Getting it wrong
Pollsters whose job it is to talk to people and predict the outcomes of elections, sometimes get things badly wrong: the general election in Britain in 2015 when Labour performed unexpectedly poorly; the Brexit vote this year which took everyone by surprise; and finally, the US presidential election, which shocked the whole world.
Pollsters got these so wrong because they rely on sampling and extrapolating techniques. Statistical analysis is accurate only if the sample is representative of the whole. In the US presidential election, for example, voter samples excluded entire voter segments such as white working class voters and were also disproportionately skewed towards other segments such as politically conscious urban voters (who don’t get fed up of being called again and again considering there were many of these polls just before the elections). Yet another problem was that voters were not entirely honest and may have held back their choice for whatever reason.
Surveys from Princeton Election Consortium that used statistical “Bayesian” model developed by Professor Sam Wang, Neuro / Data Scientist, and Professor at Princeton, the model which correctly predicted 49 of the 50 states in the 2012 election, confidently rated Hilary Clinton’s chances of winning at 99%.
Predicting the unpredictable
Provided the sample is statistically relevant, determining which way people will vote is still the easier thing to do.
Predicting the outcome of events that do not entirely depend on individual decisions is far harder. In such cases, polling is useless. No one who is polled has any specific information to share. When something is unknown, it does not matter how many people you ask, the larger the poll, the larger the pool of guesses.
But for events that are not entirely random there are different bits of relevant information that may be known to different people. Some people believe that asking a group is more accurate than asking a single individual.
James Surowiecki, the author of the best-seller “Wisdom of Crowds”, argues that “under the right circumstances”, a group of people is better able to estimate, forecast, or predict than single individuals in that group.
Economists believe that the best way to predict the future, particularly where politics are concerned, is to let the markets do it for you. Betting companies have been doing this for years relatively successfully.
But betting or “prediction” markets failed to predict the Brexit vote. Polls and markets shifted towards “leave” in early June but not enough to overtake “remain.” On the day of the referendum, markets priced “an 85% likelihood” for Britain to remain in the EU. Markets also failed miserably to predict Trump to win the Republican nomination. But perhaps bruised from these miscalculations, markets clearly predicted the outcome of the US presidential election.
Prediction markets are considered more accurate than polls because they take into account the “weight individuals’ beliefs by conviction as well as frequency”. This means that people who are convinced of the outcome often place high wagers than others who are simply just placing a bet. These punters can also adversely influence the market if they are driven by closely held biases rather than rational judgements something known as “cognitive bias”.
Specialised predictive markets, set-up for the purpose of trading the outcome of future events, are exchange-traded markets best positioned to benefit from aggregate individual predictions.
The Iowa Electronic Market is an academic market dedicated to elections and limits the positions its members can take to $500. iPredict operates out of New Zealand and trades in a number of political events relating to that country. NewsFutures counts major corporations such as Siemens, Renault, and Pfizer as its customers. Microsoft launched its own predictive markets and Google uses the concept internally to identify “discontinuous product ideas.” Other large companies that use predictive markets internally include France Telecom and Hewlett-Packard. The Hollywood Stock Exchange, now part of Cantor Fitzgerald, trades in the prediction-shares of movies, actors, and directors and has made many accurate Oscar predictions.
The caveat Surowiecki inserted in his argument about the wisdom of crowds: “under the right circumstances”, is made up of 4 conditions:
- Diversity of opinion – each individual must have some method or information which they leverage to arrive at their decisions
- Independence – Opinions of individuals are their own. Unlike crowd behaviour where people follow the crowd or where the crowd speaks as one voice after conferring, individuals must make their decisions independently and on their own. Voting preferences are heavily influenced by opinions and biases of others such as the media. Even stock markets, where investors do their own research, are heavily influenced by media buzz and rumours that lead to bubbles and crashes
- Decentralisation – Individuals are able to do their own research and use their own resources to acquire whatever information they need
- Aggregation – some method or mechanism is available to aggregate the individual decisions accurately
It seems that the conditions for the wisdom of crowds to work are almost tailor-made for blockchain based distributed ledger (DL) systems.
DL systems can be developed to aggregate individual opinions that have been arrived at independently by a very diverse range of people. That is exactly what distributed ledger based systems do.
DL systems are designed primarily for decentralised aggregation of diverse but independent input in order to arrive at collective consensus-driven decisions enabling some processes and even organisations to operate autonomously.
Of course there is no way to ascertain how independent these opinions are because everyone is exposed to public opinion or influenced by general market sentiment, or worse, collective biases which often override reason, but certain questions may work better than others.
Online betting exchanges are not all,owed to operate in the United States but they are available elsewhere such as the world’s largest exchange market, Betfair.
Augur is an exchange prediction market that uses blockchain concepts to predict the future set to go live in 2017. The way it works is simple. Individuals on the platform can make predictions by trading shares in event outcomes such as the outcome of the presidential election. If the odds are even, the share is bought at 50 cents, if you win, you get back $1. If you lose, nothing. Just like other prediction markets.
The key difference with Augur is that it is entirely open-sourced and decentralised and allows for the execution of contracts on the Ethereum platform. This means that anyone, anywhere can set up a prediction market of their choice and it will manage itself autonomously.
Notable supporters of the project include Intrade co-founder Ron Bernstein, the Thiel Foundation and Vitalik Buterin.
While prediction markets are usually better than other methods of divining the future, these will provide the ultimate test for blockchain concepts to offer an open-for-everyone method for assembling collective wisdom and if the Augur project delivers results, it may end up as the most useful application of blockchain technology – more so than its currently discussed uses in others sectors such as finance and payments.