InSightful
At 8:30 PM on Dec. 13, 2003, American forces captured Saddam Hussein in a compound outside the village of ad-Dawr, a surprisingly peaceful culmination of a months-long search. It seems that no one could have predicted this secret operation and its successful execution, except for the one company that did: InTrade.

InTrade was one of the early entrants in the emerging field of prediction markets, platforms that allow participants to buy and sell contracts on a variety of events. Currently, users can bid on contracts as diverse as the Vice Presidential nominee of the Republican Party, the odds of a U.S. recession in 2008, and Europe’s chances of hosting the 2016 Summer Olympics. At the culmination of the period, contracts clear at zero or 100 points depending on the outcome, with winners receiving the full value of the contract (typically $10). Prior to closing, traders buy or sell (short) the contracts at values corresponding to the market’s prediction of the likelihood of the event.

A Political Future
The most widely discussed use of prediction markets in recent years has been in forecasting election results. Despite various concerns as to their accuracy, market performance has historically predicted winners more accurately than traditional tools like pre-election polls or exit polls. In 2004, InTrade markets correctly predicted the presidential winner in every state. The 2006 Democratic takeover of the House and Senate surprised many pundits and political practitioners, but the markets had already shown Missouri and Virginia as likely Democratic Senate pickups. In fact, InTrade correctly predicted the outcomes of all of the Senatorial races.

Over the summer and fall of 2003, the value of the contract “Saddam Hussein will be captured or neutralized by the end of December 2003” hovered around nine points as the search appeared to be dragging on with no appreciable end in sight. However, on December 11th, despite no relevant publicly available news, the trading volume suddenly increased and the price jumped to 30. Two days later, American troops captured Saddam and the contract cleared at 100.

The uncanny ability of prediction markets to forecast events like Saddam’s capture, as well as their past successes in predicting election results, increasingly attracted public attention amidst the furor of the 2008 primary season. Alongside ardent supporters, market detractors surfaced with a series of worries. While proponents pointed to greater accuracy of the markets over polls, opponents cautioned against over-confidence, highlighting spectacular failures like the market’s inability to anticipate Hillary Clinton’s victory in the New Hampshire Primary.

Hype, Up; Market Failure, Down
Justin Wolfers, professor of economics at the University of Pennsylvania’s Wharton School of Business, analyzes the efficiencies and failures of political prediction markets and spoke to the HPR about some of his findings. “The less liquidity, the more open [the markets] are to manipulation, but there have been few real examples of manipulation. While someone could [manipulate prices] for a couple of hours or a day, it would be very expensive and very difficult to do for a number of weeks.” The efficiency of the market comes from its ability to self-correct.

While numerous academic studies show popularly cited concerns over manipulation and distortional effects of a small trading volume to be overblown, the political prediction markets do fall prey to some of the failures that occur in many markets. Their predictions most often fail when people over-respond to information coming in at the end of the contract period, and when predicting unlikely events.

“On election day, the markets go a little crazy and people make the wrong forecast or put too much stock in information that comes out at the last minute,” Koleman Strumpf, a professor of economics at the University of Kansas and expert on prediction markets, explained to the HPR. He cited the example of market over-reaction to the early exit polls from Ohio in 2004 (which showed Kerry likely to win) as emblematic. The other major market failure occurs when markets attempt to predict unlikely events. “The favorite long-shot bias (widely known to occur in other betting markets) places too much weight on an underdog winning—when markets get to ten or fifteen percent chance of winning, in practice those candidates almost never win.”

The Future of Futures Markets
While legal challenges (particularly surrounding internet gambling) may limit significant expansion of the current system, proponents are looking to other avenues for development. One such option involves popularizing usage of fake money markets, which site officials hope will produce more accurate results by widening the pool of participants. Other avenues for growth include emerging internal markets in companies, in which employees can bet to predict the timeliness of project completion or the success of a new product launch.

Beating the Alternatives
The main value of political prediction markets lies not in their absolute predictive ability, but in their forecasting accuracy relative to other options like polls or pundits. “Like the Dutch tulip market, or the real estate bubble, markets don't always get it right,” John Delaney, co-founder of InTrade, shared in an interview with the HPR. “These are prediction markets; they aren't guaranteed. A ninety percent probability is just that—one out of ten times it won't come through. Wouldn't it be more surprising if they came through all the time?”

While having a majority outside the margin of error in a poll indicates that a candidate should certainly win, a prediction market simply gives that candidate’s odds of winning. “[In New Hampshire] markets had her [Hillary Clinton] at seven, which means she had a 7% chance of winning, making her a long shot, but like in horse racing, sometimes long shots win. The odds made her an unlikely winner, but not impossible as the polls would have suggested,” Wolfers explains.

Empirical evidence widely supports the claim that markets outperform polls, for which Wolfers offers several reasons. “Markets include much of the same information as polls, as well as other sources, like who did well in last night's debate and what the pundits are saying. Markets have a better ability to draw on various information sets. Polls are quite crude, with sample sizes of only a few thousand.”

So while political prediction markets fail at times, often in highly publicized circumstances, in general they tend to be quite accurate, especially when compared to available alternatives. While imperfect, they can serve as a valuable predictive tool, especially when their predictions are taken as just that. John Delaney aptly summarizes the take-home message: “Are they perfect? No. Do they get it right most of the time? Yes. And they are better than almost anything else that is available."]]>