TOPICS > Politics

Political Polls, Professors and Election Markets Predict the Presidential Race

October 12, 2012 at 12:00 AM EDT
With such a close presidential race, many would like to look to pollsters and their political crystal balls for predictions. But what stats are the most reliable and accurate for predicting what will happen come Election Day? Paul Solman examines how polls and election markets parse data to come up with a winner.

JEFFREY BROWN: Polls that say this, polls that say that, professional oddsmakers, and folks who just throw a dart at a dartboard. So who’s got the most accurate forecast when it comes to the presidential race?

Our economics correspondent Paul Solman went on a quest for answers. It’s part of his ongoing reporting Making Sense of financial news.

PAUL SOLMAN: Want to know who’s going to win the presidential election?

WOMAN: Do you approve or disapprove of the way Barack Obama is handling his job as president?

PAUL SOLMAN: We would like to know the winner too. But who to ask? The pollsters?

Doug Schwartz runs QuinnipiacUniversity’s Polling Institute.

DOUGLAS SCHWARTZ, Director, Quinnipiac University Polling Institute: Right now, Pew has Romney up by three points. Gallup has Obama up by five points in their seven-day track. But since the debate, they have Obama up by just three points.

PAUL SOLMAN: Or how about asking old friend Ray Fair, an economist we have been visiting every four years since the George H.W. Bush administration? Fair has got an economic model.

RAY FAIR, YaleUniversity: So, right at this moment, its 49.5 percent of the two-party vote for Obama.

PAUL SOLMAN: So, the polls and models call it just about even. But, curiously, the betting public tells a very different story.

Online betting prediction markets, like Ireland’s Intrade, which takes bets from Americans, and England’s Betfair, which doesn’t, have made President Obama the strong favorite for months.

The lone domestic and entirely legal betting haven is the University of Iowa’s Electronic Market.

Ever frugal, we asked Hari Sreenivasan, our man in the Midwest, at least last week, to drop in.

JOYCE BERG, Iowa Electronic Markets: This is our server right here.

HARI SREENIVASAN: So the entire prediction market is in there?

JOYCE BERG: That’s right, in that little box.

PAUL SOLMAN: A visual letdown, perhaps, but this is what electronic betting looks like everywhere, a box.

Through this one, the Iowa Market, 1,600 people place their bets, with a $500 limit per punter. The odds when Hari was there?

Director Joyce Berg:

JOYCE BERG: Right now, the markets are favoring Obama. If you look at the winner-take-all market, that market is showing our traders believe there’s an 80 percent chance that Obama will get more than half the popular vote.

PAUL SOLMAN: That was before the first presidential debate, however. The odds are now much lower.

But they’re still above 60 percent that President Obama will win the popular vote.

So, who’s right, the professors, the pollsters, or the plungers?

Let’s start with economics professor Ray Fair, who has based his equation on a century of election data.

He finds that just four variables can pretty reliably pick the winner: incumbency — sitting presidents have an edge — party — being a Republican has been an advantage — inflation rate — the lower the better — rate of income growth per person — the higher the better.

So, this year?

RAY FAIR: The model says a very close election predicted. If the economy were booming, the equation would be predicting a pretty substantial Obama victory. If the economy were going into a recession, it would be a substantial Romney victory. The economy’s more or less in between.

PAUL SOLMAN: Unlike most of us, Fair has a professional interest in the outcome.

RAY FAIR: If it’s a close election, the equation will have done quite well. It’s been predicting a close election for over a year, so it will be a rather substantial plus for the equation, the fact that — if we have a close election, regardless of who wins.

PAUL SOLMAN: So everybody else, pretty much, has a horse in this race, but your horse is your equation?

RAY FAIR: I’m an econometrician, not a pundit, right? So the horse you’re interested in, OK, is the — is my equation.

But I have another side of me, which I’m not talking to you about, which is what I personally would like to see.


PAUL SOLMAN: And you’re not going to tell me what that is?

RAY FAIR: I’m not going to tell you what that is, no.

PAUL SOLMAN: How accurate has the model been?

RAY FAIR: My interest is in vote shares. And on that metric, looking at the predictions the average error over these elections is 3.6.

PAUL SOLMAN: Three-point-six percent?

RAY FAIR: Correct, within this margin of error, yes.

PAUL SOLMAN: As it turns out, that’s about the same margin of error as for the second major source of presidential predictions, the polls.

DOUGLAS SCHWARTZ: In 2008, we polled on a bunch of swing states. We got within three points of President Obama’s actual margin in Ohio, Pennsylvania, Florida, Colorado, Wisconsin.

MAN: Is you mind made up, or do you think you might change your mind before the election?

PAUL SOLMAN: Earlier this week, Quinnipiac surveyed voters in three of this election’s swing states: Colorado, Wisconsin and Virginia.

MAN: Your telephone number was randomly generated by a computer, and I need to pick someone at random from your household to participate in this survey.

PAUL SOLMAN: To make sure they’re polling a scientific sample of likely voters, these callers use random-digit dialing of landlines and cell phones. But that’s just for starters, says Doug Schwartz.

DOUGLAS SCHWARTZ: When we call a household, we don’t automatically speak to whoever picks up the phone, because they tend to be women. We speak to whoever has the next birthday in the household.

We also call over several days, because we don’t want to just reach the easy-to-reach people. For example, seniors, people with young children are easier to reach. So we will call a phone number at least four times between 6:00 and 9:00 during the week, but then we will also try on weekends.

We also conduct interviewing in Spanish. So, it all gets back to this principle of making sure that everyone has an equal chance of being selected.

PAUL SOLMAN: But, of course, not everyone wants that chance.

Do people get angry at you?

DEETA MOUNING, Quinnipiac University Polling Institute: Oh, yes, they do. They blast us out.


PAUL SOLMAN: They do blast you out?


PAUL SOLMAN: And what do they say when they blast you out?

DEETA MOUNING: Oh, I can’t say.

PAUL SOLMAN: In fact, out of 100 phone calls, callers reach only about 10 eligible respondents.

WOMAN: Hello?

PAUL SOLMAN: And, of those, only about half cooperate.

WOMAN: I would say, out of 100, I would probably get, like, five interviews.

MAN: I would say about three or four.

PAUL SOLMAN: Although those who do take the calls seem to do so gladly.

ARTHUR ROY, Quinnipiac University Polling Institute: Generally, people who are interested in talking about politics love to engage, so they will talk to us quite a bit.

PAUL SOLMAN: But since every pollster claims to come up with a scientific random sample eventually, how can the results be Romney up by four and Obama down by three at the very same time?

A hefty margin of error, says newly minted Ph.D. economist David Rothschild, who works at Microsoft Research in Manhattan.

So, Rothschild incorporates models, polls, and markets into a constantly updated forecast on his blog

Last we looked, his data mash-up gave President Obama a 65 percent chance to win.

DAVID ROTHSCHILD, Microsoft Research: I never sweat too much about any individual poll because there’s too much noise.

PAUL SOLMAN: Too much noise meaning?

DAVID ROTHSCHILD: Too much movement.

PAUL SOLMAN: Take the so-called bumps after a vice presidential pick, or a convention. Or take the first presidential debate. Say you’re a Democrat the morning after.

DAVID ROTHSCHILD: A pollster calls you, you don’t want anything to do with it. You’re just — you’re upset. You hang up that phone.

They call another guy, exact same demographics as you, because they’re trying to fill some sort of demographic hole or whatnot. He’s a Republican. He’s stoked. He wants to yell for how excited he is for Romney.

The same person on paper looks like he switched to Republican, but what happened really is, is that the Republicans want to talk that day vs. a Democrat who is like, I’m not even answering the phone. I’m in a bad mood. Obama screwed up last night. So…

PAUL SOLMAN: Between prediction markets and polls, which do you trust?

DAVID ROTHSCHILD: Prediction markets, because prediction markets have all of this polling information available to them, as well as additional information.

PAUL SOLMAN: The very latest economic statistics, for example, or a secretly taped fundraising talk.

WOMAN: Right about here is Romney, that secret video taping of the 47 percent.

PAUL SOLMAN: We found one such better, MITSloanSchool student Andre Gloria.

ANDRE GLORIA, student, MITSloanSchool: The question that you’re really answering when you’re placing a bet is, who do you think will win, not, who will you vote for?

PAUL SOLMAN: Gloria plays betting markets the world over. He’s already won enough on this election to holiday in style.

ANDRE GLORIA: I will have a nice Christmas break thanks to this, yes.

PAUL SOLMAN: And though this MBA student doesn’t play in low-stakes Iowa, he’s very much like her speculators, says Joyce Berg.

JOYCE BERG: Our market traders are a lot different than the general voting population. They’re predominantly male, wealthier than the average voter, younger than the average voter, more well-educated than the average voter.

PAUL SOLMAN: And as we take one last look at the Iowa, the quietest bookie in America, one last question: How accurate have the polls proved to be?

JOYCE BERG: So, we have gone back over time since the ’88 election, looked at over 900 polls, and compared the IEM prediction to the poll prediction, and found, in 75 percent of the cases, the IEM was more accurate than the polls.

We also look at night-before forecasts, and there, in that case, we are usually within about 1.3 percent of the actual election outcome.

PAUL SOLMAN: For the latest odds from Iowa, Intrade and Betfair, just go to our Making Sense website, where we link to all of them in real time.