Why And When Political Prediction Markets Fail.

In their article "The behaviour of betting and currency markets on the night of the EU referendum" Linton and Auld examined the efficiency of the Betfair and pound/dollar markets as the results of the United Kingdom European Union membership referendum were announced. They concluded that - " Not only was the currency market informationally inefficient, so too was the betting market, and both markets violated semi-strong EMH on the night of the vote.....Our results suggest that market participants suffered a behavioural bias as the results unfolded. It appears that traders and gamblers simply could not believe that the UK was voting to leave the EU. " Two effects were very cleary at work on the night: The Dunning–Kruger effect and the The Zollman Effect.

Despite considerable evidence to the contrary the notion that people who trade political betting markets have access to privileged information continues to dominate the narrative on social media. Even the so-called experts fall in to the trap. Professor Leighton Vaughan Williams Head of Economics Research and Director of the Betting Research Unit and Political Forecasting Unit at Nottingham Business School famously tweeted the following tweets in 2017 just before the Conservative Party lost their majority in a snap general election.

The notion of informed money backing the Tories at 1/5 on smacked of a man who had not only drank the Kool Aid, but who had succumbed to wishful thinking, confirmation bias and motivated reasoning. They still quote him on Bloomberg and refer to him as a betting expert. And they still put forward the notion that political betting markets are full of rational, sophisticated traders who upadate their beliefs in response to new information and who collectively provide a quantifiable assessment of the wisdom of the crowd. Sure it makes for a great story, boosts egos and keeps the bookmakers in profit. Alas there is no evidence whatsoever of a wisdon of crowds effect at work in the political betting markets.

In the run up to the 2016 U.S. Presidential Election Professor Leighton Vaughan Williams even took to Twitter to tweet to Nate Silver about the supremacy of the betting markets.

Leighton V.Williams: Betting markets outperform polls, models and experts.

The communications theorist Paul Watzlawick wrote that a "phenomenon remains unexplainable as long as the range of observation is not wide enough to include the context in which the phenomenon occurs. Failure to realise the intricacies of the relationships between the matrix in which it takes place, induces one to attribute to his object or study certain properties the object may not possess." Betting markets ascribe probabilities to political events - but, more often than not these probabilities are being scrambled together on the back of incomplete information and are actually nothing other than the product of guesswork and wishful, emotional laden thinking.

One hour after polls closed in the Brexit Referendum, the implied probability, on Betfair, the world's largest betting exchange, that the UK would vote to remain in Europe was 94%! Many people accordingly went to bed, believing that the vote was a done deal.

The most surprised, come the next morning, were the FX traders, who awoke at 4am to take their taxis into work, to find that the UK had actually voted for Brexit. Many of them had put their faith in the betting markets, and bet that the pound would rise significantly on the back of a vote to stay in the European Union.

The following two tables reveal the extent to which sterling volatility eased off as a remain vote appeared to look more certain (the second table is a snapshot of the Betfair EU Referendum betting market with sixty five million pounds traded and hardly a penny looking to get on the Brexit outcome.)

The stance that had been adopted by the majority of FX traders, represented a serious case of overconfidence - they had fallen into the age old trap of believing that they knew more than they actually could have known: or, putting it another way, they had been duped into believing that betting markets were omnipotent and contained some additional inside information about the outcome of the EU referendum that the polls had missed, and they chose to slavishly follow them. (Kahneman; "When they come together, the emotional, cognitive and social factors that support exagerated optimism are a heady brew.")

It was not the first time in the recent past that the betting markets had failed to adapt in response to new information. A previous, and recent humiliation had occured in relation to the Greek Referendum vote. Traders on Betfair, in particular, got it badly wrong, with the betting market on the event showing a 75% implied probability of a YES vote, as polling closed, seemingly ignoring a Guardian's poll of polls that had predicted a No vote.

The result of the Greek Referendum also represented a significant blow for the prediction market fanboy Justin Wolfers, who in an article in the NYT on the Greek Referendum (quoting Betfair prices) had written; as they adjust betting odds in response to the flow of money, their odds come to represent a quantifiable assessment of the conventional wisdom. The problem, for Wolfers, being that the conventional wisdom simply did not adjust in response to the polls, indeed if anything, those of us that closely watched activity in the market ahead of the vote announcement, noted a very distinct homophily effect at work, with the betting market simply moving in line with the pronouncements being made by other traders on Twitter that had voted yes. Contrary to the notion of prediction market traders as being actively open-minded thinkers, willing to change their minds and to abandon their old views easily - the example of the Greek Referendum pointed to them as being prone to anchoring down and failing to respond to new information.

Greek Referendum 2015 Polls*

Outcome	Implied Probability
Yes	48%
No	52%

Source: Guardian (Average of Six Polls).

Greek Referendum 2015 Betting Exchange*

Outcome	Implied Probability
Yes	75%
No	25%

Source: Betfair.

Closer to home, on the eve of the 2015 UK General Election, the implied probability of a Conservative Majority (the actual outcome) on Betfair was 7%.

The implied probability of there being no overall majority in the House of Commons was 91 per cent. Six days later the Tories had won 330 of the 650 seats — an overall majority.

Source: Betfair Betting Exchange.

Whilst speaking about his pre-election forecast on the 2015 UK General Election, Nate Silver said:

" The forecast assigned too little of a chance to an outcome like this one, especially given that there have been significant polling errors in the UK before. It is a good lesson as we begin to plan our coverage for the 2016 U.S. election. "

In July 2015 Silver wrote as regards Donald Trump's Presidential propsects; " In the long run — as our experience with past trolls shows — Trump’s support will probably fade. Or at least, given his high unfavorable ratings, it will plateau, and other candidates will surpass him as the rest of the field consolidates."

On 7 October 2016, Donald Trump saw fit to mock Silver: "He's always been on the right side of what happened and he didn't predict me."

The notion that betting markets are omnipotent and somehow capable of pricing in all of the available information pertains, not least on twitter. Speaking on the subject of prediction/betting markets, Justin Wolfers recently wrote in the New York Times; "their odds come to represent a quantifiable assessment of the conventional wisdom." Professor Leighton Vaughan Williams, of the Political Forecasting Unit at Nottingham Business School wrote; " The power of the betting markets to assimilate the collective knowledge and wisdom of those willing to back their judgement with money has only increased in recent years as the volume of money wagered has risen dramatically."

Poll supremo Frank Luntz, who famously took to Twitter to declare that Hilary Clinton will be the next President of the United States later went on to declare that the polls had been incorrect because Trump voters refused to participate, because they didn't want to help the media/establishment.

The statistician and survey scientist Drew Linzer, another who got it badly wrong, took to Twitter to say (one hopes with a sense of humility and indeed humour) There's no sugarcoating that my forecast of a Trump loss with 215 EVs is much lower than the 306 he will likely end up with. Save that one for the grandchildren Drew.

For Sam Wang (whose organisation ascribed a totally ridiculous 99% probabilty to a Hillary Clinton victory!) and Nate Silver, who got it wrong but less wrong, it was the late-deciding voters that delivered victory to Trump.

The following table shows the Betfair betting exchange betting market on the night of the 2016 U.S. election. Hillary Clinton is seen trading at 1.14, which tanslates into an implied probability of 86%.

Betting on the 2016 U.S. Presidential election on the Betfair betting market.

In betting on individual States, the traders on the Betfair betting exchange, taking their steer from the polls, had Clinton nailed on in Florida, Michigan, Pennsylvania and Wisconsin. She lost them all to Trump, and it was these states that swung the election for Trump. State polls had systematically underestimated Trump's standing in these states, and this oversight then fed into the super bullish implied probabilities that were attached to Clinton chances vis a vis winning the Presidency: 85% at the New York Times, 71% at FiveThirtyEight and 99% at the Princeton Election Consortium etc...

STATE	DEMOCRATS	REPUBLICANS	OUTCOME
Florida	68%	32%	Republican
Michigan	80%	20%	Republican
Pennsylvania	80%	20%	Republican
Wisconsin	83%	17%	Republican

Source:Betfair.

One of the chief advocates of the term prediction markets , Justin Wolfers, was left with a significant amount of egg on his face in the wake of the U.S. election result.

In the run up to the election, Wolfers had taken to Twitter to take a swipe at Nate Silver, both for underestimating Hillary Clinton's chances and for running more than two models on the event, suggesting that this could give rise to the temptation to point to one of them ex-post.

Justin Wolfers takes a dig at Nate Silver for running two models.

Wolfers also suggested, that in the light of the high probabilities ascribed to Clinton's chances in other polls, that Silver's model must be wrong.

As the table below reveals, at the time of Wolfer's attack, Nate Silver's organisation had Clinton with an implied probability of 86%!

Wolfers also wrote an academic paper with Eric Zitzewitz, in which the pair claimed that a Trump win in the election would tank the financial markets: Market participants quoted in the financial press suggest an apparent consensus that a Trump victory would lower equity prices, weaken the economy, and increase risk. Our calibration of these movements using prediction market movements during the debate suggest that the magnitudes of these effects would be much larger than in past Presidential elections. The estimated magnitudes of the Trump discount are more comparable to those that accompanied the Brexit vote or 2003 Iraq War. As per, Howard Marks of Oaktree Capital; ... the U.S. stock market had its best week since 2014! The Dow Jones Industrials rose almost 5% for the week, taking them to a new all-time high. The Dow was up every day last week. It rose on Monday and Tuesday, when Clinton was expected to win. And then it rose Wednesday, Thursday and Friday, after she had lost.

The New York Times, for which Wolfers occasionally writes, and which has been rightly lambasted for its extermely jaundiced coverage of the election, embraced the prediction business itself during the election, happily advertising a model that ascribed an 86% probability to a Clinton presidency.

The New York Times ascribes a 86% implied probability to Hillary Clinton.

Bryan Cranston, an online lecturer in politics, and PhD candidate at Swinburne University, Australia, rose to fame after penning an article in the UK's Independent online newspaper under the heading Can Donald Trump win the election? Here’s the mathematical reason why it’s impossible for him to become President. It seemed that the traders on Betfair believed Cranston, for at one point during the evening of election day, the implied probability that Clinton would become the next U.S. President was 88%.

There is only one problem with this notion of conventional wisdom enshrined in a super smart betting market - the betting market does not always act smart. Trump was right - nobody saw him coming. On the 7th of July 2015 the implied probability on Betfair of Donald Trump becoming President was 4%.

2016 U.S. Election Implied Probabilities.

Outcome	Implied Probability
Hilary Clinton	47%
Jeb Bush	22%
Marco Rubio	10%
Scott Walker	7%
Bernie Sanders	6%
Donald Trump	4%

Source: Bettingmarket.Com analysis. 07/07/2015.

When betting opened on the next Labour Party Leader after the departure of Ed Miliband, Jeremy Corbyn was not even quoted in the betting. When he did enter the betting he was quoted at 100/1 and he was friendless in the betting market.

Next Labour Leader Implied Probabilities

Outcome	Implied Probability
Andy Burnham	29%
Yvette Cooper	20%
Dan Jarvis	14%
Chuka Umunna	13%
Liz Kendall	8%
Tristram Hunt	6%
Rachel Reeves	5%
Lisa Nandy	4%
Jonathan Ashworth	4%
Gloria De Piero	2%
Owen Smith	2%
Alison McGovern	1%

Source: Bettingmarket.Com analysis. 08/05/2015.

When betting markets act irrationally, the so-called experts will always find an explanantion as to why something other than what they have predicted has occured; although strangely, the explanation never comes to light until after the event! Self proclaimed Bremain supporter Leighton Vaughan Williams, famously dismissed the collective intelligence of the 17.4m people that voted for Brexit with an artcle declaring that it was the Sun and Mail wot won it. Oh, if only he had shared this wisdom with his devoted Twitter followers before the vote had taken place. This of course ties in neatly with Tetlock's assertion that experts rarely if ever admit that they are wrong, and that when they are forced to admit that they were, they are always at hand with a large collection of excuses; timing, an unforseen event had occured, or, they had been wrong but for the right reasons. (the "I knew it all along" or "hindsight bias," sometimes translated as "I was simply too clever by half for my own good.")

The betting market arbitrageur is the one punter that is never left holding the baby; he has left the room before the first squeal breaks out. He has won regardless of the outcome. He has surfed the wave of delusion, and parked up his surfboard, long before the tsunami has hit the shore. He and he alone, knows that an enhanced illusion led people to become unrealistically bullish and that a majority of traders are simply prone to herding and over and under-reacting to breaking news and dislay a strong collective confirmation bias.

When we look back at the Brexit vote, the key question that we must ask ourselves, is just why it was that betting and financial markets diverged so significantly from the polls (as per the folowing chart from Bloomberg).

A further mysterious aspect of the entire episode was the news that hedge funds had received private exit polls prior to the close of the polls, telling them that the UK had voted in favour of Brexit. They had traded on this information and made hundreds of millions, but such was the weight of the mug money supporting Remain in both the betting and FX markets that their trades were able to seemingly pass under the radar. The market failed to respond to the presence of insider information.

The polling average missed the final outcome by around only 4%, and as the following table demonstrates, five of the eleven polls at the close of voting actually had leave winning. They were simply ignored by a majority of those trading in the betting market.

A significant aspect of the narrative fallacy put forward by those that opposed Brexit, was, that given that a majority of the polls that came out in favour of Brexit were online polls, they had to be wrong - people, it was contended, were seemingly more likely to reveal their true voting intention when contacted by telephone (paradoxical logic par excellence).

Throughout the Brexit Referendum campaign a narrative fallacy took hold amongst a majority of betting market participants; betting market discipline had allegedly harnessed the wisdom of the crowd, who for some reason (perhaps only best known to them, or to the people that like to speak of prediction markets) were allegedly in receipt of some special causal insight that was not available to the pollsters, or indeed to anybody else for that matter; betting market participants believed that they were receiving information that was somehow privileged, or at least extremely insightful and from this they were able to construct a coherent story in which the implied probability suggested by the betting market actually made sense (In another context such behaviour might be labelled schizophrenic). This was confirmation bias and motivated reasoning at its finest. Most traders believe themselves to be less prone to biases than other traders - otherwise why would they bother to trade.

The tendency of people to employ memory bias to construct self-serving narratives of the past was also in evidence. The recent result of the Scottish Referendum served to feed the delusion and crystalise the viewpoint that when it came to the actual Brexit vote people would choose to vote in relation to their best interests (as defined by the traders) (recency bias and availability heuristic at work). Such a belief highlights the risk of mistaking what’s happened in the recent past for some sort of iron law, and ties in neatly with Kahneman's assertion that Any recent salient event is a candidate to become the kernel of a causative narrative....The core of the illusion is that we believe we understand the past, which implies that the future should also be knowable, but in fact we understand the past less than we belive we do.

There was also of course significant anchoring at work. If the betting market says that an event has an implied probability of 90% or more, then people are going to be strongly influenced by this number (the somebody must know something bias). Big probabilities breed isomorphism - birds of a feather fock together. Few, despite their best efforts manage to escape from the ubiquitous character of the anchoring bias. Anchors act as unconscious cues and influence information processing that biases judgements toward the anchors, whilst also serving to feed the notion that conformity of opinon makes something more likely to be right. It has been clearly demonstrated that people's subjective interpretation of probabilities is affected by the extent to which they want the actual event to occur. Most people, incorrectly view something with a probability of 70% as a near certainty and act accordingly.

When reviewing just why it was that the betting markets performed so poorly when it came to adjusting to new information, contrary to Bayes' rule, one would do well to remember that Tetlock found that people who spend their time and earn their living, studying a particular topic, produce poorer predictions than dart throwing monkeys. In Expert Political Opinion Tetlock found that the average expert's predictions were no better than a random guess; moreover, he also found that there is something about being a high ranking expert that interferes with forecasting. Human affairs are mostly random and intractable (and often the product of luck): even Tetlock's superforecasters got Brexit and Trump wrong.

Kanheman notes that errors of prediction are inevitable because the world is unpredictable. Secondly, he states that high subjective confidence is not to be trusted. Thirdly, it is unlikely that lessons will be learnt, because facts that threaten livelihood and self-esteem are very very quickly forgotten. Kanheman notes People can maintain an unshakable faith in any proposition, when they are sustained by a community of like-minded believers.

In what was surely one of the greater ironies of the entire EU Referendum betting market campaign, Mike Smithson, who is regularly quoted in the media as being a betting guru, tweeted that things looked good for Bremain as far as market on Betfair was concerned, but that he had no idea what this actually meant!! Whilst Smithson may well have saved face with this comment, it was hardly a ringing endorsement for his beloved political betting markets.

When we first visited the French Presidential Election Betting Market in July 2016, Alain Juppe was the warm favourite in the betting market, with an implied probability of 36%.

Juppe:36%

Outcome	Implied Probability
Juppe	36%
Le Pen	22%
Sarko	20%
Hollande	13%
Macron	9%
Fillon	1%

Source: Bettingmarket.Com analysis. 25/07/2016.

On the 7 November 2016, the implied probability that Juppe would be French president had climbed to 68%.

Juppe:68%

Outcome	Implied Probability
Juppe	68%
Le Pen	25%
Fillon	1%

Source: Bettingmarket.Com analysis. 07/11/2016.

Spring forward to 30 November 2016 and Juppe has gone, replaced at the top of the market by Francois Fillon, who back in July was a 100/1 shot in places (an implied probability of 1%).

Juppe Gone

FRENCH ELECTION - NEXT PRESIDENT
Outcome	Implied Probability
Fillon	69%
Marine Le Pen	18%
Others	13%

Source: Bettingmarket.Com analysis. 30/11/2016.

In the run up to the first round vote, which took place on Sunday 30 April 2017, Macron, who the betting market had failed to spot, consolidated his overall lead in the Betfair French Presidential election betting market, with Le Pen moving back into second place. The Melenchon bubble had been well and truly popped. However, in betting on who would win the First Round vote, Betfair traders had well and truly nailed their colours to the Le Pen mast, affording a Le Pen victory in the First Round vote a 60% probability. When the results were announced, Macron had in fact won the first round vote, with a 24% share, against a 21.3% share for Le Pen. Once again the so called wisdom of the crowd was found wanting.

Betfair Exchange: To win The First Round

Outcome	Implied Probability
Le Pen	60%
Macron	33%
Fillon	5%
Melenchon	2%

Source: Betfair. 30/04/2017.

The Richmond Park by-election was a UK parliamentary by-election in the constituency of Richmond Park, held on 1 December 2016. The Polls had Zac Goldsmith, an independent candidate and anti-Heathrow campaigner, as the clear front runner (see below). As the polls closed, the implied probability in the betting market that Goldsmith would win the seat was 68% - he duly lost.

Source: Betdata

Presidential elections were held in Austria on 4 December 2016. Almost without exception the Polls had Norbert Hofer of the Freedom Party of Austria in the lead. On the morning of the contest the implied probability that Hofer would be the next president of Austria was 73% - he duly lost.

Date	Poll	Hofer	Van Der Bellen
Opinion polling for the Austrian presidential election, 2016
16 November	Gallup	52	48
11 November	SORA	51	49
4 November	Gallup	51	49
3 November	Akonsult	52	48

The 2017 UK General Election represented another significant blow for the so-called prediction markets. On the eve of the poll, the implied probability of a Conservative majority on Betfair was 87%. The betting market also suggested a Conservative majority of between 75-99 seats. The 2017 General Election actually resulted in a hung parliament, with no party able to win a majority of the 650 seats in the House of Commons. The Conservative Party made a net loss of 13 seats.

As regards the polls, in their final published poll ICM put the Tories on 45%, a 11% lead on Labour. ComRes predicted the Tories would score 44% and achieve a 10-point lead over Labour, whilst BMG Research put the Conservatives on 46% and achieving a 13% lead over Labour. YouGov put the Tories seven points clear of Labour as did Opinium. The prize went to Survation who put the Conservatives on 42% and Labour on 40%, very close to the actual result. Their poll was simply ignored by the traders in the betting market, even though recent events had clearly suggested that polling companies were actively engaged in statistical smoothing and herding.

The 2019 federal election in Australia represented a further failure for both the polls and betting markets. On the eve of the election YouGov/Galaxy, Ipsos and ReachTEL polls had Labor ahead 51-49 on the two-party preferred vote. The average implied probability of a Labour victory with the bookmakers was 82%. One punter lost $1MILLION on Labor, whlst another lost $850,000. Australian online sports betting operator Sportsbet, owned by Flutter Entertainment, paid out AU$1.3m to bettors who’d backed Labor to win before the result was called, based on the company’s assertion that 70% of its election wagers were on Labor and "punters rarely get it wrong on elections."

Fully rational traders in efficient markets should not react with delay, and when and where they do their mistakes and their cognitive biases should be ironed out by the disciplining presence of arbitage. In their article "The behaviour of betting and currency markets on the night of the EU referendum" Linton and Auld found that - " The betting market took 2 hours to reflect the information contained in the vote whereas the currency market took over 3 hours. There was a close to risk-free arbitrage opportunity in the two markets. The arbitrage result suggests that a violation of EMH in the weak form has occurred. " And this still happened despite the fact that a number of hedge funds actually knew that the UK had voted to leave the EU before anybody else.

People still confuse the likes of horse racing betting markets - predictive - with political betting markets - not predictive. They drink the Kool Aid - cheer when the political betting market favourite wins, and acclaim that the betting market has harnessed the wisdoms of crowds. Political betting markets are not horse racing betting markets - there is no form to go on, where insiders exist they are crowded out in a sea of mug money, there is very clear evidence that all known information is not aggregated and assimilated in to the closing price and as Mutalik said: "Aside from potential clues gleaned from a fluke result, it would take hundreds of U.S. presidential elections to definitely conclude that one election forecasting model is superior to another." The phenomenon is far too complex and chaotic to be modelled with any great precision; every election has its own one-off contingencies and indetermination, deviation, non-linearity and endless feedback loops are the order of the day for those seeking to model them. There is never enough informational structure in the environment to allow precise predictions to be made, as Sam Wang so ably demonstrated with his infamous predictions on Clinton.

There is little doubt that we have now come through a period during which people simply had an over-heightened expectation of what it was that political betting/prediction markets could and would deliver. In general, and certainly in relation to the two biggest and most liquid events of recent times - Trump and Brexit - observed trader behaviour in these political betting markets deviated starkly from prediction market model expectations. Bayes' rule, which holds that rational agents will always update their beliefs on receipt of new information was simply not followed. More often than not, traders were seen to be swayed by unconscious cues in the environment and to be prone to every cognitive bias in the book - evidenced by continual displays of herding; over and under-reaction to breaking news, motivated reasoning and strong evidence of a collective confirmation bias at work, whereby a majority of traders selectively sampled information in order to support their own preordained hypotheses. (In the wake of the Brexit vote one FX trader told the author; None of us knew anybody that voted to leave.).

The likes of Hanson and Wolfers, more than happy to buy into the ridiculous notion of the wisdom of crowds, fostered an essentialist mode of thinking that reframed political betting markets as prediction markets in the process ascribing to them a truth seeking power that they simply did not have. The shocking performance of these markets vis a vis Trump and Brexit - where they simply failed to follow Bayes' rule - (at the very least they could have been in line with the 538 models)- clearly demonstrated how easily essentalist thinking can lead to mistaken predictions about future outcomes. A majority of those that trade the political betting markets are in fact nothing but shining examples of the Dunning-Kruger effect. They are guilty of giving undue weight to the prices that pertain in the political betting market at a particular moment in time as though they are somehow a direct and unmediated reflection of objective reality, when in fact they are nothing more than a product of noise and changes in the expectation and sentiment of a group of people all of whom are trading on the back of incomplete/assymetric information.

Truth be told our behaviour is always conditional and context dependent and generally speaking we are unreliable decision makers. Our beliefs are distorted by motivational, ideological and cognitive biases, such that at any particular moment there are often a whole series of misperceptions at work. We rush to judgements,failing to take account of the most salient variables, misconstruing noise as information and being heavily influenced by irrelevant factors, such as our current mood, our last bet, our current bank balance; the pseudo-signals that are being sent out by the betting market itself and by the so-called betting market experts.

When traders in political betting markets make predictions their so-called wisdom is actually interspersed with and filtered through significant amounts of signal, noise, and bias in unknown proportions. Accordingly and unsurprisingly Subjective Probabilities In Political Betting Markets, Like Beliefs, Are Nothing More Than An Epistemic Anchor Set Down In a Sea of Ontological Complexity. Because almost nobody knows how to build a good political forecasting model (and it is becoming increasingly more difficult to do do because of polling issues)................................ people's priors are not very strong and accordingly they do not actually have something concrete to anchor on to. In the absence of strong anchors there is nothing but noise at work and subjective probabilities generated under these conditions are aleatory (depending on the throw of a dice or on chance;random) fanciful, speculative and merely the result of guesswork. A tweet to the effect of "the subjective probability of X winning the x election on Betfair is x%" should be taken with a pinch of salt.

Final Predictions at Opening of Polling 2016

Outcome	Betfair	Nate Silver*
Hillary Clinton	82%	72%
Donald Trump	18%	28%

NOTES:Why And When Political Prediction Markets Fail.

1. Herding or herd behaviour represents the tendency for an individual to follow the actions of a group, whether those actions are rational or irrational (out of fear of being alone or missing out).

2. Confirmation Bias is the tendency to search for, interpret, favour, and recall information in a way that only confirms one's pre-existing beliefs or hypotheses.

3. The Dunning-Kruger Effect is a cognitive bias in which people wrongly overestimate their knowledge or ability in a specific area. This tends to occur because a lack of self-awareness prevents them from accurately assessing their own skills.

4. The "Zollman Effect" is very much applicable to political betting markets. It holds that it is sometimes worse for communities to communicate more. In particular, groups with more network connections will be generically less likely to arrive at a correct consensus when they are all simply trading on the back of incomplete information.

5. It is clear that there is not a sufficient informational structure in place in the political betting market environment to allow precise predictions to be made. Neither is there timely and effective feedback mechanisms in place that facilitate error-correction and learning in real time.

6. It should never be forgotten that the reasons as to why most people trade are not conducive to optimal decision making at the best of times (never mind under conditions of ambiguity)- people trade in order to relieve boredom, so as to be able to deal with indentity defusion, on the back of impulsiveness and sensation seeking, because of a desire to increase ones self-esteem by being top dog, a fear of missing out on the action etc. Political betting markets attract a particular type of trader; loud, a heavy user of social media, keen to secure bragging rights.

7. Until the event is terminated the betting market continues to throw up an unusual concatenation of circumstances. A majority of those trading in the market are trading on the back of incomplete information; assymetric information flows are the order of the day; numerous constraints (cognitive and otherwise) persist; beliefs and assumptions are not updated in the face of new information. Strong emotions drive attentional focus. Dysregulated dopamine function drives aberrant salience attribution. Stress, induced by financial and temporal constraints contributes to deficiencies in critical and rational thinking - people display an impaired ability to think about thoughts and emotions. Prior expectations attenuate the discovery process. The pattern seekers seek patterns as a means of reducing uncertainty, ambiguity and confusion. The stick is a snake is a stick.. It stands to reason that when confronted with uncertainty we should always be wary of believing the first solution that springs to our mind.

8. People are generally unaware of their own unawareness - they more often than not misundertand their own misunderstandings and are ignorant of their own ignorance. They fail to detect the differences that make a difference. They reach their conclusions, burdened by memories of the past, limited by the shortcomings of the power of foresight and heavily influenced by their unconscious biases and yet still they manage to convince themselves that the conclusion that they have arrived at are the right one under the circumstances.

9. Behaviour is situational: the social setting that we find ourselves in shapes our behaviour independent of our personality and our perceived sense of self. We all have our own unique set of cognitive and affective traits and it is how these play out in particular situations, that influences our behaviour the most. It is our context-specific beliefs about what to expect that makes us prone to reject any new information that violates our expectations. This is why we cherry pick data and make spurious inflated correlations and adopt a one-dimensional approach to causation. We are almost all guilty at some time or other of seeing spurious correlations when none exist, making incorrect judgments about the extent to which a particular stimulus is a plausible cause of a given response and of specifically over-emphasising evidence that inflates the veracity of our misguided beliefs.

10. Research suggests that the explicit reasoning we offer to ourselves and to others after the event is more often than not merely rationalisation, that we actually act intuitively, on the back of our instincts, inclinations, stereotypes, emotions, neurobiology, habits, reactions, evolutionary pressures, unexamined principles, or justifications other than the ones we think we're acting on, then we tell a post hoc story to justify our actions to ourselves and to others. We tell the story that seems the most plausible to both ourselves and others in the circumstances.