- What is a prediction market
- Interesting examples
- Why prediction markets are so accurate
- Futarchy and decision markets
- What prediction markets need to be successful
What is a prediction market
A prediction market is a speculative market that asks a question about the future. In the case of a yes/no question, participants can buy and sell contracts on either side of the market. The yes contracts pay out if the answer to the question turns out to be yes. The no contracts pay out if the answer is no.
If it sounds like a betting market, that’s because it is. The main difference is that prediction markets are set up so that the market price is a direct prediction of the probability of an event occurring. Also prediction markets offer trading mechanisms that are normally not available on betting markets.
An example of a prediction market is “ Will Elizabeth Warren win the 2020 US presidential election?” Yes contracts can be bought and sold at a price between 0 and 100, set by supply and demand. If Ms. Warren wins the price of the yes contract goes to 100. If she does not win the price goes to 0.
The price of the yes contract is the market’s prediction of the likelihood of Ms. Warren winning. For example if the yes contract is trading at 38, it means the market predicts Ms. Warren has a 38% chance of winning.
Prediction markets use the wisdom of the crowd because all the participants in the market set the price, and the price is a prediction of the likelihood of a future event. Prediction markets have proven to be a very accurate forecasting tools, more accurate than experts or polls.
Here is a great video explanation
Incredibly accurate election forecasts
In 2008, online prediction market Intrade predicted that Barack Obama would win 364 electoral college votes. In fact he won 365. For the 2012 election Intrade correctly predicted the results in 49 of 50 states. Did Intrade predict the Trump victory in 2016? Intrade was shut down after being sued by the US Commodity Futures Trading Commission. Thanks Obama!
US Navy uses prediction market to find missing submarine
The below quote from Rational Markets: Yes or No? The Affirmative Case is edited for brevity.
In 1968, the submarine USS Scorpion was officially declared missing. Five months later, after extensive search efforts, her location was still undetermined. The Navy’s top deep-water scientist had all but given up. He asked Navy personal to bet on the probabilities of different scenarios that could have occurred. Averaging their responses, he pinpointed the exact location (within 220 yards) of the missing sub.
This episode was described in Sontag, S., and C. Drew 1998. Blind Man’s Bluff: The Untold Story of American Submarine Espionage. New York: Harper/Collins.
The truth hurts inside corporations
I love hearing and reading everything Dr. Robin Hanson has to say about prediction markets [blog][academic papers][videos]. Particularly when he describes running prediction markets as an outside consultant in large companies.
Suppose there is an important deadline for a project and management is signaling that this deadline should be met. Then Dr. Hanson runs a prediction market for all employees, which predicts the deadline has a 5% chance of being met. This is embarrassing for management and might affect employee morale.
Dr. Hanson describes prediction markets as autistic in that they blurt out the truth with no regard for the consequences, social norms and who might look bad or be offended. Prediction markets don’t play politics, exactly what we need sometimes!
Prediction markets have proven to be more accurate than experts at predicting movie ticket sales, with 96.5% accuracy. In 2007, players in the Hollywood Stock Exchange correctly predicted 32 of the 39 major-category Oscar nominees and 7 out of 8 top-category winners.
Internal prediction markets at Hewlett-Packard forecast sales of different printer models better than HP’s own sale projections formed by HP sales and accounting managers. [Harvard Business Review][Yang, UCSD]
Why prediction markets are so accurate
Prediction markets are so accurate because they have built in economic incentives that encourage accuracy. There is a price for being wrong and a reward for being right. People with the best information participate the most and therefore have the most impact on the prediction.
If someone has good information that the market will result to yes they will buy yes contracts causing the price of yes contracts to go up. The better the information they have the more yes contracts they will buy, the more the market will predict yes.
The people with the worst information whose opinions we want to take into account the least, are the people who will participate the least. Therefore these people affect the market’s prediction the least.
Remove noise and amplify better information
If someone posts on social media, that they are sure Elizabeth Warren will win in 2020, there is no cost if they are wrong. Their opinion has no backing and there is no way to weigh it against other opinions expressed on social media or in conversation.
If we used social media sentiment to make a prediction we would give more weight to the opinions of the loudest people. If we used polling we would give equal weight to all respondents opinions. Prediction markets improve on this by giving more weight to the opinions of people who have better information and are willing to risk more.
If someone buys 10 m฿ worth of predictions that Elizabeth Warren will win in 2020, there is a cost if they are wrong. We assume they have good information backing their opinion. Even more so if they buy 1000 m฿ worth of predictions.
Let’s consider 2 individuals and how much we want their opinion to affect the most accurate prediction possible on Ms. Warren’s chance to win the 2020 US presidential election.
- Average Joe votes how his parents voted, follows the news a bit and watches most presidential debates every 4 years.
- Washington Sue is Elizabeth Warren’s Doctor. She knows that Ms. Warren’s undisclosed terminal illness is worsening and that this will be made public soon.
To make an accurate prediction of Ms. Warren’s chance to win the next election we want to take Washington Sue’s opinion into account more because she has better information. In a prediction market she will participate the most meaning she will affect the prediction the most.
Average Joe has no reason to think he can beat the markets’ prediction and make money, so he won’t participate and his low value opinion will not be taken into account.
The prediction market does not need to actively seek out the people with the best information. The market does not have to know anything about Average Joe and Washington Sue or make a value judgment on who has the best information.
Because of the economic incentives those people will self-select and tell the market how good their information is by how much they choose to participate.
If someone with bad information mistakenly thinks they can beat the market’s prediction and make money, the market will reflect this temporarily before being corrected by other participants.
Participants feeding bad information into the markets would lose money and therefore be less likely to keep participating. Participants feeding good information into the markets would make money and therefore be more likely to keep participating. This will make the markets more efficient with more accurate predictions over time.
The same is true for people trying to game the market for the sake of appearances, like making it appear Ms. Warren has a greater chance to win than she actually has. An efficient market will quickly correct this distortion.
People may counter that how much money someone risks is more related to how much money they have to risk rather than how good their information is.
This is not wrong but it is not relevant to the accuracy of the markets’ prediction. It applies equally to both sides of a market and averages out if the market has enough participants. On average all participants in the market have the same amount to risk and participants with more and less than average will be spread roughly equally on both sides of the market.
Also, whether we like it or not, people with more money have advantages over people with less money so it is reasonable to assume they might have better information. Perhaps because:
- They can afford to buy better information.
- They have more time available to study the market.
- They have a higher level of education.
- They are more likely to have insider knowledge or have contacts with insider knowledge.
Futarchy and decision markets
Decision markets are prediction markets set up to predict the effects of potential decisions. Allowing us to then take the decision with the best predicted effects.
Futarchy is a governance model where actions are determined by decision markets. It could be a model for society where many people and organizations use decision markets to determine their actions.
Let’s run prediction markets inside large corporations like
If [current CEO] is still the CEO on January 1 2019 will the share price be above [current share price]?
So for Apple that would be
If Tim Cook is still the CEO on January 1 2019 will the share price be above $150.27?
If the market trades under 50% it is a clear signal that the Board of Directors should fire the CEO.
You can imagine how much enthusiasm this idea gets from CEOs. Even Boards might not like the idea because it may be seen as overly aggressive and cut-throat. It may also force the Board into making a decision they otherwise do not want to make.
You can add nuance by running 2 related decision markets. Suppose new legislation called The Jobs Act is proposed, with the stated aim of reducing unemployment from its current 5%. We would run decision markets that ask:
If The Jobs Act is put into effect will the unemployment rate on January 1 2019 be under 5%?and
If The Jobs Act is not put into effect will the unemployment rate on January 1 2019 be under 5%?
If the markets predict putting The Jobs Act into effect increases the chance of the unemployment rate dropping it should be put into effect. If the markets predict it decreases the chance of the unemployment rate dropping it should not be put into effect.
Suppose military expense proposals had to pass a decision market testing whether they would decrease the number of citizens killed by enemies.
Questions of crypto currency governance are the perfect environment in which to use decision markets. A good example would have been when the Ethereum network was agonising about whether or not to bailout the DAO. Markets such as these could have been run to help make the decision.
If we bailout the DAO, what will the value of 1 ether be on January 1 2019?
Min: $0. Max: $100,000
If we don’t bailout the DAO, what will the value of 1 ether be on January 1 2019?
Min: $0. Max: $100,000
Until now we have discussed prediction markets with yes/no questions. That is not the only type of outcome prediction markets can deal with. They can also handle scalar markets where the prediction is some number within a range.
In the above example, due to the min and max values, the contract trading at 2 would be a prediction of $2000, and the contract trading at 14 would be a prediction of $14,000.
What prediction markets need to be successful
So why aren’t prediction markets widely used? For prediction markets to provide accurate predictions and be widely used they must meet 3 conditions:
- Well run on a good platform.
- Uncensorable and anonymous.
- High liquidity and participation.
Intrade had 1 and 3 but not 2. New decentralized crypto currency prediction markets like Augur may have 1 and 2 but not 3.
Well run on a good platform
For a prediction market to be well run on a good platform it needs:
- Clear rules on how the market, trading and payouts work, which are reliably applied.
- A clear well defined question for the market to trade on, that will have a result that is easy to determine and verify.
- Be easy to use and accessible to everyone.
This condition has already been met by many government currency and crypto applications, from betting exchanges to consumer financial products.
Uncensorable and anonymous
When prediction markets had to use government currency, they could be censored and participants could be identified. Crypto changes this.
Before crypto we have only had prediction markets that are run by centralized organizations. These organizations have been censored from hosting markets that are politically or legally sensitive.
Intrade was censored from offering commodity price prediction markets by the Commodity Futures Trading Commission. The Pentagon was censored from running prediction markets about events in the middle east by Congresspeople [New York Times][Wikipedia].
Crypto allows for decentralized, uncensorable prediction markets on any topics. Including topics which will make embarrassing and unpopular predictions.
Washington Sue would not want to participate in a government currency prediction market because her trades against Elizabeth Warren could be linked to her identity as Elizabeth Warren’s Doctor. This could be bad for her career and land her in legal trouble.
Crypto’s anonymity allows the people with the best information to participate by keeping the economic incentive of financial gain while removing the social disincentive of possibly being identified.
High liquidity and participation
Prediction markets need liquidity and a lot of participants to be accurate, useful and attractive. Without these prediction markets:
- Lack economic incentives for people to participate.
- Do not reflect the wisdom of the crowd because the crowd is not participating.
- Are easier to game.
- Are less efficient at correcting market distortions caused by single large participants.
The solution to this problem is increasing the wider adoption and acceptance of crypto. We are at the tipping point and crypto is about to be mainstream enough to satisfy this condition. This is another reason to try to spread crypto’s adoption.
This will be the first time ever that all 3 conditions for prediction markets to be successful and widely used will be met. Another way crypto can change the world for the better. What a time to be alive!