Can betting odds predict elections?
With so much at stake in the US Presidential Election it’s no surprise that a huge amount of attention is paid to any potential predictors of the outcome. Opinion polls have historically been relied upon to predict US Election sentiment, but their track record is mixed. With the rise of online gambling bookmakers are increasingly seen as the Oracles for political markets, but can betting odds predict elections?
What do US Election odds represent?
Probability represents the likelihood of an unknown event — such as the winner of the US Presidential Election — providing a measure of the chance of the potential outcomes. Odds are derived from that underlying probability to answer the question — how often are those events expected to happen? This is why odds are offered as a ratio.
If US Election odds for Hillary Clinton are 1:2 (1/2) this means that over a sequence of elections under similar conditions the bookmakers would expect her to win two out of every three; this gives her a 66% chance. Donald Trump at 2:1 (2/1) would be expected to win one out of three elections — with a chance of roughly 33%.
You can see the latest odds for the US Presidential Election at www.USElections.bet which is polling live odds from a selection of leading UK bookmakers to produce a live % chance of Clinton or Trump winning the Presidential Election.
Probability can be objective — like a coin toss where the probability is known to be 0.5 or a 50% chance on either side — or subjective, where complex situational factors make knowing the exact chances impossible.
Football matches are a good example of subjective probability, and given that they are also very entertaining, everyone has an opinion on the outcome which makes betting — further increasing that excitement — so popular.
Football matches are also ideal for bookmakers to set odds. They can rely on the increasing amounts of reliable data from previous results — known as ‘form’ — to help make their judgement on the probability. They then feed this into a computer model and turn this probability into odds to enable betting, and a margin for their efforts — the cost of a placing a bet.
Why elections aren’t like football matches
The difficulty with setting the odds for political events like the US Presidential Election is that — unlike football matches — the dynamics of the campaigning and election process are too complex to model.
Elections have very little meaningful form because they occurr so infrequently, and the time-lag between elections means the situational factors are always in flux, making comparison meaningless.
Though context can change in football matches — such as injuries — the model can account for this by for example assigning values to players’ influence and these changes tend to be small in impact and isolated.
By contrast, the context for an election can change dramatically at any moment with what is known as an ‘Apocalypse Event’. This presents a real issue for bookmakers. An even bigger issue for bookies is that the key voters in elections are the undecideds.
Voting intent of undecideds cannot be known or measured, it can only be guessed, which is like not knowing which side a striker will be playing for until kick-off.
These challenges don’t deter bookmakers from setting odds. Despite the difficulties in understanding prospective chances UK bookmakers formed a unanimous opinion on the opening market odds for the 2016 US Presidential Election which from the beginning favoured Hillary Clinton’s chances of winning.
The important question is how did they reach that opinion and can the odds and their subsequent movement be relied on as a predictive tool?
Wisdom of the crowd
The principle of collective predictions — in effect what betting markets are — was established by Sir Francis Galton back in the early 1900s when he uncovered the Wisdom of the Crowd phenomenon.
Galton was at a livestock fair watching a competition to guess the weight of a butchered ox. No-one guessed the exact weight but Galton calculated the median of guesses as being within 0.8% of the answer.
“The middlemost estimate expresses the vox populi, every other estimate being condemned as too low or too high by a majority of the voters.” Sir Frances Galton, 1906.
From Cows to Election outcomes
A betting market expresses the vox populi in the same way as the competition Galton witnessed; the best estimate of the underlying probability is reached by the average of odds bettors want to wager at.
For this reason betting markets are generally very good indicators of underlying probability, however, the best estimate approach is not universally accurate, and the circumstances that lead to its failure are especially prevalent in elections, referendums and sentiment based markets.
A Big Enough Crowd
In Galton’s example the accuracy of guesses of the Ox’s weight would be proportional to the number of entrants in the competition. So the number of bets is important, but more so the volume of the money wagered.
Ten £5 bets on Hilary Clinton are not the same as one £1,000 bet on Donald Trump. This can be summed by the phrase ‘Skin in the Game’.
Importantly, the volume of bets cannot be discerned just by looking at advertised odds, and are only known to the bookmaker but it can be inferred through the movement of the odds which react to volume not bet count.
In relation to US Election betting, the concern of a big enough crowd can safely be ignored given the fact it is expected to break all records for political betting. However, the wisdom of the opinion demonstrated through those bets will be augmented by the experience and knowledge of those person placing them.
If it is an experienced US Election bettor, or someone with access to privileged information, it has much more weight that a wealthy casual fancying a large punt. But is that the case here?
It isn’t widely known but the majority of modern bookmakers profile their customers. If a customer shows signs of being sophisticated or consistently winning, their betting limits will be severely limited or their account closed. This practice has the potential to diminish the collective wisdom of that betting crowd, as does diversity.
A Diverse enough Crowd
Diversity is a key factor that can impact the accuracy of election betting odds and is a critical factor in the US Presidential Election. For the average of opinions on the US Election odds to be truly accurate, those opinions have to be representative of the electorate.
Though gambling is heavily regulated in the USA many states offer opportunities to gamble but none can legally offer odds on national or local elections (to avoid manipulation and rigging) and equally it is illegal for US citizens to bet outside of the US.
This should mean that the betting activity driving the US Election odds is being driven by bettors from outside the population who’s opinion won’t count in the actual US Election.
That would naturally cast some doubt over the efficacy of this vox populi because the sample group — the voice being expressed — can’t be assumed to be correlated to the voting behaviour of US Citizens.
The reality however, is that despite large-scale prohibition gambling is still thriving in the US with estimates for sports betting in 2015 put at $145billion — with only a fraction of that within sanctioned onshore operations.
A simple Google search returns a plethora of options of US citizens, but the issue of diversity doesn’t then disappear, it simply takes a different dimension. Are the kind of registered US voters who also gamble in this grey market representative of the electorate as a whole? There is no definitive answer to this but the most reasonable assessment is probably not.
US Election betting as a loss leader
Modern bookmakers make most of their money by attracting sports bettors and encouraging them to play in the casino. The risks associated with casino games are fair easier to manage so, despite the challenges associated with setting US Election odds, bookmakers may offer them and not be concerned if they lose money.
Bookmakers know there is a huge demand and may be seen as a loss-leader to attract in new customers who will then be enticed with more profitable products. The accuracy of the odds in that situation is therefore, less important than their job in acquiring new customers.
Information Cascade & Anchoring
A betting market has to start somewhere. The opening odds — sometimes referred to as the ‘tissue price’ — are decided by the bookmaker. The importance of this opening assessment and how it is calculated are another reason to be cautious of the accuracy of the US Election odds. They are called tissue for a reason, because they are so fragile.
As mentioned above, a bookmaker would normally look to the available form and a suitable mathematical model to help work out the underlying chances. Though there are some reliable patterns to elections — such as a natural inclination for voters to look less favourably on parties serving multiple terms in office — alone, they aren’t enough to calculate a probability.
The next best guide after historical results is Opinion Polls, but there is plenty of reasons to doubt the accuracy of polls, not limited to:
- Statistically insignificant samples
- Unrepresentative samples
- The increasing irrelevance of land lines (for phone polls)
- Biases in the kind of people likely to answer polls
- Context — polls reflect opinion at that time only
Beyond opinion polls the bookmakers will look to financial markets which will build in the financial implications of differing outcomes in national elections, but the specific influence of the election outcome has to be isolated from all the other factors moving the markets and quantified.
What this means is that bookmakers have to choose from a number of potential indicators — none of which are necessarily accurate — as reference points for the assessment of the opening odds.
Importantly bookmakers don’t act unilaterally. Many modern bookmakers are actually getting their odds from a third-party and simply adding their preferred margin, and those that set their own odds will naturally watch what their competitors do. This provides a potential risk of what are known as Information Cascade and Anchoring.
An Information Cascade takes place when inaccurate data is passed onto others who may have reason to disagree, but simply assumes it to be correct.
Given the huge number of stakeholders and vested interests — not just bookies but Hedge Funds and governments — the Information Cascade theory maybe hard to countenance especially with Hedge Funds often commissioning their own private polls, but the sub-prime mortgage bubble of 2008 (the effects of which we are still feeling) had large elements of information cascade. Why couldn’t the same happen in political prediction markets? Our separate Brexit case study gives very good ground to support that idea.
The problem for the Hedge Funds and the bookmakers may have been that the information they sought — the voting sentiment of undecided or first-time voters that proved so crucial — was not known and perhaps could not practically have been known. It may have come down to timing.
Timing — Voter Uncertainty & Intent
The behaviour of bettors is not necessarily equal to that of voters. Bettors are making their opinion known for months even years, but many US Election voters will make up their mind at the point of casting their vote. The betting markets cannot not know what the waverers may do, so when that group is large enough, the odds can be wrong.
This is exactly what happened in the 2015 UK General Election when a huge number of voters decided on the day, so their intent could not have been factored into the odds. This issue of timing is a crucial drawback in betting markets but so is intent.
Bettors may have simply seen a binary question — who will win the Presidential Election Trump or Clinton — but a large number of voters may actually vote in protest of a wider issue of disenfranchisement and disillusionment in the value of Washington. Taken together, the issues of timing and intent could be key omissions from the betting odds.
The Favourite Doesn’t Always Win
The last and perhaps most important comment to make on the accuracy of betting odds in predicting elections is a simple one. Sometimes the favourite doesn’t win. Imagine the outcome of the US Election is decided by picking one ball from a sack of 100 balls. 70 are blue (Democrat/Clinton) and 30 red (Republican/Trump)- these correspond to the current implied chance as of October 1st. It makes it easier to conceive that the underdog can win without the proportions/odds being wrong.
All of this doesn’t mean that betting odds cannot provide a useful indication of potential outcomes of elections, it just means that they should be treated with caution.
More like this
Read more like this at www.uselections.bet.
