Trump still in the fight if polls are as wrong as in 2016

Washington | Donald Trump, after another weekend of frenetic big-crowd campaigning, could be sitting on a greater than one-in-three chance of winning the US presidential election seven weeks from now.

That's the view of an Australian analyst who correctly picked the outcome in every US state in the 2012 election.

But there's a big caveat: that the current polls will be as wrong in 2020 as they were in 2016.

Because if they're accurate, Mr Trump's chances of victory are just 5.3 per cent, says Simon Jackman, chief executive of the University of Sydney's United States Studies Centre, and data researcher Zoe Meers, in a report to be released on Tuesday.

The findings, which imply the race could be far closer than the headline polls suggest, come as Mr Trump takes his campaign for a second term to America's west, aggressively targeting Latino voters at a controversial indoor rally in Henderson, Nevada, on Sunday night (Monday AEST).

The indoor event is the first time Mr Trump has spoken to a large crowd in an enclosed venue since his June 20 rally in Tulsa, Oklahoma, which health experts believe was responsible for a spike in COVID-19 cases in the weeks that followed.

In response to criticism from doctors about the latest rally, Mr Trump's spokesman, Tim Murtaugh said: “If you can join tens of thousands of people protesting in the streets, gamble in a casino, or burn down small businesses in riots, you can gather peacefully under the First Amendment to hear from the President of the United States.”

Joe Biden is also stepping up his campaign and will fly to Florida on Tuesday (Wednesday AEST) for a series of events.

Both candidates are adding events to their schedule at a hectic pace, driven by concerns the race is far closer than most believe.

Bernie Sanders has reportedly become "quietly worried" by Mr Biden's campaign and is urging the Democratic presidential nominee to make clearer his message on how his administration would raise wages, create jobs, cut prescription drug costs and expand health care coverage.

He has also reportedly urged Mr Biden to campaign with progressive favourites such as Alexandria Ocasio-Cortez.

A Fox News poll published on Sunday has Mr Biden ahead with 51 per cent of the national vote to Mr Trump's 46 per cent. However, the democratic contender continues to lag the President on who would best handle the economy.

Those numbers are echoed by the RealClearPolitics average of polls, which has Mr Biden on 50.5 per cent against Mr Trump's 43 per cent.

But across six so-called "battleground" states, the difference between the two men is just 3.6 points, ranging from a 6.3 percentage point lead for Mr Biden in Wisconsin to a near dead-heat in North Carolina.

Polls in swing states systematically underestimated Trump's strength in 2016.
— Simon Jackman, University of Sydney

With memories still fresh of 2016, when a generation of pollsters and political pundits were savaged by reality, Professor Jackman's study includes an estimate of how wrong the numbers were.

Across 13 swing states, he estimates, polls underestimated Mr Trump's support by an average of 3.7 percentage points.

If those polling errors are repeated by the same amount as four years ago, Mr Biden would likely win 305 electoral college votes out of a total of 538. To win the presidency, a candidate needs an absolute majority of 270 votes.

"Polls in swing states systematically underestimated Trump's strength in 2016," Professor Jackman writes.

The shock outcome in 2016 was particularly striking as poll-based forecasts across those same swing states were much more accurate in 2008 and 2012.

In those years, the polls tended to only modestly overestimate the leads of Republicans John McCain, in 2008, and Mitt Romney four years later, against Barack Obama by 0.9 points and 1.7 points respectively.

Only one of 13 swing states was forecast incorrectly across both election cycles.

"If swing state polls are as wrong in 2020 as they were in 2016, then the 2020 election result will be much closer than recent polls suggest, with a 34.9 per cent chance of a Trump win," Professor Jackman writes.

"If the accuracy of 2020 swing state polls is closer to that seen in 2008 and 2012 than seen in 2016, then the current polls imply only a slim chance of a Trump win: just 5.3 per cent."

Among the potential causes of the polling errors four years ago were an unusually large number of undecided voters who broke heavily for Mr Trump in the final week of the campaign; failure to recognise demographic changes; and the fact that college-educated voters were more likely to agree with surveys.

Professor Jackman believes forecasts for the November 3 election should be "considered especially uncertain" due to a number of unusual circumstances.

Fallout from the pandemic, concerns over mail-in voting, strong voter engagement and likely turnout, and a historically low number of undecided voters mean anything's possible.

"Who will actually vote this cycle – and how – and whether their ballot will be considered legitimate and be counted is a live and controversial question this election," he says.