How reliable are election betting markets?
With so much at stake in the upcoming US Election it is no surprise that a huge amount of attention is paid to any potential predictors of the outcome. With the validity of opinion polls increasingly questioned, will US Election betting markets assume that role?
Opinion polls have been the mainstay for measuring election sentiment, but their track-record in recent years has meant that they are increasingly treated with caution.
Opinion polls were catastrophically inaccurate for three of the biggest election events in recent years - the UK’s 2016 EU Referendum, the 2017 UK General Election, and the 2016 US Presidential Election.
With so much doubt cast on the usefulness of what was the standard barometer for pre-election sentiment, betting markets - in various forms - are increasingly filling that role. But are they any more reliable, and what does that mean for bettors looking to profit from the uncertainty?
Why the 2020 US Election Will Break Betting Records
It has been estimated that betting on the 2020 US Election will eclipse any political event in history. Disregarding the unique context within which it takes place, Donald Trump on his own is a huge draw.
A political maverick who divides opinion, Trump is predictable in his unpredictability. Yet, this year’s POTUS has so many other dynamics beyond merely the liking or loathing of the great orange one.
The choice of his opponent has in itself been a fascinating saga, with the DMC searching its soul to decide who is best-placed to take on Trump, veering between a more radical left-leaning candidate (Bernie Sanders) or a safer centrist option. It appears they will end up with the latter, with Biden nailed on to be crowned at the July Convention.
But is Sleepy Joe the right choice? He has two failed POTUS runs behind him, is 77 with two major brain surgeries behind him, leading to his lucidity being questioned. He has a dangerous raft of sexual allegations hanging over him, with the prospect of more skeletons finding their way out of the closet, while his son Hunter, is a constant source of embarrassment.
Then of course we have the COVID-19 pandemic, its impact on the economy, the rising trade feud with China, Middle East tensions - specifically Iran - and Russian influence.
The list of themes that might decide the election are endless and will make the POTUS debates box-office viewing. So what do the US Election betting odds at Cloudbet tell us about the potential outcome, and how reliable are betting markets as opposed to polls?
US Election Betting Odds - The Current Situation
Cloudbet is offering a host of exotic US Election markets, but the main binary option is Who will be the next President of the USA? This is where all the action is.
As of May 12th the current odds are as follows:
If you aren’t familiar with betting odds, we can turn that into something much more useful - the implied percentage chance winning. Here is how that is calculated.
Calculating Implied Chance from the Odds
Odds represent the likely frequency of an unknown event. In this case voting sentiment in the US Presidential Election.
You can turn these odds into an implied probability – in other words, the chance the odds are based on – by doing the simple sum (1/decimal odds)*100.
For Trump's current odds that means (1/1.81)*100 = 55.3% chance.
Here are the implied chances for all the available options at Cloudbet:
Donald Trump - 55.3%
Joe Biden - 44.44%
Hilary Clinton - 4.4%
Andrew Cuomo - 2.3%
Michelle Obama - 2%
Mike Pence - 1.6%
Elizabeth Warren - 1%
Notice that the chances add up to more than 100%. That is because they include a margin. Learn about betting margins from this article.
Should we trust the predictive accuracy fo election betting odds?
The question that anyone unfamiliar with betting will ask is ‘Why should we put faith in betting odds?’ To answer this we have to go back to a livestock fair in Plymouth in 1906 attended by Sir Francis Galton who, in observing a competition to guess the weight of a butchered ox, uncovered the Wisdom of the Crowd phenomenon.
Wisdom of the Crowd is the process by which dispersed, independent opinions coalesce around the most accurate assessment of an unknown variable - in that case, the accurate weight of the ox - in our case, the winner of 2020 US Presidential Election.
Wisdom of the Crowd is the process by which dispersed, independent opinions coalesce around the most accurate assessment of an unknown variable
No-one guessed the exact weight of the ox, but Galton calculated the median of guesses as being within 0.8% of the answer, commenting “the middlemost estimate expresses the vox populi, every other estimate being condemned as too low or too high by a majority of the voters.”
A betting market acts as a vox populi by the measuring willingness of participants to risk their money predicting the outcomes of events based on quoted probabilities, presented in the form of odds.
Just as in the competition Galton witnessed, the best estimate is reached by adjusting the odds in proportion to the aggregate volume of wagers bettors are willing to risk on either side.
So US Election betting odds represent the vox populi of the participating bettors, and the same for every other type of betting market, or prediction of an unknown which applies to the price of bitcoin or guessing the number of sweets in a jar. But they are not without their limitations, which are important to understand.
The limitations of betting odds as predictors of election outcome
Though betting markets are very good indicators of underlying probability, they aren’t perfect, and going back to Galton’s experiment we can illustrate why in certain circumstances they should be treated with caution.
The accuracy of guesses of the ox’s weight would be proportional to the number of entrants in the competition -- the more guesses, the more accurate the median.
Equally, from a betting perspective the number of bets is important, but not all bets are equal in terms of their influence on the odds.
Ten €5 bets on Trump are not the same as one €1,000 bet on Biden. A shorthand way of describing this is ‘Money Talks’. Betting markets react to the volume wagered rather than the volume of wagers. So they are not a proxy for votes, they are a proxy for sentiment as represented by the willingness to risk money.
So if money moves markets, should that opinion always be trusted? If it is an experienced political bettor it has much more weight than a wealthy casual bettor placing a large punt, but you will only see the output - the odds - and not know the volume of bets or know anything about who is behind them.
This is true for all betting, but politics is niche compared to mainstream sports, and for all the reasons discussed here, is unlikely to be a market a sharp bettor would get involved in.
The importance of diversity
In the case of US Election betting, there is no certainty that those betting are drawn from a wide audience representative of the electorate. For a start, Cloudbet don’t accept US customers, and the same is true of the majority of bookmakers worldwide.
So is the behaviour of bettors for POTUS markets outside the US a useful proxy for actual participating voters?
This will be the first US Election since the relaxing of betting regulations in America, so there is an argument that the betting markets will be more robust than in the past, as actual voters will be helping form the betting odds. And the odds at Cloudbet don’t exist in isolation from bookmakers elsewhere; there may be some difference in the odds, but if there is enough liquidity - which should certainly be the case if, as expected, records are broken - then markets should be efficient, and therefore exist within a relatively small deviation.
Even if US bettors are helping shape the market for the next POTUS, are they representative of US voters? This is unlikely to be true. Betting tends to be indexed heavily towards men aged between 18 and 40, which clearly doesn’t represent the whole electorate.
Taking this argument further, even if those US bettors truly represented the US electorate, given the way that the election works, what the betting is actually gauging is the likelihood of Trump or Biden winning in a handful of key swing states. Even if every citizen of California were to place a bet on the outcome, would that provide a more accurate assessment of the outcome in Florida, given California has consistently voted Democrat every year since 1992?
This illustrates that the way elections and referendums are decided is far more complex than would you expect. Rather than a first-past-the-post system, which would be easier to predict, the US uses the anarchic Electoral College system.
The crowd is only as wise as its participants and by corollary, the accuracy of betting markets is determined by the domain knowledge of those participating.
Information & disinformation
The idea of knowledge is crucial in the Galton example. If the crowd guessing were children, not farmers, the outcome would have been very different. The crowd is only as wise as its participants and by corollary, the accuracy of betting markets is determined by the domain knowledge of those participating.
Knowledge depends on information. In a Premier League game for example, the result is unknown but the majority of the relevant information is quantitative (previous results, in-game data, goal expectation, weather, attendance) and in the public domain, and so a reasonably accurate assessment of the underlying probability of the three outcomes can be quickly reached, especially given that this data is fed into complex models which quickly improve the market’s efficiency.
This means that what are called the ‘opening odds’ are pretty accurate, but as time progresses those very informed bettors who have access to proprietary information (not in the public domain and derived from private networks or complex models) place bets and the market moves to maximum efficiency close to kick-off. This is why ‘closing odds’ are always more accurate than opening odds.
But the US Election isn’t like an EPL game, because it is so hard to take a quantitative approach. If you built a model (like the EPL example) what inputs would you decide as relevant in determining the winner, and how would they be weighted?
Should you use data from previous elections, even when they happen so infrequently and the context is completely different -- as are the candidates as well as the electorate itself, with new voters entering the system as they reach 18 and a similar proportion dying.
With all these challenges it is easy to see why polls have been the go-to yardstick, and increasingly, focus groups in marginal/swing areas, where the results are closely guarded.
We are in an era of ‘big data’, so you might think that a big enough computer could crack this nut, and though Google Search Results and trending Twitter hashtags can certainly provide insights into what people are thinking and saying, digital media has actually been used to shape public opinion rather than to measure it.
The influencing of public opinion via social media and disinformation is a massive can of worms in relation to this discussion. Essentially, the ability to influence public opinion via ‘fake news’, another input into any election betting assessment that is hugely problematic to assess, unless you simply assume all sides are equally capable of doing it, in which case there is no bias.
As mentioned above, opinion polls and betting markets called three of the biggest elections in recent memory completely wrong. In isolation, one of those results wouldn’t necessarily prove anything other than an unlikely result; in any circumstance the favourite will sometimes lose. However, three biggest upsets in such close proximity suggest something else.
Given the absence of solid data with which to base predictions of election outcomes, the market tends to focus on a set of assumed truths e.g. women/Hispanics won’t vote for Trump; or that the undecideds will still to the status quo (in relation to Brexit). Without being challenged, these ideas become unchallenged truths, fuelled by the media, and can often prove false.
That process is known as an information cascade, and was especially relevant for the 2016 US Election where so many preconceptions about the electorate’s view of Trump were completely smashed.
This is evident from the fact that he was considered a massive underdog to win in 2016 throughout the entire election process and even once voting was underway. The market only turned when the result started to come in from Florida. Up to that point, the prevailing narrative was that Clinton was gonna win comfortably, and clearly that narrative was misguided.
The relevance of timing
Based on the points raised so far, there is plenty of reason to feel that predicting elections is fraught with danger and we haven’t yet discussed the issue of timing.
In the three events highlighted as examples of how out of touch polls (and betting) have come with genuine voter sentiment, timing was a crucial factor.
The betting markets were unable to accurately account for the floating voters who made their mind up on the day, and the first-time voters who were an even greater unknown. Both groups were likely more susceptible to the influence of messages closer to voting.
Nowhere is this idea more apparent than in the admission of Vote Leave Machiavelli, Dominic Cummings, who used Data Science techniques to identify messages most likely to push people to vote leave in the 2016 EU Referendum. In his own words:
‘For example, one of the few things about advertising which seems logical and has good evidence to support it is — try to get your message in front of people as close to the decision point as possible. That’s why we spent almost the whole campaign testing things (via polls, focus groups, online etc) then dropped most of our marketing budget in the last few days of the campaign’.
The Prostitute Factor
Leaving the best to last, one of the reasons that bookmakers themselves are wary of election betting markets is something rather bluntly known as the ‘prostitute factor’. It refers to the danger that at any time a skeleton might fall out of a candidate’s closet and completely derail their chances.
It doesn’t necessarily have to relate to sex but was especially relevant for Donald Trump, with multiple allegations of affairs, sexual misconduct and misogyny.
What was so surprising was the allegations themselves, which became so numerous as to lose impact, but that his voters didn’t care - a similar story with Boris Johnson. This doesn’t necessarily apply going forward, scandals can come in many forms and have devastating consequences on a candidate’s chances of winning. But because there are no clues ahead of time, they can make election markets a potential minefield for bookmakers who are used to markets moving by degrees, not wild swings.
No One Really Knows
The 2017 UK General Election is a great example of how badly out of whack opinion polls can be. Take this excerpt from the relevant Wikipedia page:
“Opinion polls had consistently shown strong leads for the Conservatives over Labour. From a 21-point lead, the Conservatives' lead began to diminish in the final weeks of the campaign. In a surprising result, the Conservative Party made a net loss of 13 seats despite winning 42.4% of the vote (its highest share of the vote since 1983), whereas Labour made a net gain of 30 seats with 40.0% (its highest vote share since 2001 and the first time the party had gained seats since 1997). This was the closest result between the two major parties since February 1974.”
We will never know definitely why the outcome was so different to the pre-election polling and betting markets. Theresa May’s U-turn on the so-called ‘dementia tax’ is considered to have played a large part, but that is purely conjecture, and in essence all elections are.
This is actually good news for bettors. Bookmakers don’t like uncertainty and struggle to model the odds accurately. If you are interested in betting on the US Election, or any other for that matter, step outside of your echo chamber, ignore opinion polls, absorb as many diverse opinions as possible and keep your popcorn handy for results night, as no one really knows what the outcome will be.