Much has been made of how a multitude of polls could be so wrong in showing that Barack Obama would easily defeat Hillary Clinton in the New Hampshire primary. Even more interesting, however, may be that prediction markets – where people actually invest in their beliefs, not just share them with a pollster – got it so wrong as well.

Prediction markets, although not well-known among the public, are numerous and growing. They cover everything from sports scores to election results to whether Osama bin Laden will be captured to whether some celebrity couples will wed before others. Some allow participants to bet on outcomes with virtual money with prizes awarded to those whose predictions are closer to reality. Others involve real money.

One of the oldest is the Iowa Electronic Markets, run by the business school at the University of Iowa and open to anyone. Traders can invest up to $500, buying and selling shares based on who they believe will win elections.

On the eve of the New Hampshire election, Sen. Obama’s shares were valued more than three times as highly as Sen. Clinton’s in the Iowa Democratic Nomination Market.

The Wall Street Journal Political Market gave Sen. Obama an even greater advantage. It showed the Illinois senator with a 92 percent chance of winning New Hampshire, compared to Sen. Clinton’s 7 percent.

This led Justin Wolfers, an assistant professor at the University of Pennsylvania’s Wharton School and national expert on prediction markets, to proclaim in the Wall Street Journal that “it is no exaggeration to term the results truly historic.” Because Harry Truman was rated as having an 11 percent chance of winning the presidency in 1948 compared to Sen. Clinton’s 7 percent in New Hampshire, professor Wolfers termed Sen. Clinton’s New Hampshire victory more significant than Truman’s presidential win 60 years ago.

The markets, of course, now reflect Sen. Clinton’s victory, surprising or not, and give her a 6-in-10 chance of winning the Democratic nomination, up from as low as 1 in 4 before Tuesday’s New Hampshire vote.

All the polling and trading can’t compensate for the fact that it is humans, who are notoriously difficult to pin down, who ultimately cast the votes. But a closer look at the New Hampshire results shows weaknesses in a market approach to predicting who will be the next president.

Sen. Clinton outpaced Sen. Obama among women, those who earned less than $50,000 a year and those who had not graduated from college and older voters. None of these groups is

typically largely represented in

prediction market traders.

Although prediction markets, over time, have better records at foreseeing outcomes than polls or professional analysts, the New Hampshire experience shows that predicting human behavior remains a gamble.