THE US presidential election looks like a done deal. Joe Biden leads Donald Trump by 10.5 per cent in the FiveThirtyEight average of polls and by nine per cent in the RealClearPolitics average. The pattern has been clear and steady for several months.
Of course, you will have heard this all before, as in the Daily Telegraph of October 24, 2016 …
‘With only 16 days until the election, two separate polls have given Mrs Clinton a 12-point lead over Mr Trump, with the real estate mogul’s support tanking among key voter groups.
‘An ABC / Washington News poll released on Sunday corroborated the findings of a study published earlier this week by the Monmouth University Polling Institute which showed Mrs Clinton leading Mr Trump 50 per cent to 38 per cent in a four-way contest with two minor candidates.
‘Mrs Clinton’s long-held advantage with the fairer sex has increased to 20 points following allegations that Mr Trump sexually harassed multiple women, according to the most recent ABC / Washington Post poll.’
And the experience of four years ago makes even those most desperate to see the end of President Trump hedge their bets. The Guardian’s analysis of the polls contains a long disclaimer: “We must caution that the polls – particularly some swing state polls – severely undercounted Trump supporters in 2016. We are not certain, despite assurances, that they have corrected this. Additionally, they may be over-counting Democratic support (more people may say they will vote for Biden than actually turn out).”
Part of the problem is that the polls look too good for Biden. The actual national turnout lead in recent Presidential elections has been two per cent for Hillary Clinton, four per cent and seven per cent for Obama, 2.5 per cent for Bush, 0.5 per cent for Al Gore, 8.5 per cent and 5.5 per cent for Bill Clinton, eight per cent for Bush Senior.
One has to go back to Ronald Reagan’s destruction of Walter Mondale in 1984 (by 18 per cent) to find a bigger winning margin than that predicted for Joe Biden.
The polls are confounded by all the non-polling evidence pointing to a Trump win. ‘It’s the economy, stupid’, said one of Bill Clinton’s election strategists – and it is indeed rare for any government with a track record of economic success to lose an election.
A Gallup poll published early this month found 56 per cent of Americans believed they were better off than four year ago. This is the best-ever figure recorded by Gallup at the end of a president’s first term; the last sitting president not to be re-elected, George Bush Senior, achieved only 38 per cent on this measure.
Gallup found 89 per cent of voters thought the candidates’ position on the economy was either extremely or very important to them, besting terrorism and national security (83 per cent) and the response to Covid-19 (77 per cent).
Joe Biden generates negligible interest when he ventures out of his basement to a public engagement. It’s estimated that since May 1, some 240,000 people have attend Trump rallies, but for Biden it’s only 84 (sic) when one excludes staffers and the press.
A Biden / Harris event in Arizona (a key swing state) attracted zero interest, not even on the sidewalks outside, though their campaign claimed it wasn’t meant to be a public event.
Rather than trusting polls with their dubious track record, one might prefer to rely on the election forecast model developed by Professor Helmut Norpoth at Stony Brook University, New York State.
Professor Norpoth runs a ‘Primary Model’ to forecast the winner of the presidential election by looking at how long a party has been in power and candidates’ performance in the party primary elections. Analysing data back to the first primaries (held in 1912), the model has correctly predicted the election winner 25 out of 27 times.
The only occasions on which it has picked the wrong winner were in 1960 and 2000. In both instances, the candidates’ primary results were very close and so were the presidential elections; John F. Kennedy may have only won due to chicanery in Illinois and Texas, and George W Bush lost the popular vote but sneaked in thanks to hanging chads – punch-card ballot papers not counted by voting machines – in Florida.
The most extraordinary feature of Norpoth’s model is that he announces his prediction about six months before polling day. In March 2016, Norpoth stated Trump would win the election with 87 per cent certainty. This March, Norpoth gave Trump a 91 per cent chance of winning re-election, predicting the most likely electoral college split would be 362-176.
Norpoth completely ignores polls, but it’s difficult to discount the evidence of so many polls saying Trump is going to lose, and lose bigly. Pollsters are quick to state that they have made ‘adjustments’ to their methodology since the last election, in particular addressing the under-sampling of non-college educated voters in their polls, a fault they acknowledge from 2016.
However, tricksy voters don’t like to be boxed and categorised; having been told for four years that association with Trump is like having something nasty on your shoe and being told by Trump that the mainstream media are ‘fake news’, it is easy to understand why Trump supporters might not reveal their true voting intentions to pollsters.
And there’s some evidence this might be the case. A Harris poll in September found Biden with a two per cent national lead, but when people were asked who they thought their neighbours were most likely to vote for, Trump had a seven per cent lead.
The Gallup poll that found 56 per cent feeling richer than four years ago also found the same percentage believed Trump would win, with only 40 per cent believing Biden will be President. Perhaps those unwilling or ashamed to say they are supporting the Bad Orange Man are projecting their voting intentions on to others.
One polling body is bucking the trend and finding Trump in the lead. The Democracy Institute’s October poll put him one per cent ahead, down from three per cent ahead in September. As the electoral college arithmetic hurts Biden (because Democrat support is heavily concentrated in California and New York), a small national lead to Trump means the Donald is almost certain to win.
The Democracy Institute correctly called the 2016 election and the EU referendum. It recognises the phenomenon of the shy Trump voter and tries to ask questions in a manner to counter this, in part by using recorded questions and push-button rather than verbal responses (to reduce the embarrassment of admitting to another human being that you’re voting for Trump).
It found that 87 per cent of Biden supporters said they were ‘comfortable with relatives, friends, co-workers knowing how you vote’, but only 22% of Trump supporters said the same. The polling panjandrums at FiveThirtyEight and RealClearPolitics don’t include the Democracy Institute polls in their figures. The final Democracy Institute poll comes out on November 1.
The betting markets are often turned to as a reliable indicator of a result. Trump’s current odds of winning vary from 11/8 to 2/1. This is hardly overwhelming; I had a couple of small bets on Trump in 2016, at 11/4 and 4/1 respectively.
Betting odds reflect the amount of money placed on an outcome, not the number of bets placed, something which can fool the unwary to draw the wrong conclusions. Remain was the bookies’ favourite to win the EU referendum in 2016, but the number of individual bets on Leave was greater.
I can’t find data on betting numbers for the presidential election. However, if you enjoy a little gamble and are prepared to trust Professor Norpoth’s model, you can get 14/1 on Biden getting less than 179 electoral college votes. Fancy a bet?