Rolling Craps in New Hampshire

The storied prediction markets fared little better than the polls

Journalists, politicians and statisticians continue to scramble to figure out why, exactly, the polls predicting an easy victory for Barack Obama going into the New Hampshire primary on Tuesday were so terribly wrong. But polling was not the only device that failed to foresee Hillary Clinton’s comeback.

Prediction markets-which ask people not who they think will win, but who they would bet on to win, and which are often extolled as more accurate than polls-were also wrong. As my colleague, Liz Cox Barrett, reported Wednesday, part of the scramble to figure out what happened involves a certain amount of “self-flagellation,” and John Tierney of The New York Times is now doing his share. On Monday, he predicted on his blog that Obama and John McCain would take New Hampshire for the Democrats and Republicans, respectively. His prophecy was based on tips from anonymous sources at Intrade, a leading online prediction market that allows people to bet on a number of future events, from elections and business deals, to military actions and celebrity life. On Monday, Tierney called Intrade “the greatest time-saving invention of this century” when it comes to analyzing the presidential horse race.

“I guess I picked the wrong day to tout the Intrade futures market,” he reflected in another blog entry, posted Thursday in the wake of the upset. The market’s poll-like failure is not enough for Tierney to abandon such devices altogether (Intrade was remarkably accurate in the Iowa caucuses), but he wrote that he will consult alternate sources of information in the future before hazarding an electoral prediction:

As one Lab reader, Tony, observed after the results came in last night, the traders apparently fell victim to the same Obama-is-unstoppable cascade that had journalists and politicos writing off Hillary Clinton’s chances.

But the traders did at least realize their mistake even before the polls closed (presumably because some of them were insiders who got hold of the closely guarded exit polls), and Mrs. Clinton’s upset victory was obvious on Intrade long before it was proclaimed on television.

Tierney is not the only journalist likely to continue to rely, to some degree, on prediction markets. Last year, Slate launched a feature that will remain active until the 2008 general election called “Political Futures: Betting on the Election,” which amalgamates the data from four prediction markets, including Intrade and the largest of such operations, the Iowa Electronic Markets, which was founded in 1988. The introduction to Political Futures on Slate’s Web site reads, “If a single prediction market is wiser than the pundits and the polls, imagine how wise all the prediction markets are together.”

Ultimately, points out David Plotz, Slate’s Washington editor, Political Futures is just one element in the publcation’s larger electoral package, which also includes an amalgamation of polling data called Election Scorecard, a database on candidates’ campaign travels, and, of course, news coverage. “All these things are components of how you can look at the campaign,” Plotz said. “Anybody that says that prediction markets can predict everything is wrong. What we’re looking at is giving readers all the different points of entry.”

Prediction markets do not receive anywhere near the media attention that polls do. Similar devices, called futures markets, have been common in the business and commodities world since the mid-nineteenth century, when standardized contracts were introduced in Chicago for trading. Today, when the press writes about prediction markets, it often focuses on some of the odder items to bet on (the Hollywood Stock Exchange, in which traders use play money to wager on the outcome of films, is one of the largest and most popular prediction markets).

But in December, Justin Wolfers, an associate professor at the University of Pennsylvania’s Wharton School of business, published an op-ed in The Wall Street Journal that went into some detail on prediction markets’ performance in recent elections. Wolfers points out that they foresaw the collapse of Howard Dean’s candidacy and the Electoral College winners in every state in 2004, as well as the victor of every Senate race in the Democratic sweep of 2006. The traders that ended up disappointing the Times’s Tierney this week might have bet more cautiously (or avoided some measure of Obama-hysteria) had they read this paragraph from Wolfers’ column in advance of the primary:

In a truly efficient prediction market, the price will come to reflect the influence of all available information. For instance, those discouraged by Ms. Clinton’s recent polling in New Hampshire are probably selling, while those who believe endorsements by the Iowa Register are crucial are buying; Ms. Clinton’s campaign to increase her likeability may lead some to buy, while recent mis-steps by her campaign may lead others to sell.

Unfortunately, by the eve of the New Hampshire primary, Wolfers was back in the Journal, writing this time that the newspaper’s own prediction market, WSJ Political Marketplace, run by Intrade, was showing that New Hampshire might be the “death knell” for Clinton and a couple other candidates. After a bet like that, in Vegas they’d say, “craps.”

Has America ever needed a media watchdog more than now? Help us by joining CJR today.

Curtis Brainard writes on science and environment reporting. Follow him on Twitter @cbrainard.