The potential of prediction markets is well known to anyone who’s read James Surowiecki’s bestseller, The Wisdom of Crowds. Well-designed markets can help draw out knowledge contained within disparate groups, and research shows that when people have money on the line, they make better forecasts. That explains why Google, Microsoft Corp., and even the US Department of Defense have used prediction markets internally to guide decisions, and why university-linked political betting sites such as PredictIt, where wagers are capped at a few hundred dollars, sometimes outperform polls. So what are the prediction markets and what perspectives do they open in front of us?
Prediction markets, also known as the ‘wisdom of crowds’ can make predictions in a number of domains, including sport, entertainment, politics or any other event or outcome that can be objectively verified ex post. They enable individuals to trade ‘bets’ on whether a specific outcome will occur or not. If many participants are trading in such a market, market prices will generate a prediction of the outcome, based on the aggregated information of the participants. While they have existed for decades, their field of use is only getting broader due to blockchain-related inventions and decentralized governance models,.
The magic power of money
Prediction markets are not an invention of our time. In Europe, as early as the 16th century, people placed bets on who would become the next pope, and in Great Britain and Ireland since the 18th century, such bets were mainly based on the likelihood of political events.
The prize for a correct prediction and the payment for a wrong one was not always measured in money. At first, even those who had no money were allowed to bet, and in case of a loss, the players had to entertain the audience with some kind of performance. However, it was the cash prize and the desire to avoid losing money that gradually turned betting into a way to look into the future. Economists found an explanation for this already in the 20th century: the chance to make a profit and the desire not to lose money makes bidders carefully analyse the ongoing events and significantly improve the quality of their forecasts.
In the United States in the 19th century, small bets on the outcome of the next presidential or local elections began to be made in billiard rooms. For instance, The Curb market, which became the American Stock Exchange at the beginning of the 20th century, was originally a crowd of street traders who offered, among other things, bets on the outcomes of the upcoming elections and demanded a cash advance. In the 1900s, electoral betting in the United States was already reminiscent of a real market: central newspapers published bets on the outcome of the presidential election, and it was quite possible to determine the future winners of the election by who was betting more.
But the First World War, its aftermath, and the shift in the 1930s to political forecasting through public opinion polls stepped back prediction markets for a while. They began to revive only with the development of the internet, but even during the heyday of Intrade in 2004, the exchange collected from players who bet on the winner of the next American election (in which George W. Bush won), was much less than their predecessors almost a century ago – about $25 million.
Why are prediction markets necessary?
The main purpose of prediction markets is the aggregation of beliefs over an unknown future outcome. Because they incorporate a wide variety of thoughts and opinions, prediction markets have proven to be quite effective as a prognostic tool. The classic example one can use to explain the value of prediction markets is political elections. Prediction market platforms allow the creation of a poll-like market where the participants can trade the outcomes in a way which is similar to sports bets. So if a business owner thinks that a certain politician being elected would negatively affect the revenue of his business, he could bet on the event of a successful election and thus hedge against the disadvantageous outcome. In this case, decision-makers do not directly vote on policies but rather on desired outcomes (or “KPIs” for the management). Prediction markets are set up for various policies to predict which policy is likely to have the highest impact on this metric and which will be the one that actually gets implemented. Thus, these markets can directly advise important policy and governance-building decisions, by giving more accurate estimates of the consequences of those decisions. But it’s clear today that the potential impact of this concept could go far beyond betting.
With individuals being financially incentivized to predict the outcome/decision most likely to find consensus among all stakeholders, solutions that satisfy all parties can be found much more efficiently. This is because the fight for personal interests in a decision-making process gets balanced out by an economic self-interest of predicting a consensus-reaching alternative.
How do prediction markets work?
The ability of prediction markets to provide accurate and reliable forecasts is due to three reasons:
- The market mechanism of such exchanges is an algorithm that aggregates information.
- Knowledge of the situation gives a monetary reward, so the players are interested in having the information that reflects it most fully.
- Constant updates of the information are required to continue the successful trade.
Imagine that someone who is involved in forecasting has information that the probability of a certain event is much higher than the quotes reflect. In this case, this player and those who are guided by his opinion will pull prices up.
Now let’s imagine that another group of participants has the opposite information. This means that they will sell their contracts, which means they will move the price down. As a result, the price on the market will reflect the balance of information of both profile participants and those who have information that profile experts do not own. It is the combination of these two sources that can improve the accuracy of forecasts and provide smaller statistical errors than professional forecasts and exit polls.
The simplest possible prediction market can be explained by the example of the elections. In that case, it will have the following events: Event A: Candidate A gets elected Event B: Candidate B gets elected. To simplify this case, let’s say it is only possible to participate in this market using US Dollars. After the market is set up, participants can invest for example $100 and receive 1 “A-token” and 1 “B-token” in return. Both types of tokens automatically pay out $100 each in the event that the respective outcome happens.
If the outcome does not happen, $0 will be paid out for this type of token. So if no action is taken, $100 (the initial investment) will be paid out with a 100% certainty. However, these tokens can also be sold freely with other participants. So similarly to how stocks represent the aggregated investors’ prediction on a company’s future performance, these outcome tokens will be priced according to supply and demand and represent the aggregated probabilistic predictions of the respective event. For example, “A-tokens” could be priced at $65 while “B-token” trades at $35. This can be read as a 65% probability that Candidate A gets elected versus a 35% chance of Candidate B taking over the office. If anyone disagrees with this probability distribution, he is economically incentivized to buy the undervalued or sell the overvalued token, each of which will have an effect on the price. As time goes by and more and more people buy and sell the tokens, the prices will fluctuate depending on the combined information held by market participants.
Oracles and obstacles
In traditional centralized prediction markets, the company or individuals running the market would fill the oracle role when the event has occurred and pay out the profits to the correct predictors. In decentralized prediction markets, oracles are needed to submit and verify information on real-world events & outcomes to the blockchain for the smart contracts to initiate the right payouts. Oracles can come in different forms such as software, hardware, or humans and are necessary to control the payouts of costs.
Regarding the criticisms that prediction markets face, one of the biggest ones is that their realm is often misunderstood, seen as akin to sweepstakes, and used primarily for predicting gambling. In addition, one can question the moral aspect of betting on the likelihood of violent events, such as wars, since it allows to capitalize on someone else’s misfortune. Moreover, there is another moral hazard: the prediction of a certain violent event, such as a political assassination, can provoke someone for the sake of winning to make such a scenario come true. And if the forecasts of such a market are taken into account when formulating a policy, then there are risks of facing manipulation: there will be those interested in directing the market towards a certain result.
At the same time, most of the world’s think tanks resort to prediction markets. Today, the use of big data and data processing applications along with prediction markets is underestimated. Given that people have a financial stake in deducing the right answer, they tend to consider the question seriously, weigh their own knowledge and the publicly available information about the outcome, and answer honestly.
Prediction Markets as a Forecasting Tool
The crowd wisdom begs questions such as, ‘‘why would the markets be more accurate than forecasting tools?’’
It is likely that the capabilities of markets, such as broad access to other types of information, access to real-time information, trader anonymity, truth-telling, and other issues provide prediction markets with the ability to generate highly accurate forecasts.
Perhaps the primary benefit deriving from the use of markets for internal purposes compared to other forecasting tools is that participants may have access to information sources that they may not have had access to prior to implementing the market. Since the markets involve a wide range of participants, information and knowledge are gathered from a broad range of sources, potentially some that are not part of the normal reporting process. As a result, prediction markets potentially open new communication channels, as new traders join. In addition, they create a new medium for interacting with those information sources.
Besides, market forecasts gather timely information on a continuing basis, as long as there is new information and as long as the market is continuing. Alternatively, a forecast using a sophisticated approach is likely to have limited data and operate over a limited time horizon.
If markets have such excellent accuracy, should we see them as stand-alone tools or should they be used in conjunction with other forecasting tools?
If there are sophisticated forecasting tools, then any trader, with access to that information, will be able to embed the forecasting information in the market. Further, those traders will bring the official forecasting information in with any additional information that the trader feels is important. However, if there is no forecasting tool, then no additional information other than the expectations and other knowledge of the traders will be embedded in the price.
Accordingly, from an information perspective, more and different information is embedded in the market if there are traders who also have access to forecast information. As a result, we generally would expect better performance when markets are run in conjunction with other sophisticated forecasting tools. Prediction markets are especially suited to situations where there is sparse data otherwise available that may be used to define a forecasting model. If there is sparse information, a market can act to pull the information together and provide an aggregation function. Besides, traders can trade on information asymmetries with the result being the integration of that information in the market price and a loss of asymmetries.
Prediction markets are emerging as a valuable forecasting tool in diverse application areas from sales forecasts to project success. This social analytics strategy could potentially help resolve a number of uncertainties, especially where prior data may be sparse or the situation is so unique that other forecasting tools are less useful. However, it is crucial to remember that while the technology for markets is easy to implement, aggregating a number of traders and involving them in a market process to resolve uncertainty in an organization is more difficult. The application of these methods to forecasting requires broader leadership and an effective process for organizational buy-in. But to sum up, the emergence of a regulated prediction market could be useful to attract additional money to niches and conduct effective risk management in vast fields from culture to government modelling, cybersecurity or natural disasters.
