Polls' performance in election scrutinized

Several political polls had predicted that Democratic challenger Joe Biden had a big lead over President Donald Trump heading into Election Day, but as the votes were tallied Tuesday, it became clear the gap was smaller, with the race headed to a nail-biting finish.

The country woke up a day after the election with no declared winner as votes still were being counted in several states.

The major polls that make election predictions now are facing scrutiny over what they got right and what they got wrong.

The muddled polling results follow the debacle in the 2016 US presidential election, when pollsters were notably wrong.

Most predicted then that Democratic candidate Hillary Clinton would win. But it was Trump who prevailed.

Pollsters vowed they would get it right in 2020 — but did they? They also had Biden ahead by a wider margin than they had Clinton.

RealClearPolitics, which aggregates polls, and FiveThirtyEight monitor some of the biggest political polls in the US.

FiveThirtyEight analyzes several influential polls by NBC News/Wall Street Journal, YouGov, Reuters/Ipsos, John Zogby Strategies/EMI Strategies and others to get its results. The Rasmussen poll is also well respected and came the closest to predicting the 2016 outcome.

On Wednesday, Rasmussen appeared to have made the most accurate 2020 forecast. It does not use humans to gets its results; instead, it uses a pre-recorded voice.

In the days before Election Day, polls by RealClearPolitics showed Biden with a seven-point lead. FiveThirtyEight said he was ahead by eight percentage points.

NBC News/Wall Street Journal said Biden had a 10-point lead. YouGov said Biden was nine points ahead. John Zogby Strategies/EMI Strategies said Biden was up by six points. Each poll has a margin of error of 2 percentage points.

But in Florida, polls by FiveThirtyEight had Biden with an 8.4 percent lead. When the state began reporting results Tuesday, however, they showed Trump and Biden would be in a much tighter race. Trump ultimately won the state with 51.2 percent to Biden's 47.8 percent.

FiveThirtyEight got Florida wrong. Rasmussen got it right; it had predicted that Trump would win the state by more than one percentage point.

Jeffrey M. Berry, a political science professor at Tufts University, told The Boston Globe: "There appears to have been a broad and systematic error in predicting the election outcome ... it's clear that the pollsters underestimated the high Republican turnout."

In crucial states like Pennsylvania, which has 20 of the 270 electoral votes needed to win the White House, RealClearPolitics predicted that Biden was up by 1 percent. FiveThirtyEight said he had a five-point lead. On Wednesday evening, Trump was leading although thousands more ballots had to be counted.

In Wisconsin, RealClearPolitics showed Biden with a 6.5-point lead. FiveThirtyEight said he was eight points ahead. On Wednesday morning, Biden had a 5 percent lead. The former vice-president ultimately was declared the winner in Wisconsin by midafternoon, with 49.6 percent of the vote to Trump's 48.9 percent. The Trump campaign immediately requested a recount.

One theory on the poll discrepancies is that some Trump supporters may be "shy" to admit whom they are voting for in a climate of political division with Democrats.