The Science Behind PPX

The Election-Night Crystal Ball The betting site intrade.com offered a security that would pay out $100 if the Republicans held their majority in the U.S. Senate and nothing if they lost. For weeks, it had been trading at around 70, reflecting the consensus view that there was a 70 percent chance that the Republicans would hold power. But as the results came in, the market-not the media-was first to smell the upset. Paul Wootton

Here at PopSci.com, we've launched the PopSci Predictions Exchange–the first online market dedicated to trading the future of science and technology. It's totally free, and you don't need any experience to sign up (although reading this feature on the science that makes the market tick might help a bit). Check out ppx.popsci.com after you're done here to begin trading today.

Should you believe anything that NASA announces? More specifically, should you believe that anything the space agency says will get done by a certain date will in fact happen by that date? NASA's track record is not good. Shuttle launches are almost invariably postponed, by an average of about two weeks. The International Space Station, which depends critically on the shuttle for its construction, was supposed to take five years and 37 shuttle flights to build. It's been a decade, and we're 16 missions away from finishing a vastly pared-down project. So when NASA says it's going to finish building the ISS by the end of 2010 (as indeed it must, because that's the last time the shuttle will be allowed to fly), you could be forgiven for questioning just how likely that really is.

But how can we know? How, for that matter, can NASA know? Building a $100-billion multinational orbiting space platform is, after all, a rather complicated business; uncertainty is everywhere. One of the shuttle's many thousands of parts could malfunction, or Congress could order a review, or an ill-timed storm could strike central Florida. And sometimes the worst happens, and tragedy delays the program for years.

There are far too many variables for any single person-not the NASA administrator, not the shuttle program manager-to fully comprehend. But if no one can absorb all the available data, how can anyone expect to make accurate predictions about what that data implies? How can we know what's going to happen if we can't even know what's happening now?

Here's how: We can ask the market.

ENGINEERING AN ORACLE

A prediction market is like a stock exchange, except people trade not stocks but predictions. The first and most famous prediction market, established at the University of Iowa business school in 1988, was designed to forecast the outcome of that year's presidential election-which it did, with remarkable precision. Anyone could join the market and buy or sell propositions such as "What percent of the popular vote will George H.W. Bush receive in the presidential election?" Traders who thought Bush would get 60 percent or more of the vote would buy the shares if they were less than $60. Traders who disagreed would sell. The market price reflected the consensus view, which turned out to be more accurate than any of the six polls released in the week before the election. But if predicting a landslide Bush I victory over Michael Dukakis seems like fish-in-a-barrel stuff, consider that the Iowa Electronic Markets have correctly predicted the winner of the popular vote in every national election since then, and with greater accuracy than the Gallup poll in all but one of those years. It achieved this success by relying not on the oracle-like knowledge of a single "expert" or small group of pundits, but on the power of the market to efficiently gather the intelligence of traders and agglomerate them into collective wisdom.