Taking stock of political markets
Posted: Wednesday, June 04, 2008 7:20 PM by Alan Boyle
IEM
This chart traces the political fortunes of Democratic candidates on the Iowa
Electronic Markets. Yellow stands for Barack Obama, blue stands for Hillary
Clinton, green stands for John Edwards and red stands for the rest of the field.
|
Now that the presidential primary season is over, economists are analyzing how the political prediction markets sized up against the pollsters - and looking ahead to the bigger campaign ahead.
Like the polls and the pundits, the markets were sometimes thrown for a loop during this season of political surprises. But in the judgment of economist Justin Wolfers, who has been monitoring the ups and downs of political fortunes for years, "the markets got it the least wrong." And the markets already have picked a clear favorite for the White House prize.
Prediction markets let investors put money down on the chances that a particular person (or party) will win out in a campaign ... or, for that matter, the chances for an actor to win an Oscar, or for a football team to win the Super Bowl. It sounds a lot like betting - but the difference is that shares in the proposition can be bought or sold (at higher or lower prices) as long as the futures contract is open.
If the proposition pans out, each share pays a set amount - for instance, $1 on the Iowa Electronic Markets, the only prediction market that has the federal government's blessing to operate in the United States. If the proposition doesn't come true, the shares become worthless.
The shares for Barack Obama, who became the Democrats' presumptive nominee on Tuesday, are worth more than 90 cents each. The shares for Hillary Clinton, his biggest rival for the nomination, are worth less than 10 cents each. However, those prices didn't just get to be that way on Tuesday. They've been virtually unchanged for the past few weeks.
"To our market, yesterday's news wasn't really news," IEM director Joyce Berg, an accounting professor at the University of Iowa, told me today.
Similarly, the prediction markets have favored the Democrats over the Republicans for months. Wolfers, a professor at the University of Pennsylvania's Wharton School, said he was struck by the way the market for the general election race reacted to Tuesday's news.
IEM
This chart from the Iowa Electronic Markets shows the values for shares that would
pay off in the event of a Democratic presidential win (in blue) vs. the shares for a
GOP win (in red). The trend line indicates that the Democrats have consistently
been given more of a chance for victory, although the margin has varied.
|
"The Democrats' chance of taking the White House jumped by quite a large amount, considering that not that much actually happened," he told me today.
He speculated that investors were reacting to the fact that Clinton seemed to be taking her setback in stride, even though she hasn't formally conceded the nomination. "The real news is that it appears this thing is not going to drag on until the convention," Wolfers said. "She has not come out fighting."
Are markets magic?
Unless something happens to shake up the markets, which include offshore establishments such as InTrade as well as the IEM, the trend appears to point toward a future Obama administration. But how good are those markets, really?
Last month, Wired magazine's John McQuaid wrote that prediction markets are going into the general election campaign "battered from a tough primary season." He cited New York Times columnist Paul Krugman's observation that the markets "know no more than the conventional wisdom."
However, the numbers tell a different story. Krugman wrote what he did just after Clinton's win in January's New Hampshire primary - which was not predicted by the pollsters or the pundits. In contrast, "the market said it was maybe a 1-in-14 chance," Wolfers said. High odds, to be sure, but not all that high.
No one really claims that the prediction markets have some mystical power to aggregate information. But year over year, the data indicate that prediction markets work better than political polling, perhaps in part because the participants in the market have to put money where their mouth is.
That record has been borne out by research going back to the Bush-Dukakis race back in 1988, said George Neumann, an economics professor at the University of Iowa who was a co-founder of the IEM and is now on the venture's board of governors.
"Our standard for many years was to compare ourselves to the polls, but that's just like shooting fish in a pond," he told me today.
The numbers game
This year, the Caveat Bettor blog compared InTrade's predictive powers against the Zogby polls, and last month InTrade was declared the winner with a record of 7-3 vs. Zogby's 3-7 (plus 11 state-by-state "ties"). The result was "unsurprising to those who know a little bit about the scholarship, economics and/or track record of prediction markets," according to the blog.
Neumann sizes up the performance of the markets from a different perspective: How soon did the investors settle on a consensus pick? "The thing that impressed me was how quickly the markets selected the eventual winners," he said.
In the Democratic as well as the GOP race, the winner rose to the top by mid-February. "That gives you some idea how far in advance a knowledgeable investor knows what's going to happen. And that's sort of startling," Neumann said.
He said past studies have indicated that the pace of financial contributions, as reported by the Federal Election Commission, was a significant market-mover. "I bet if we look ... we'll see that the sharp upturns are related to how much money Obama raised, or how much money Clinton had to give her campaign to keep running," Neumann said.
Beyond the horse race
But this isn't just about the campaign. Neumann said the research that goes into political markets is being applied to other fields as well. For yeasr, the IEM has been working on an Influenza Prediction Market for health-care professionals, and Neumann said the experiment could do for public health what the political markets are doing for campaign prognostication.
The key challenge is to predict the course of a flu epidemic two weeks in advance, which would give agencies enough time to scale up vaccinations and head off bigger problems. "Two weeks in advance, our market certainly outperformed statistical methods," Neumann said.
He said the cost-benefit analysis for flu prediction hasn't yet been worked out in detail, but "just on the raw question, 'Can you learn faster?' ... the answer appears to be yes."
The IEM's Berg said the experiment also showed that "the people with information aren't necessarily the people you think of as the experts."
She said school nurses and pharmacists had the best record for predicting the course of a flu epidemic - which makes sense. They're the ones who actually see the leading indicators: an uptick in sick students, or in the number of prescriptions filled.
Higher-level questions
Multiple market outcomes can be combined to suggest answers to higher-level questions, even in the political realm, Neumann said. For example, would Obama or Clinton do better against John McCain, the presumptive GOP nominee?
"If we did this right, we could figure out who's the stronger candidate. ... We didn't get around to doing that this year, [but] that turned out to be one of the bigger issues in the campaign."
To address that question, researchers would set up three markets: one market for weighing the Obama-Clinton market, a conditional market for Obama vs. McCain, and another conditional market for Clinton vs. McCain. If Clinton didn't face McCain, the investors in that market would simply get their money back.
Market trends could be analyzed to arrive at a prediction about who would do better against McCain. A similar combination could be used to ask, for example, how the passage or non-passage of gun control legislation might affect future murder rates.
It all sounds like an economist's dream. Or a politician's nightmare.
"The conditional probability times the marginal probability of the condition is equal to the joint probability," Neumann said. "`I'd hate to be the person who had to educate people on that, because it would drag them back to those old college courses."
Do you think prediction markets will turn out to be the wave of the future, or is their usefulness essentially limited to handicapping horse races? Feel free to add your comments below.
