American Scientist

Those who champion public support for basic scientific research often point out that if the government didn't fund fundamental work in certain fields, nobody would (a few wealthy patrons aside). Indeed, by some definitions, the lack of commercial interest in an area of scientific study is what delineates basic from applied research. But an economic vehicle of growing popularity—prediction markets—stands to turn this notion on its head.

Although markets in which traders attempt to profit by predicting the likelihood of future events have existed in one form or another for decades, they have truly taken off in the past few years. They are commonly used, in essence, to replace opinion polls and often prove more accurate than such surveys because there are concrete costs and rewards associated with being right or wrong.

One of the pioneering prediction markets is an entity called the Foresight Exchange, where buyers and sellers trade contracts whose value depends on various propositions about the future—some of which are purely matters of scientific interest. For example, the Foresight Exchange trades contracts that will pay owners at a future date should at least one gamma-ray burst be discovered to have originated within 10 megaparsecs of Earth. The value of such contracts fluctuates with the consensus of the scientific community about whether such a discovery is likely to be made.

In the late 1990s, astronomers came to believe that no gamma-ray burst could have originated in Earth's cosmic neighborhood, in part because such a burst might well have destroyed life on Earth. The market reflected this appraisal, valuing contracts on the possible nearness of a gamma-ray burst at only a small fraction of the payout value. But in 2002 the person judging this claim for the Foresight Exchange noted that astronomers have come to believe that some large spinning stars release bursts of gamma rays along their rotational axes when they explode as supernovae. Thus it has become reasonable to suppose that such a supernova could have occurred within 10 megaparsecs, shooting its lethal gamma rays in some other direction and sparing Earth—yet still qualifying as a nearby gamma-ray burst. The judge's statement apparently caused the value of the bursts-are-near contracts to rise. They now trade at more than 80 percent of their payout value.

The rub is that the Foresight Exchange operates with play money, not real dollars. It must restrict itself in this way, because markets for predictions about the future fall into an awkward legal limbo. These operations share some elements with gambling, most forms of which are prohibited by state laws, and some with futures trading, which is regulated at the federal level by the U.S. Commodity Futures Trading Commission.

But suppose for a moment that it were perfectly legal to run a scientific-prediction market with real money. An enterprising astronomer might use it to fund her research. Tom W. Bell, a professor at the Chapman University School of Law in Orange, California, outlined how this might be done (using a different example) in a 2002 article in the Chapman Law Review, one titled, "Gambling for the Good, Trading for the Future: The Legality of Markets in Science Claims."

It goes like this: The investigator would approach a bank, asking it to issue, for example, 1,000 certificates that would each pay, say, $100 to the bearer if by a specified date the hypothesis is judged to be true—and an equal number of certificates that would pay $100 if the proposition is judged to be false. The bank would charge only 1,000 × $100, or $100,000, (plus, presumably, a modest fee) for doing so, knowing that, no matter what happens, it will only have to pay out on half the notes.

The market-savvy astronomer could then polish up her theory that proves beyond a shadow of a doubt that no gamma-ray burst could possibly have originated close to Earth. Rather than publishing it immediately, she would sell certificates good for redemption if gamma ray bursts are discovered to have gone off nearby. Because that proposition is consistent with the prevailing consensus about what is possible, she might get a decent price, perhaps $50. With the $50,000 dollars she earns in this way, she writes a stunning paper (perhaps while vacationing in Hawaii) and submits it to Nature. As soon as it is published, the community is swayed by her arguments, and the price of contracts that pay if gamma-ray bursts do not prove to come from nearby shoots up to $99 (while the price of those that she had just sold for $50 drops to $1).

She can now sell her remaining contracts for $99,000 and easily pay back the second mortgage she took out to fund the issuance of all these certificates in the first place. And she'll have almost $50,000 left over to pay for the research needed to come up with her next blockbuster idea in astrophysics.

Could such a wild scheme for funding science ever be possible? Perhaps, but not unless the legal status of prediction markets is sorted out. Right now, the only real-money prediction markets operating in the United States—the Iowa Electronic Markets—are run by the University of Iowa as an academic exercise, with rules that limit the amount of money each participant can put at risk to $500.

Forrest Nelson, a professor of economics at the University of Iowa, helped to found the Iowa Electronic Markets. He and his colleagues faced considerable push-back from their university's lawyers in setting up a real-money exchange, but they persevered and eventually were granted "no-action" letters from the Commodity Futures Trading Commission. Why go to all that effort? "Economists think that incentives matter," he says, noting that when traders use play dollars, "It's all just matchsticks."

Nelson does, however, accept the findings of studies showing that the predictions of some play-money markets are just as accurate as their real-money counterparts. But he points out that to get a good answer with play money requires more participants to average out a lot of ill-conceived and extreme positions. Play-money markets, Nelson says, tend to show excessive volatility.

Nelson's Iowa group is mostly known for dealing in future election results. But in recent years it has set up markets for scientific predictions too, ones about public health (influenza outbreaks) and meteorology (hurricane landfalls). For example, the price for a futures contract that pays 10 cents should the World Health Organization confirm 400 cases of H5N1 "bird flu" before July 1, 2008, was running at 6.1 cents on the Iowa Health Prediction Market when this article went to press. This valuation suggests that the group of expert epidemiologists invited to participate feel that the proposition has a 61 percent chance of proving true. But with them gambling only a few dollars of somebody else's money (in this case, that of the project's sponsor: the Robert Wood Johnson Foundation), it's hard to know how much serious study goes on behind the trading. Presumably, the results would be much more reliable if participants stood to make true fortunes with correct predictions.

At a February conference held at Google's headquarters in Mountain View, California, Bell suggested to his audience of mostly programmers what he called "a legal hack," a maneuver that might protect the organizers of prediction markets from prosecution—at least for ones set up and funded by a single philanthropic sponsor. The first step, Bell says, is to establish an in-house market that deals in contracts clearly related to some public good. For example, Google could sponsor a market in contracts that pay if earthquakes of certain magnitudes strike the Bay Area within specified time frames. Google employees would be required to play the market using play money ("gbucks," say), and those who made the best predictions would be awarded real-money bonuses at the end of the year.

Bell suggests that Google could then ask the courts for a declaratory judgment about whether their in-house market in earthquake futures is legal. Presumably, the answer would be yes—play-money markets with prizes being common these days.

The next step, Bell says, would be to "expand the bounds of the firm." Google would enlist scientific experts to play, engaging them nominally as independent contract researchers, but they would receive only a pittance for playing, perhaps 1 cent per trade. Those who earn the most gbucks in a year would, however, be rewarded with lucrative bonuses. In this way, the nominally play-money market would provide sizeable real-money incentives to come up with good scientific predictions. Bell suggests that Google could then approach the courts again to obtain a second declaratory judgment—saying in essence, "I'm just double checking here"—because the operation of this prediction market hasn't changed in any fundamental way since the first judgment.

Bell notes that, strictly speaking, Google would need to do this in all 50 states. But he says that, in actuality, it would probably be sufficient to obtain court judgments in California alone, because other states would be likely to follow suit.

Will such legal calisthenics be necessary? It's easy to see how robust markets in scientific predictions could provide a valuable public service, but it indeed remains difficult to forecast whether larger society will soon embrace these operations or treat them as so much highfalutin bookmaking.  It wouldn't surprise me, though, if someone soon set up a market to help gauge the answer to that very question.