The Prediction Markets Stumble

The "wisdom of crowds" crapped out in Election 2006.

Katherine Mangu-Ward | November 13, 2006

It's not often one finds oneself in a position to eat crow on behalf of more than 50,000 people—but here I am. My beloved election prediction markets have failed me this cycle, and so I'm here to apologize for the inaccuracy of the predictions I stole from those markets, and for the failings of the tens of thousands of traders who were supposed to illuminate the truth for us all. 

Usually, when people—especially pollsters—predict election outcomes wrong, they say something like "No one could have guessed that the turnout would be so high in the southwestern part of the state" and leave it at that. They don't get fired, and they only occasionally even concede that they might have something to apologize for. And maybe they don't. After all, they're in the business of entertaining and informing, but perfect accuracy isn't required.

For a brief, shining moment last week, politics junkies had lots of facts at their disposal, and it made them giddy. Judging by the frequency with which polling data was released, Americans spent the entire first week of November answering pollsters questions about their voting habits, opinion on the issues, and preferences for the more handsome candidate. Oceans of data poured forth from every television, every political blog, every newspaper. With all that information, talking heads swelled. Already confident in their predictive abilities, the presence of such a profusion of information sent election season seers into a frenzy of confident explication and prediction. Conversations about politics inevitably degenerated into poll swapping—"Zogby says Allen is ahead." "Yes, but CNN's poll favors Webb." And when things got really desperate (either on air or over drinks) surprising fervent discussion of margins of error would flame into existence.

Thorough all this chaos, I calmly cited online election prediction markets. Standing zen-like above the fray in that chaotic week, I'd casually drop a mention that "InTrade has the likelihood of Republicans holding the Senate at 70 percent." I delivered tiny, smug lectures on the superior ability of markets to aggregate information, name-checking Hayek. I sat aloof, murmuring the old TradeSports motto to myself "Put your money where your mind is."

This strategy has served me well in the past: In 2004, InTrade traders correctly called all 50 states the weekend before the election. The Iowa Electronic Market, a highfaultin', academic futures market in the same vein, has frequently boasted a lower margin of error than polls since its creation in 1988.
 
But this time around, InTrade and its other electronic market cousins didn't acquit themselves very impressively. Iowa Electronic Markets had Republicans holding the Senate, with an all-Republican Congress trading high for most of the history of this cycle's market, and an all Democratic Congress trading very low—below 20 percent probability for most of the market's history. And while markets devoted to individual races in the Senate tended to be correct, the overall prediction markets for the balance of power in the upper house miscalled the race. These outcomes were still better than a heck of a lot of pundits, but not good enough to justify my serene pre-election confidence.

Weirdly, the McLaughlin Group, the fustiest of all the talking head shows, had one of the best records this cycle, with Eleanor Clift, Lawrence O'Donnell, and John McLaughlin all predicting Democratic takeover of the Senate and calling nearly all of the close races correctly. Still, that old line about stopped clocks comes to mind.

Now my favorite moment of the political cycle is upon us: Total ignorance. The whole ecosystem of pundits, pollsters, and politics-obsessed bloggers has been floating in a lush primordial soup of political data. But information-rich environments are not the natural habitats of such creatures. As soon as the marathon election night coverage ends ("We are calling the Senate for the Democrats") the punditry returns to a state of Eden-like ignorant bliss, unburdened by data.

And for now, I'll go back to my old strategy. After all, markets still have a better track record than pundits. Sen. John McCain (R-Ariz.) is trading at just over 53 cents on the dollar at InTrade today as the prospective 2008 Republican nominee. After a lull during the congressional election when investors were in wait-and-see mode, McCain's chances spiked up over 50 percent, as Allen contracts plummeted to less than one cent on the dollar.

And InTrade has Sen. Hillary Clinton (D-NY) on a steady upward trajectory since early September. After virtually no trading in the first week of November, she's up over the 50 percent mark for the Democratic 2008 nominee. So, McCain versus Hillary it is.

Fortunately, even if I'm wrong this time around, I don't have to do more than offer a casual apology. I'll know not to be quite so cocky, and all those stupid traders I had to apologize for this time around have lost all their money, so they'll be sufficiently chastened (and impoverished) when considering their bets for 2008. 

Katherine Mangu-Ward is an associate editor at Reason.