Originally written on 5th August, 2008 for IIM Calcutta’s IMZine:
—————–
Imagine that you are an associate at the Boston Consulting Group, working for a project in the oil refining sector. Your client requires your help in estimating the oil prices three months down the line. What do you do?
You could either try to dig up information from the knowledge base that the Group has developed over the years and from other associates and partners, build an Excel model and pray that it gives you true values. Or, you can create a prediction market. That is to say, create some sort of a virtual market place where people (possibly other consultants and domain experts) can bet on what the prices of oil would be in the future. Then you seed the market with some initial funds, and let the natural tendency of the market play out – the market price will gravitate towards what people feel is the most optimum price. The lure of making profits in the market make people bet for what they believe will be the most appropriate oil price. And that is the prediction you wanted.
Prediction markets work because they can tap into the “wisdom of the crowds”. People generally have a lot of tacit information about a lot of things that would normally be held in isolated silos, rarely if ever available to the one who needs it. Prediction markets can tap into this information and bring to light a more comprehensive picture. Financial journalist James Surowiecki, who wrote the 2004 book The Wisdom of Crowds, explains: “under the right circumstances, groups are remarkably intelligent, and are often smarter than the smartest people in them.”
Prediction markets are essentially speculation markets developed for the purpose of making predictions. The market trades “Event Futures” (Will India sign the nuclear deal?) or “Parameter Futures” (oil prices after three months). The current market prices are interpreted as predictions of the probability of the event or the expected value of the parameter. Prediction markets are thus betting exchanges, and present no risk to the bookmaker.
A lot of companies, including ArcelorMittal, Best Buy, General Electric, Google, Hewlett-Packard, Intel, Microsoft, Nokia, and Samsung, have started understanding the power of such markets in predicting public reaction to new products, sales revenues of the next quarter, future prices of a commodity or even shipping dates of software deliverables. Naturally there is a growing demand for software services that allow companies to set up such predictive markets. Companies such as Consensus Point and Inkling and open source software such as Zocalo are competing for a pie of the business. Chris F. Masse, a financial consultant in Sophia Antipolis, France, who specializes in prediction markets, says: “By 2010, 10 percent of Fortune 500 companies will have gone public about their use of internal prediction markets, and probably another 10 percent will be testing some projects.”
So how accurate are these prediction markets? Research suggests that they are at least as accurate as, if not more than, other models of predictions employed by leading analysts. For example, the Iowa Electronic Markets set up by the University of Iowa College of Business to predict the outcome of the US Presidential Elections has been accurate to the extent of 98.63 percent averaged over three elections. Tradesports, a commercial marketplace, was able to predict every single of the 33 US Senate contests held in 2006 – a feat unmatched by any public opinion polls. Hewlett-Packard Co., in Palo Alto, California let selected individuals bet on future sales of some of the company’s printer products. They found prediction markets to be “a considerable improvement over the HP official forecast.”
The accuracy of the market depends largely on the fact that real money is being used to keep the bettors honest. The price one pays is set by the market’s opinion on the odds of that outcome. Even in corporate prediction markets, some sort of real incentives or trinkets are needed to prevent people from choosing random values out of sheer boredom or to advance their personal agendas. However, because in some scenarios, the people who bet are the ones who can change the outcome of the events, most corporate markets limit the maximum winnings an individual can receive.
Of course, some individuals will still think they know more than they really do and will make lousy bets. But this is by design rather than a failure. These erroneous bettors provide the market with fodder that the more accurate ones can use, and so long as the later group trumps the former with additional wagers of their own, prediction markets will continue to succeed. If everyone was an expert and knew the final outcome, there would hardly be any betting at all, and everyone would only win what they bet for in the first place.
To increase betting, the market bookmaker can contribute an initial amount of money into the pot or can provide subsidies to the bettors. Further, it is important to choose a diverse set of user pool in order to bring dynamism in the market. Every buyer must have a seller and vice versa for the double auction to work effectively. Even more important is the wording of the contract: Improperly worded contracts that leave things to open interpretation (vague definitions of success of an event), or those that don’t have clear cut objective outcome choices (“Will my wife get pregnant?”) will fail spectacularly; frustrating users who will lose money and no true prediction will come out of the exercise.
Despite all the positive evidence, many companies are still reluctant to use prediction markets. Managers have a hard time letting go of the fact that they can’t really control decisions that the market is predicting for them. It is difficult to see a negative prediction come true. Such a prediction can severely undermine a manager’s efforts to reverse the situation. For those companies that have no formal prediction methods in place, two major adjustments are necessary: one, deciding to make formal predictions about the future, and two, choosing to use prediction markets as the methodology. The first is obviously a larger issue, and needs buy-in from every stakeholder.
Further, it is difficult for vertically structure hierarchical organizations to accept the subversive outcomes of predictive markets. For example, increase of the market prices to questions such as “Should the CEO be forced to step-down?” when responded to by the workers of the company seriously threaten the established order. Truth, as they say, is bitter. Not everyone can digest the hard facts. Prediction markets can also be highly controversial, especially when tackling questions relating to politics or public policy and opinion. Another serious issue with accepting the predictions is that such markets are often counterintuitive. A group of people will always guess, fairly correctly, the number of jelly beans in a jar. Once managers start trusting the prediction markets and get past their initial doubts, they can create markets that can foretell problems when there is sufficient time to do something to rectify them. But such acceptance is hard to come by, given the human nature.
People often compare prediction markets to gambling, stating that they really are not any better off. However, they must note that stock markets too were initially thought to be worse than gambling. In fact, a lot of financial instruments, including short selling, options trading, commodity futures, derivatives and even hedge funds were thought of as illegal and unethical. However, these are not zero-sum games. They help the industry decide what to produce, how much to produce and when to produce. It is vital that the social utility of such instruments and markets not be lost amidst the fear of misunderstanding them.
