Pollsters were chastised for incorrectly predicting a large Hillary Clinton win in 2016, yet they have they have got it wrong again.
Millions of dollars are spent on looking at who will win a US presidential election, and by how much.
Ahead of this year’s election, Joe Biden was predicted a safe win, with the Democrat leading by up to 16 points a month before, which narrowed down to eight points the day before the election.
It was to be the Democrats’ largest lead on the eve of an election since Bill Clinton in 1996, according to polling analysis website FiveThirtyEight.
US election 2020 live: Follow the latest updates
But, that has not been borne out by the results.
It could be some time before the final handful of states declare but getting the all-important 270 electoral college votes has come down to the wire.
If, as the pollsters predicted, Mr Biden had had a strong lead, he would have reached that number within 24 hours of the polling booths closing.
Hispanic voters
It became clear early on that the pollsters had not hit the mark when Donald Trump won the most-prized swing state of Florida.
There was dancing on the streets of Miami when Donald Trump overtook Joe Biden for the swing state of Florida
Mr Biden had been expecting a three-point lead here but Mr Trump is thought to have won thanks to the state’s large Cuban-American vote, NBC News exit polls found.
Pollsters are felt to have underestimated a shift towards the president by certain demographics, especially Hispanic voters in specific states.
Hard-to-reach voters
Like 2016, another factor was undercounting hard-to-reach voters who tend to be less educated and more conservative – and more likely to vote for Mr Trump.
Nick Beauchamp, assistant professor of political science at Northeastern University, wrote that he is not as convinced by the “shy Trumper” claims of Trump supporters deliberately misrepresenting their intentions to pollsters.
There was a record turnout for the election
Margin of error
Some pollsters are saying the differences are within the three to five percentage points margin of error, and as some states, such as Florida, were predicted to only be won by Mr Biden by a small percentage, that is within the realms of error.
However, Mr Beauchamp said: “But it seems awfully large to me for just that.”
Historic turnout, mail-in voting and online polling
Another factor could be the historic turnout and a surge in mail-in voting, with the sheer numbers making it more difficult to establish who people will vote for.
Polls are also more difficult now as telephone polling is not as effective because voters are less likely to answer phone calls on their mobiles when they do not recognise the number – unlike when landlines were called and the number did not show up.
As a result, pollsters often rely on online panels to approximate random samples, making it harder to target all types of demographics and therefore more difficult to get a good representation.
:: Subscribe to the Daily podcast on Apple Podcasts, Google Podcasts, Spotify, Spreaker
These are all possible factors, but the real reason why the pollsters got it wrong will most likely be figured out by sometime next year.
“A full examination of what went wrong with polls this year is going to take a while,” Doug Schwartz, director of polling at Quinnipiac University in Connecticut told the Boston Globe.
“After the 2016 election, it took six months for the American Association for Public Opinion Research to release their findings about polling errors.
“I would expect a full evaluation of 2020 to take at least as long.”