Two weeks from the US presidential election, the average of the surveys on the RealClearPolitics site, Joe Biden continues to lead the outgoing Republican president by nine points nationally. But, in the United States, a candidate can access the White House by winning the majority of large voters at the state level, without winning the popular vote at the national level – as Donald Trump in front of Hillary Clinton in 2016.
This year, six states are deemed likely to tip the victory: Florida, North Carolina, Arizona, Wisconsin, Pennsylvania and Michigan. There too, Joe Biden has the advantage, although he is sometimes within the margin of error, going from +1.4 points in Florida to +7.2 in Michigan.
Why were the polls wrong in 2016?
In 2016, the polls had correctly photographed, the day before the vote, Hillary Clinton’s slight national advance. But they
got it wrong in some of the Midwest hub states who gave victory to Donald Trump, explained to AFP, Chris Jackson, of the Ipsos institute.
Among the causes, he mentions an under-representation in the samples of
white people without a university degree, who eventually came to put a Trump ballot in the ballot box.
Have things changed this time?
Most institutes claim to have corrected their methodology to eliminate this blind spot.
Key states that were not surveyed last time are also the subject of much more and repeated studies.
In addition, pollsters point to great stability: since the spring, Joe Biden has led with an average lead that has never fallen below four points. By comparison, the sawtooth Clinton-Trump curves had crossed twice, illustrating an uncertain race.
Finally, in an extremely polarized country, there are far fewer uncertainties likely to change the game at the last moment.
Are there any hidden Donald Trump voters?
The thesis emerged from Trumpist voters
shy, who, questioned by pollsters, would prefer to keep their choice silent as their champion is controversial.
The polls were wrong last time and they are even more wrong this time, martèle ainsi Donald Trump.
Trafalgar Group, a polling institute close to Republicans that prides itself on a methodology supposed to circumvent this bias, was almost the only one, in 2016, to give Donald Trump a winner in Pennsylvania and Michigan.
But this time, even this pollster gives Joe Biden the advantage in crucial states like Pennsylvania and Wisconsin.
Four years ago, the businessman new to politics was a novelty, and the novelties are still difficult for pollsters to grasp.
Today, everyone has made their opinion on him, there is no longer really any surprise effect around Donald Trump, says Chris Jackson.
What is the margin of error?
The New York Times did the math: even if the current state-to-state polls were as wrong as they were four years ago, Joe Biden would still win by a large margin.
Mr. Biden is closer, in our average, to winning Texas, a Republican stronghold, which would result in a
tsunami in his favour,
that President Trump is not victorious in more traditionally disputed states like Pennsylvania or Nevada, wrote Nate Cohn, the daily specialist recently.
Are things done?
Pollsters and analysts are always careful to remember that voting intentions are not a prediction and that there is a margin of error.
Above all, the campaign is a dynamic. The last presidential election was probably played in the home stretch, according to the news. With 16 days before the poll, the FiveThirtyEight site forecast gave Hillary Clinton an 86% chance of winning, almost like Joe Biden today.
And no election is like the last.
In the United States, voter registration varies widely, making it particularly difficult to predict turnout. Donald Trump invokes the enthusiastic crowds of his meetings to announce a momentum in his favor, but will this be reflected in the polls? The Democratic camp, little mobilized for Hillary Clinton, unpopular candidate who seemed to have won in advance, will he unite behind the most consensual Joe Biden to drive out a hated president?
Finally, one unknown remains: the impact of the pandemic.
Advance and postal voting are at historic levels. We don’t know what effect it will have, explains Chris Jackson, citing factors
difficult to take into account by surveys.