Trading on the Wisdom of Crowds

Most polls have the Democratic nominee winning in November, some handily, but here's a prediction you can bet on: John McCain is at nearly even odds to win the popular vote, at a 46% likelihood, up from the mid-30% range in the summer. This is the latest spread from the Iowa Electronic Markets, where traders put their money where their mouths are.

Political columns in these pages will cheer or jeer a close race. The purpose here is different: to embrace the counterintuitive notion that predictions by large numbers of people betting their own money have more value than ostensibly scientific measures such as polls. Markets rule, even in politics.

This could be the election cycle that legitimizes Web-based prediction markets. These services – which also identified Barack Obama's popularity among Democrats early on – show how the Web lets people combine their views so that, in the aggregate, they're eerily likely to get it right. James Surowiecki, in his book "The Wisdom of Crowds," points to predigital examples such as a contest at a country fair to guess the weight of an ox, which combined guesses from experts and nonexperts, resulting in a nearly perfect answer. This collective intelligence also accounts for why Google results, determined by an algorithm reflecting the popularity of Web results matching a search, are so relevant. Likewise, Wikipedia entries edited by many anonymous citizen "editors" can be remarkably accurate.

We think of forecasting stock-market performance and presidential outcomes as entirely different exercises. When stocks are assessed for future earnings, we don't look to opinion polls to define their price. Why shouldn't active, anonymous trading on politics, backed by money, work just as well in setting the odds of a political outcome?

Indeed, before polling became the accepted currency for information about likely voting results, in the 1940s, newspapers carried the latest betting odds as the most relevant source of information. Consider this Wall Street Journal report from Election Day, Nov. 7, 1916: "Odds on the election as quoted on the Curb showed a distinct tightening yesterday morning. There was a good supply of money to back both candidates, with Wilson offerings heavier than for some time past."

Some $165 million in today's dollars was wagered in Wall Street on the 1916 election – twice the amount spent on the election campaign itself. This suggests an active market, with known information well reflected in prices. And, indeed, Woodrow Wilson did narrowly beat Charles Evans Hughes.

"Betting odds are generally taken as the best indicator of probable results in presidential campaigns," this newspaper explained in 1924. Numerous firms "make a business of receiving and placing bets, and many of the principal banking houses call these firms up daily to secure information on the betting trend."

In contrast, polls depend on getting a representative sample, truthful responses to poll questions, and proper use of statistics. Election-predicting markets seem to work so long as there are enough traders whose aggregate information is fully reflected in bets. There have been cases where supporters of candidates have swayed the prediction odds, though economists note that this is more a risk when play money is used. (Rooting for a candidate is one thing; losing money on a bet quite another.)

As other forms of social media such as wikis are better understood, perhaps regulators will let predictive markets become more efficient. For now, these markets are suppressed by the Web's antigambling laws. The Iowa Electronic Markets operates through the University of Iowa, which managed to get a no-action letter in 1993 from the Commodity Futures Trading Commission, as an "experimental and academic program," with its bets limited to $500. More can be wagered through private operations such as Intrade, based in Dublin, but these services operate with a low profile and modest volumes. Intrade also powers a political futures market at using play money.

Iowa Electronic Markets claims that its results have been more accurate 75% of the time versus some 1,000 opinion polls since the early 1990s. Intrade's traders correctly called the results of 49 states in the 2004 presidential election. Academics analyze the topic in the Journal of Prediction Markets, and numerous blogs track the results. A group of economists led jointly by the American Enterprise Institute and the Brookings Institution has urged deregulation of predictive markets as different from gambling on poker or other games.

Maybe deregulation can create more-perfect predictive markets in time for the 2012 presidential contest. Meanwhile, candidates who rely on polls get second-best answers on how they're doing or what positions are helping their cause. A politician with faith in markets and the populist wisdom of crowds could look to predictive markets instead. Such a candidate, if one emerges, might just be the best bet.