The Fizz-dom of Crowds
If prediction markets are so great, why have they been so wrong lately?
When political junkies first encountered prediction markets in the early 1990s, it was as if the opinion poll had been perfected. Academic studies showed that the markets—in which individuals bet real or fake money on futures contracts that expire when an election is over—consistently did a better job of predicting electoral outcomes than even exit polls. They've been an important part of Slate's political coverage for years.
But lately, prediction markets have been getting some big questions very wrong. Hollywood Stock Exchange and its cinema brethren are great at some things but have consistent trouble making accurate best supporting actor and actress predictions, including this year. In January, Slate's Dan Gross wrote about how the prediction markets failed during the crucial New Hampshire Democratic primary. Instead of beating the conventional wisdom, the prediction markets trailed it. The Intrade contract for Obama to win the New Hampshire primary rose as high as 95 cents; on the day that voters went to the polls in New Hampshire, Intrade still had Obama's chances of being the Democratic nominee as more than 70 percent. But as results favoring Hillary Clinton began to come in, the contracts became a dead heat. That's hardly a prediction—it's simply tracking the day's results.
On the evening of the California primary, the prices on InTrade indicated that Obama would win; he didn't. When Barry Ritholtz, who runs a quantitative research firm called FusionIQ and writes the popular Big Picture financial blog, wrote about New Hampshire in January, he reminded us that TradeSports had also failed to correctly predict the Democrats' recapturing the Senate in 2006, and that the Iowa Electronic Markets had Howard Dean as the favorite to win the Iowa caucus in 2004. Citing Ritholtz, the New York Times declared in February that "a little backlash has begun" against prediction markets.
So, what's going on? Is there something fundamentally wrong with this one-time darling of pop economists? Here's an examination of the leading pet theories for why prediction markets fail:
They're too small. That's one of Ritholtz's biggest gripes. Political-prediction markets, he wrote in January, "are thin, trading volumes are anemic, the dollar amounts at risk are pitifully small. Thus, these markets are subject to failure at times."
There's no question that tiny markets are bound to yield inadequate results. Yet by the usual standards of prediction markets, the trading on political futures markets has been pretty robust. In the days before the Iowa caucus, for example, several thousand dollars changed hands in the Iowa Electronic Markets, and tens of thousands of contracts were sold.
Yes, that level of participation is far lower than, say, the 2-billion-plus shares that are traded on the New York Stock Exchange most weekdays. But it's not so small as to alone make a prediction market inaccurate. Thomas Malone, who studies prediction markets at MIT's Center for Collective Intelligence, says that his rule of thumb is that only "somewhere around a dozen" participants are necessary to give a prediction market sufficient liquidity.