From Ring Knockouts To Court Deliberations: How Markets Predict

A prediction market is, by standard definition, a market that exists for the primary purpose of making predictions about a future event. In fact, though, any market can be employed to make predictions. If I know that the odds makers for an upcoming heavyweight championship fight are laying odds 1: 2 in favor of the reigning champ, then I can declare that the markets have ‘predicted’ his victory.

In the boxing-match case, prediction is a collateral consequence. The mechanism that creates those odds doesn’t exist to predict, it exists to keep the bets flowing, by creating as much demand for the challenger’s side of the proposition as there is for the champ’s side. If too many gamblers/voters favor the champ, even if for sentimental home-town-boy reasons, the odds will end up skewed in his favor. (The odds makers and their employers won’t be at all put out by that predictive failure.)

To get somewhat closer to a true prediction market (though not quite there) one can point to the Saddam Security. This was a market established by a sports-betting operation prior to the Iraq War that allowed for wagers on whether Saddam Hussein would still be the President of Iraq by a given date. If you were fully (100 percent) confident that Saddam would be deposed by a certain date, you might rationally have bet as high as $9.99 for each $10 incremental payout on his deposition and still secure a profit – assuming away transaction costs and the house’s take. If you thought it only 50 percent likely that his deposition would take place by a certain date, it would not have been rational for you to pay more than $5 for the prospect of a $10 payout. In this way, the Saddam Security market aggregated opinions – not any old shooting-the-mouth-off opinions, but the opinions of those who were willing to risk their money thereon.

At the start of 2003, the buying and selling prices indicated a majority chance (75 percent) that Saddam would be ousted by June 2003, but only a minority chance (35 percent) that he would be ousted by March 2003. As hindsight reveals, and as is usually the case, the long shot bet did not pay off, but a bet on the favorite did. Saddam was in fact formally deposed by the Coalition Provisional Authority on April 9, 2003.

Prediction markets have existed for years. The best known examples (especially the Iowa Political Markets) aim at predicting political events, such as election outcomes. Now, Jeff Joseph, Dave Karnes, and Daniel Diermeier plan to apply the principle of their operation to stock market indexes, via The Daily Delphi. Diermeier expressed confidence in a recent interview that they would have the system up and running in September 2012.

It will work in this way. Every day, before the market opens, TDD will ask its users where they believe the market, as measured by a specific index, will close that day, their degree of conviction, and whether they plan to trade the market that day. TDD will then aggregate this information using proprietary algorithms. Every user will receive the results: a forecasted daily closing price, a degree of conviction, and the range of the forecasts.

Diermeier calls this index-tied use the “first iteration” of a system that will later find other applications.

Consider the sort of information that was coming out of the Supreme Court building in Washington as oral arguments over the Patient Protection and Affordable Care Act (ObamaCare) proceeded last week. A lawyer, arguing that the recently enacted health care bill is constitutional under familiar commerce-clause precedents, found himself on the receiving end of rather aggressive questions from a certain Justice – a Justice with a reputation for being the pivotal figure in close cases. Instantly, BlackBerrys and iPhones conveyed this news to the world.

Suppose, then, that you have a substantial investment in a health insurance company, or in a pharmaceutical concern, or in any of the many industries that may be affected directly or indirectly by such a decision. Should this new datum matter to you? And, if so, how do you figure out how much it should matter?

This is precisely the sort of issue, Diermeier says, that might be incorporated into later iterations of The Daily Delphi. It might be re-worked to make predictions on any number of court cases with significant impact on businesses, in such fields as antitrust, intellectual property, tort liability, merger approval, and so forth. Good applications of TDD’s model arise “where there is deep expertise, but where aggregation mechanisms are lacking.”

There are other approaches investors might take. But the ObamaCare arguments illustrate some of the difficulties with them. Consider those tough questions from a potential swing-vote Justice. Diermeier pointed out that if you wanted to do the predicting yourself, rather than farming it out to an aggregating market, “You’d want to know, for example, whether this Justice was in the habit of asking such questions, or whether they were unusual in this case. You’d want to have a data base that could let you correlate the number and nature of the questions he asked with the way he voted in the resulting decision, and perhaps also the influence he was likely to have with other Justices in deliberations.”

That can become very tricky. “A predictive market provides an alternative to all that,” he said.