Barack Obama may have been the only one to give a victory speech, but the Democratic candidate was by no means the only winner in the 2008 presidential election. Bookmakers also had plenty to celebrate. Political futures markets like Intrade.com attracted massive amounts of money—and media attention—during the 2008 campaign. Justin Wolfers, an associate professor of business and public policy at Wharton, estimates that upwards of $100 million was wagered on the election. Though that doesn’t come close to what Americans gamble on sports, it would eclipse the total amount bet in Las Vegas sports books on last year’s Super Bowl.
With that kind of money at stake, markets gain a predictive power that rivals or even surpasses that of public opinion polls. Wolfers analyzed prediction-market trends—and one bizarre anomaly—in a string of columns for The Wall Street Journal Online beginning in primary season. After the general election he spoke with Gazette associate editor Trey Popp about the history of election gambling and its potential impact on future campaigns.
This year is the 20th anniversary of the creation of the Iowa Electronic Market, which first allowed people to buy and sell political futures. But prediction markets seemed to get a lot more attention during this election cycle than ever before. Why do you think that is?
The first prediction market in the modern era in the U.S. was the Iowa Electronic Market in 1988. But actually betting on elections dates back to at least the 16th century. In fact, there was a very active political-prediction market in the United States on the curb of Wall Street back at the turn of the century—actually, from about 1880 to 1940. And moreover, if you go back and read the front page of The New York Timeson the day before the 1916 election, or the 1908 election, you’ll see there’s no discussion of political polls—because they hadn’t been invented. The front page of the Times was talking about what was going on in the political betting markets. So in one sense the media coverage of the 2008 election looks a lot more like the 1908 election, with increasing attention to political-prediction markets.
So if these markets were common a hundred years ago, why did they disappear from view?
They did disappear from view for a while—in the United States. British bookmakers have been betting on American elections for decades. But there was this lost period between 1940 and 1988 in which there was no betting on the U.S. election in the United States. There was some crackdown on the markets on Wall Street—the legal status of those markets was somewhat unclear and the anti-gambling forces became somewhat more powerful. And people’s willingness to gamble changed a lot.
How come?
In 1968, you wouldn’t want to bet on an election against a random guy, because that random guy could turn out to be someone who works at the Gallup organization. And why would that guy be betting against you? Because he knows something that you don’t—some poll numbers he hasn’t published yet. So people were reluctant to bet with each other because everyone was fearful that they were betting against someone that knew more than they did. You go back to 1908, there were no pollsters, and as a result, information was much more democratically spread.
The question would be why in 2008 are people willing to gamble again? Well, now we’re in a world where basically there are infinite polls out there—dozens and dozens and dozens of these things. So I would actually be happy in 2008 to bet against a guy from the Gallup organization, because he might know the latest Gallup Poll, but I know the latest Zogby poll, and the latest Rasmussen poll, and the latest 20 other polls.
How much money flows through these political-prediction markets, and was the volume heavier this year than in the recent past?
My estimate is that globally there was probably about $100 million bet on the election. That’s probably about three or four times more than the 2004 race. It’s comparable to what was being bet at the turn of the century.
So looking back at the election, which seems to have been more accurate: particular polls, or poll analysts, or sites like Intrade?
There are some very clear examples of where the polls gave false signals. For instance, during the Republican Convention, the polls went very, very strongly towards McCain—despite the fact that we know that conventions usually give you a very short honeymoon, often called a post-convention bounce, which then dissipates very quickly. I think the markets were right to discount that information. So if you’re just doing a naïve comparison of polls versus markets, I think the markets outperform the polls.
What’s a little harder is that this time round, we saw the emergence of very sophisticated poll aggregations. For instance, the website FiveThirtyEight.com. What Nate Silver was trying to do there was take all the different pieces of polling information out there and put them all together into one central number. And if you think about what a market does, it takes a whole lot of different pieces of information that each of us have, and that all gets summarized into one number. Now, there’s some important differences. For one, FiveThirtyEight is garbage-in, garbage-out: it can only take account of what’s going on in the polls, it can’t take account of other factors. The night before an election, the polls probably do contain a lot of information. But three months before an election, it was still quite unclear, for instance, whether the public would be willing to elect an African-American president, or whether the war was going to be a big issue, or what was going to happen with the economy and so on. So it may be that prediction markets are about the same as FiveThirtyEight the night before an election, but maybe not when we talk about longer-run forecasts. That’s still an unknown. I think the interesting question coming out of this election cycle is not whether markets are better than polls—we think that’s true—but whether markets are better than sophisticated poll aggregates.
Markets can give strange signals, too. You drew attention in The Wall Street Journal to what seemed like fishy activity at Intrade during one of the presidential debates. What exactly was going on?
We were seeing the prices of McCain’s stock go up very sharply, for no apparent reason, at very odd hours. And there’s a concern that people might try to manipulate prediction markets, because if McCain’s stock goes up, journalists might write about it—suggesting to people that McCain was the winner—and that might bring forward greater public confidence, campaign contributions, and maybe a self-reinforcing cycle.
We’re always looking out for whether price movements reflect people who are trying to make a profit based on their information—that’s the good thing—or whether they’re trying to manipulate the market. And it’s hard to tell the two apart. Both involve buying stock in your candidate. But to the extent that they do look different, we’d expect someone who’s trying to make a profit to try and minimize the impact that their buying has on the price—to buy your stock at the lowest price you can. In this case trading was occurring that I think maximized its impact on the price. Now you can never say for sure you know it’s manipulation, but you can say it’s certainly unusual. InTrade subsequently did an investigation in which they found that it was one large client doing all the trading. Whether they were doing that trading because they were trying to hedge or manipulate the market, that’s something that I think we’ll never know.
Would this kind of thing be illegal?
This is regulated in Ireland under Irish gambling law, rather than under U.S. stock laws. So I don’t think it would be illegal.
Would it be cost-effective for, say, a long-shot candidate to bet on himself as a candidate for 2012 as a way to generate buzz?
The answer from what we saw this time is that it’s very unlikely to be cost-effective. Everything I told you so far made it sound like this guy was trying to manipulate McCain’s stock. But one thing that doesn’t make it sound that way is that he was spending thousands and thousands and thousand of dollars. And if you look for articles that say, Wow, McCain’s doing well, look at the prediction markets, you can’t find them. No one seemed to have fallen for it. So it looks like it was a particularly ineffective thing to do.
There was speculation that McCain’s selection of Sarah Palin impacted his bid, and some suggestion that the vetting process had not been as thorough as it might have been. Can you imagine a scenario in which a candidate might look at the futures market when making a choice like that, trusting that the wisdom of the betting swarm can vet a potential running mate better than a team of political advisors?
Absolutely. And there have been markets like this. For instance, if you have a market that pays off if a candidate wins both the nomination and the presidential race, you can infer the likelihood that they will win the presidential race if they’re the nominee. So you can use prediction markets to ask if-then kinds of questions.
There was an interesting market run like this in 2004, and that prediction market actually said that if the Democrats gave the nomination to Howard Dean, they would almost certainly lose the general election. If you think about how the rest of Dean’s campaign proceeded, I think that forecast looks like it was pretty good.
—T.P.
