Calling a race in these ultra-partisan times is fraught with risk, but I’m confident we can safely conclude what the final result of the election will be. Joe Biden will win the electoral college by 306-232. The exact reverse of Donald Trump’s margin over Hillary Clinton in 2016, albeit with a popular vote victory of around four to five million.
Biden fell short of a landslide
A comfortable enough victory by historical standards but not a blowout. To call it a landslide, Biden would have at least needed to take Florida.
Which brings us back to the question that has dominated the political forecasting community throughout 2020 – were the betting markets too favourable to Trump? Nobody has been more patronising on the matter than Nate Silver. This from last week wasn’t his most dismissive comment by a long stretch:
Since almost no people have the relevant expertise to build political forecasting models (it takes tons of work and even then is easy to get wrong), political betting markets are basically just a competition over what types of people suffer more from the Dunning-Kruger effect.
? Nate Silver (@NateSilver538) October 29, 2020
Even those of us who shared Silver’s expectation of a bigger Biden victory were taken aback at how touchy the FiveThirtyEight pundit seemed. How about actually running the race before judging the predictions?
Now we have, it turns out the signals from Betfair were superior to the forecasting models. Our electoral college prediction was around 290 to Biden, compared to 350-odd among pollsters.
State favourites prove reliable once again
The favourite won in 49 out of 50 states. Only Georgia broke the trend. None of the pollsters predicted Trump would win Florida or North Carolina.
This trend isn’t out of the ordinary. In 2012, every state favourite won. In 2016, 45 out of 50 did. At the 2019 UK General Election, the favourite won in over 95% of our 650 constituency markets.
Odds factor in potential for volatility
When we consider the Next President odds, I think the sceptics are vindicated again. Yes, Biden backers won. But so too will the traders who laid him at 1.51/2 on the night, with a view to cashing out as his price drifted. He went as high as 5.24/1 in-running.
Such drama seems a fundamental part of betting on US elections. The way results come in inevitably skews the betting in one direction or another, as measuring what is left to come involves a level of expertise. As explained in my trading tips piece, the big urban, Democrat-leaning cities always declare last because they have more votes to count.
I’m sure next time this volatility will be factored in and bettors will be reluctant to hold onto any short-term positions. Even I’m seriously contemplating cashing out first next time. The best plan is to have a full bank, minimal exposure and a clear head to monitor dozens of markets simultaneously. There were big-priced opportunities galore to be found.
Shy Trumper theory stands up
But more than trading motivations, the long-term betting trends for this election revealed scepticism. That Biden was a strong enough candidate, or that the polls could be trusted. Belief in the ‘Shy Trumpers’ phenomenon turned out to be justified.
Whether we give it that label, or call them ‘hard to reach’ or ‘reluctant Republicans‘, it is very hard to dispute now. Trump made this election a lot closer by energising his base like never before. For an incumbent to increase their popular vote by over eight million is quite an achievement, for which there was no conventional signal.
It was enough to defy the polls in Florida and North Carolina. But more telling is that Michigan and Wisconsin were even close, when numerous forecasts had Biden double-digits ahead.
Turnout is a perennial known unknown
This doesn’t ‘debunk’ opinion polling science as some want it to. But it does show the potential for them missing tranches of voters, especially when there’s a big jump in turnout.
The likes of Dave Wasserman, from Cook Political Report, have been flagging Trump’s scope for growth among non-college educated whites in states such as Wisconsin for years. It would therefore be inaccurate to claim pollsters aren’t aware. However it isn’t easy to reach or measure previous non-voters.
Therein lies the biggest advantage of betting markets as a predictor. Polls are, by nature, tied to a rigid, scientific process, which can be flawed if the sample is wrong. When we see massive turnout, as in Europe, they tend to be more reliable but when turnout changes significantly as in recent US or UK examples, they can be blindsided.
In contrast, bettors are making more subjective arguments, based on a wider range of factors including known unknowns. We don’t all agree by any stretch and there is no single judgement. The market is merely the product of supply and demand.
This time, that average assessment emerged looking good. The best way of identifying average market opinion is the electoral college handicap. The par line, trading at even money, was Biden -81.5 – equivalent to 310 votes. He won 306.
Follow Paul on Twitter and check out his website, Political Gambler.