now, so there’s just not quite the same level of surprise in Donald Trump,” said Jackson.
But what if …?
The New York Times has calculated that, even if the current polls, state by state, are as wrong as they were four years ago, Biden would still win.
“Mr. Biden is closer in our poll average to winning Texas, which would get him over 400 electoral votes, than President Trump is to winning in traditional battleground states like Pennsylvania and Nevada,” the paper’s Nate Cohn wrote recently.
Do uncertainties remain?
Pollsters and analysts are still careful to note that voters’ intentions are not a prediction and that there is still a margin of error.
Campaigns can be dynamic, with the last presidential election probably decided in the home stretch.
With 16 days to go in 2016, the FiveThirtyEight site gave Clinton an 86 percent chance for victory, nearly the same as Biden now.
In the United States, voter registration varies enormously, which makes it especially difficult to predict turnout.
Trump points to enthusiastic crowds at his rallies to argue that momentum is on his side, but will that translate at the ballot box?
Will Democrats who were not overly enthusiastic for Clinton, who was viewed initially as having won in advance, line up behind a middle-of-the-road Biden to chase out Trump?
And what effect will the pandemic have?
“We have mail-in voting and early voting which are going to be at historic levels,” said Jackson.
“We don’t know the effect that’s going to have. There’s a lot of really complicating factors that are entering into it that are the kind of things that are hard for polls to account for.”
By Francesco Fontemaggi