Gambling dressed up in a Wall Street suit?

WIRED:AFTER THE excitement of this week's Democratic nomination battle, eyes in the US are slowly beginning to turn to the real fight: the presidential race itself, and the battle for Congress that will take place on the same November day.

Business decision-makers all over the world crave an accurate prediction of the outcome - and none of us are short of polls and pundits willing to spout data for and against each possibility. But how do you filter all these predictions for bias and error? Anyone who has a preference (or a pet conspiracy theory) is likely to bend the facts to match what they desire, or what they deliciously fear might happen. Pollsters aren't all created equal, and even those who consistently misread the voters still stay in business.

Even the supposedly impartial press likes a fight, and tends to declare races closer than they really are. One approach that has received a lot of attention these days - with thousands of Americans visiting a Dublin-based site to gauge the will of their own people - is the use of prediction markets.

Depending on how you look at it, prediction markets are an ingenious extension of the stock market to cover contemporary events, or just gambling dressed up in a Wall Street suit. Many in the US think their local law may take the latter view, which is why one of the most popular prediction markets, Intrade, is based in Ireland, a safer offshore location for real-time wagers.

On Intrade, you don't bet using bookmakers' odds for an event such as the election - you trade in share-like "contracts", which either pay up or turn worthless when a certain result is called. Just like on the stock market, the price of these contracts depends on supply and demand. A "McCain wins the presidential election" contract pays out $10 in November if McCain wins the vote, and nothing at all if he doesn't. Right now, the price of a McCain contract is hovering about $4.80 (€3.15), showing that investors are edging very slightly in believing he will lose - but not much. Think of it as the Intrade market saying collectively that there's a 48 per cent chance of McCain winning.

Who creates these contracts? Intrade does, and it can make a profit off the gambling by charging transaction fees and keeping the interest on the first raft of purchases. If it hands out McCain wins/McCain doesn't win contracts equally, it will have to pay back half of them in the end at a full price - but that cost is balanced by all the now worthless anti- or pro-McCain contracts it also sold.

Prediction markets have been praised for their predictive accuracy - most notably in James Surowiecki's recent book The Wisdom of Crowds. But like the markets they resemble, they're not perfect. The betting on last year's Senate elections firmly predicted a Republican majority up until the last few hours of voting, when exit polls began to reveal the extent of the Democratic rout.

On the other hand, very few reliable indicators outside the markets indicated a Democratic win either. And those two facts are linked. Don't think of prediction markets as a new source of information, but an excellent summary of the current viewpoint. The buyers and sellers on the market are taking their knowledge, betting on it, and in doing so, neatly summarise the degree of their confidence in a tidy number.

Are these gamblers any better or worse informed than the average person on the street? Individually, probably not - but these markets are at least somewhat self-correcting. A fool in this market really is separated from his or her money: "Those who are right make money from those who are wrong," as Robin Hanson, one of those behind the idea, once said. The consistently wrong are chased from the market; those who have a pattern of guessing right stick around for more.

Prediction markets aren't always public. Companies such as Google, Chrysler, Microsoft and Nokia have run internal prediction markets for several years.

Google's published results show that its internal market is accurate (that is to say, a contract priced at 20 per cent will come true roughly one in five times), but often has an optimistic bias. New Google employees err on the side of optimism; others bet more optimistically when Google stock was appreciating.

Online prediction markets suffer from all of the distortions, self-deception and trickery of the real markets. Profligate types spend their money attempting to make the odds look in favour of their favourite cause; fads and herd-emotions warp the prices; insiders can profit, and, theoretically, malefactors can profit from sabotaging a real event while betting against it.

Most damaging, their audience - whether limited to Googlers, or limited to inveterate gamblers who understand the stock market - is not exactly a broad spectrum of world opinion. But if you like the idea of them as an accurate barometer of current opinion, there are worse places to gauge the probabilities of modern politics. And if you disagree - well, perhaps you might make some money betting against them?