What can we learn from prediction markets?

Blog

What do Google, Pfizer, HP and France Telecom all have in common?

Besides being hugely successful, they have all seen the benefits of opening up their decision making to more than just the board of directors. Every one of the above companies uses some kind of prediction market to help them make better decisions in their companies.

A prediction what?

Good point. I’m getting ahead of myself. But if you’ve been reading this blog for a little while, you might already be familiar with prediction markets. If you’re new, welcome.

You’d be forgiven for not knowing who Matt Singh is. Most people didn’t (sorry Matt) before 6 May 2015. The night before the UK general election, all of the polls that were conducted by the big polling companies showed that Labour would likely tie the election with the Conservatives.

But Singh had other ideas. He’d been following the polls for a while. Here’s what he told Businessweek:

“I could see that the polls as they had been were quite a way from where I expected them to end up. Look at the fundamentals: the leader ratings, the economy, the local election results. There was a big mismatch.”

In fact, he was so sure of his ideas that he put a bet on the outcome. Jokingly, he says that the win he netted was a post-Lehman level banking bonus. 💸💸💸💸.

What was the problem?

Well, there were actually a few problems.

The first one, and the most significant, is that the data which led to the most inaccurate polls since polling began in the UK 70 years ago was fundamentally flawed. When polling companies survey people, they try to get a representative sample of the population. This means that they will ask a sample of people based on various demographic features – age, political allegiance, income.

What appears to have happened in last year’s election is that companies like YouGov got it wrong because they didn’t ask enough people over 70, or asked too many Labour voters.

The second big problem, some experts suggested to denials from pollsters, was that the polling companies didn’t want to get it wrong and so they ‘massaged’ their data to bring it into line with all of the other companies – which seems a fairly reasonable thing to do if you’re looking to save face.

Unfortunately for them, in this instance, everyone got it wrong. Except some guy called Matt Singh – who got it really right.

What does this have to do with prediction markets?

Another great question. You’re on fire. You should consider using an online decision making platform to ask some of these questions⚡⚡⚡⚡.

But seriously.

The purpose of a prediction market is to even out errors like this. They serve the purpose of removing the need to get the right the answer as an individual and instead focus on the benefit of getting the answer correct as a group.

The most famous of these markets was created in 1988. The Iowa Electronic Markets project is actually a collection of prediction markets which aims to predict the outcome of elections by selling ‘contracts’ to individuals based on how they think an event will turn out.

For example, in 2015, people could have purchased a contract for the Tory party and been happily surprised when you came quids in. However, if you had bought a Labour contract, you’d have been sorely disappointed at your cash loss and also seeing a sad looking Ed Miliband.

But the real purpose of this kind of system is that if 48 people had bought a Labour contract at £1, those interested in the data could see that the public thought that Labour would get roughly 48% of the vote.

Do the markets work?

Yes and no.

There is a lot of research to show that if there’s a theoretical group of 90 amateurs and 10 experts, the group as a whole will be able to better predict with better accuracy than each of the experts alone. This alone is a pretty strong argument in favour of a system of prediction markets.

Additionally, economists believe that there is a chance that prediction markets which rely on actual cold hard cash being put onto an option rather than play money stand a better chance of winning. This is probably down to the fact that people perform better and tend to be more reserved when there is the potential of tangible loss.

However, when real world loss is involved – even at small numbers (most prediction markets have very low caps on maximum investment) – there is a very real possibility that people will choose the safe vote. Just like the pollsters were accused of doing in 2015, people might not want to be the only one to lose money and so they will buy contracts that everyone else is buying. After all, your friend might have some secret knowledge that it would benefit him financially to keep secret. You may have heard this referred to as a bubble – lots of people believe that others have secret information.

The being said, the IEM, according to James Surowiecki’s masterful book The Wisdom of Crowds is quite accurate:

A study of the IEM’s performance in forty-nine different elections between 1988 and 2000 found that the election-eve prices in the IEM were, on average, off by just 1.37 percent in presidential elections […] The IEM has generally outperformed the major national polls, and has been more accurate than them even months in advance of the actual election.

So what can we learn from them?

Prediction markets work so well because they entertain the possibility that a big group might have better insight than a small group.

Time and time again, we have seen that people who have quiet voices have valuable opinions. Those valuable opinions may be the tipping point in a decision making process which creates or destroys millions of pounds or, while we’re thinking big, changes the world for good or for bad.

The downside of prediction markets is that the opinions of others are factored into your decision making process. Individuals with loud voices tend to express themselves early in a decision making process – hiding the value or number of people making a choice can have the benefit of allowing you to vote with your own ideas rather than the thoughts of others.

(Coincidentally, doopoll allows you to make your live results visible or hidden.)

As a summary

To pull all of this together, what we’re saying is this:

  • Decisions are made more accurate with a large group
  • Anonymity encourages free expression
  • Quiet voices might have very valuable information
  • Prediction markets work well, anonymous polling works better
  • Outsiders aren’t always wrong