Online Betting Markets Beat the Odds
They have foretold the outcome of every presidential election since 1988, correctly selected Oscar winners in advance of the Academy Award ceremonies and predicted rate increases by the U.S. Federal Reserve.
No, "they" are not psychics with extrasensory perception. They are prediction markets, online betting venues that allow people to wager on the likelihood of financial, economic, political, cultural, social or scientific events.
Operating under names like LongBets, NewsFutures and TradeSports, prediction markets have demonstrated an "uncanny ability to make accurate predictions" in many areas, says Russ Ray, U of L professor of finance.
Ray says prediction markets succeed because they can flush out insider information and concentrate it in one location. Laws prohibit insider trading in the stock market and other trading markets, but there are no such restrictions in unregulated prediction markets.
![[Image]](images/russ-ray.jpg)
College of Business professor Russ Ray studies the brave new world of prediction markets.
The first prediction market was the Iowa Electronic Market, started in 1988. The market allowed anyone in the world to bet up to $500 on the presidential election. This prediction market not only has correctly predicted the outcome of every presidential election, it also has been more accurate than exit polls, expert opinion and other traditional forecasting methods at predicting the percentages of votes garnered.
"The success of the Iowa Electronic Market opened up interest in other areas for prediction markets," Ray says.
The markets work much like the stock market. A person who wishes to make a claim, such as "The Dow Jones Industrial Average will reach 12,000 by Dec. 31," offers to sell "shares" of the claim. Most sites limit claims shares to be sold for no more than $1 each. The person sets a bid price and adjusts the price until someone agrees to buy a claim. Buyers may purchase as many claims shares as they want.
If buyers agree to buy the claim for 52 cents, they send that amount to the prediction market, which forwards the bets to the claim maker who then sends $1 for each claim purchased to the exchange to cover the possibility of winning bets. The current price of the bid indicates the probability that the event will occur (52 cents indicates a 52 percent probability).
If the claim comes true, $1 is awarded for every claim share bought. If the claim does not come true, the claim seller gets back the money put up to cover the bets. In one case, the claim seller keeps the money received from buyers.
Large corporations are taking notice. Microsoft, Eli Lilly, Hewlett-Packard and others have created internal prediction markets accessible to top managers and executives to forecast sales, revenue, product success and more.
In internal corporate prediction markets, managers are given points to bet with no monetary value. The aggregation of the points tells the firm's executives what the managers truly forecast.
Ray says these forecasts are accurate because they are based on expert, inside knowledge and are done anonymously and with impunity. Managers need not fear retribution for poor forecasts or contradicting their immediate supervisors, so they can speak freely and give honest assessments.
"Hewlett-Packard found that sales forecasts made through internal prediction markets were more accurate than publicly announced official forecasts made through traditional forecasting methods 15 out of 16 times," Ray says.
The Pentagon tried to start its own Terrorism Futures Market in 2003 for people to bet on future terrorism events but Congress and the public found the idea distasteful and it was nixed. Nevertheless, Ray says it might have flushed out real information to predict and possibly prevent attacks.
Because of their accuracy, Ray says traders, speculators, regulators and policy makers should incorporate the information from these markets in their decision-making processes.
"A prediction market shouldn't be the only variable on which they should make their decisions," he says, "but it should definitely be one of the variables."
