FiveThirtyEight Combines Polls, Reporting and Baseball
If polls are the fix that political junkies crave, this campaign is no time to try to get clean. In a survey research version of the 1980s crack cocaine boom, public relations firms, interest groups and news orgs have been peddling their numbers on every street corner.
Those polls have created a relatively new dilemma for journalists and news consumers: How do you make sense of it all? (Or, to push the drug analogy, who has the good stuff?) How could Obama be ahead by 10 points in one poll and yet in a dead heat with McCain in another? Which polls matter –- state, national, partisan, battlegrounds?
A number of specialized blogs and sites have sprouted up this election to aggregate and discuss polls. One, though, goes further with a two-pronged approach: one theoretical, the other practical. FiveThirtyEight is part blog, part journalism, and fairly addictive. (I signed up for the daily text message updates, which has not improved my home life.)
Nate Silver, FiveThirtyEight’s founder, takes all the polls he can find, plugs them into calculations that adjust for factors such as reliability and state demographics and then issues his own projections of who will win. Meanwhile, blogger Sean Quinn and freelance photographer Brett Marty have been driving around the country in a 1998 Ford Escort, dropping in on McCain and Obama field offices in an effort to assess the big unknown on Tuesday: turnout.
Within six months, Silver, 30, has moved from an anonymous Daily Kos contributor to a guest on cable news shows. Quinn, who last played poker for a living, is contemplating a move to Washington to cover politics. FiveThirtyEight now gets about 800,000 pageviews each weekday (about 500,000 unique visitors), Silver said. The site was included on a list of hot political sites and blogs released by comScore last week. And devoted readers leave several hundred comments on individual posts.
There are other Web sites and blogs that focus on the numbers. RealClearPolitics offers a comprehensive list of polls and averages them. Chuck Langer blogs about polls for ABC News, and polls are a frequent topic on msnbc.com’s “First Read” blog. Pollster.com has several bloggers who clearly know what they’re talking about. And compared to FiveThirtyEight’s Excel-created maps and charts, Pollster.com’s impressive, Flash-based customizable charts (click on a state to view the scatter plot) may be the speedball of polling. (OK, that’s the last drug reference.)
But FiveThirtyEight is the only site with ties to fantasy baseball. Silver invented a system of baseball statistics, called PECOTA, that aims to predict the performance of baseball players, in part by comparing them to a library of thousands of real players’ past season performances. (While FiveThirtyEight is taking a lot of his time now and makes some money, Baseball Prospectus is his primary occupation.)
When you’re used to calculating 30 or 40 statistics for 750 major league ballplayers, Silver said, figuring out what share of the vote will be split between just two candidates isn’t that tough. “In some ways, politics is easier to look at and interpret than baseball data.”
The problem Silver saw during the primary season was that there was too much polling data and not enough analysis and understanding. Each poll is reported as news in itself, he said, although they’re “not terrifically accurate instruments.”
Understanding and Interpreting Polls
As Election Day approaches, learn how to make sense of the latest poll results, decipher the data in election polls and go beyond the horse-race numbers. Enroll now
Imagine the disparate versions of reality that would be reflected if one news organization reported a poll indicating Obama were ahead by 13 or 14 points and another reported him holding a slim lead of just 2 or 3 points. Actually, there’s no need to imagine –- we see those stories all the time. Such differences are “more than you can get through statistical fluctuations alone,” Silver said. “I think that demands explanation.”
That’s where Silver employs his expertise with baseball statistics. In Silver’s view, a completed poll isn’t the end of the process — it’s the beginning. He starts with the polls, weights them according to historical reliability, sample size and when it was conducted, adjusts for the demographics of each state, does something I don’t understand to adjust the results of older polls, and simulates the election 10,000 times to come up with probable outcomes. The folks at Pollster.com also run some calculations on figures before putting them online, though the description on its site doesn’t hint at the level of complexity that Silver describes.
Silver tells his readers whether he thinks Obama will suffer from the “Bradley effect” (not likely, he writes on his blog and in Newsweek), which states are the most likely to tip the election (Virginia and Colorado now top the list) and whether each candidate can lose in Ohio and Florida and still eke out the race (McCain, no; Obama, probably). Some of those predictions may seem elementary now; there has been plenty of talk lately in the media about how Obama is headed to victory on Tuesday. But as others have noted, Silver has been ahead of the boys on the bus, for instance correctly predicting Obama’s margins over Sen. Hillary Clinton in the Indiana and North Carolina primaries.
Weeks of reading this site has spurred me to think about how polls are and will be employed in political reporting. This may be the most polled election ever, notes The New York Times (Silver said 2,500 polls have been released this year) and the proliferation of these studies shows how they could become the modern version of “man on the street” surveys. Every reporter has done those, fully aware of the limits of such random, anecdotal interactions.
One night after the Republican National Convention in St. Paul, Minn., I conducted a “one-man focus group,” as a friend calls it, with a garbage man who lives across the Mississippi in Wisconsin. (Margin of error: plus or minus one garbage man.) Gas prices were killing him, he said, and it sure seemed like the economy was a mess. So does that mean Obama will win Wisconsin?
Raw information is everywhere -– just 500 automated phone calls or a garbage man away. Insight, however, is scarce.
Too often, Silver told me, polls are used to provide a “veneer of science” to media narratives. Whether in baseball or politics, “the basic problem is that people misinterpret random events as being meaningful.”
Silver: Journalists should “apply the same kind of fact-checking and balance to coverage of polls as they would to any other set of facts.” There simply isn’t enough skepticism of polls, Silver said — especially exit polls, which often miss the mark. That’s why he believes journalists should avoid reporting that centers on polls; when they do report the results of a poll, they should strive to place them in context with other surveys. They should, he said, “apply the same kind of fact-checking and balance to coverage of polls as they would to any other set of facts.”
“I think in newsrooms across the country, there needs to be more discussion of, and kind of respect for, the science of this,” Silver said. “There’s no reason to rely on just one when you have literally eight coming out every day.”
Even better, he said, perhaps reporters should ignore them and do what Sean Quinn is doing — going where the candidates aren’t.
Quinn, 37, admits that he’s learning on the job. He isn’t a trained journalist; this spring, he too was blogging anonymously on Daily Kos (which is how he met Silver). But he has worked as a field organizer for campaigns in Montana and Texas. He knows what’s on the call sheets in front of volunteers as they punch away at telephones, and he understands the difference between a high-quality voter contact (face-to-face) and a low-quality one (robocalls). So he started writing about the ground war.
In state after state, in cities and small towns, Quinn has been dropping in on field offices and describing the scene: volunteers working the phones, training sessions on how to canvass a neighborhood, whether the lights are even on. Quinn will have traveled through 14 battleground states by Election Day. Time and again, he said, the Obama offices have been buzzing, while McCain’s have been more low-key.
He said he has tried to report on McCain efforts just as much as Obama’s, “but the offices are closed or nearly empty, and in comparison I see Obama offices full.”
His routine is to contact the campaigns in each state when he arrives, but he doesn’t give the campaigns a heads-up on where he’s headed. “If you give them advance notice, they try to rig stuff up for you,” he said. “Even Obama people do.”
While pollsters and bloggers debate which “likely voter” scenario will best reflect Tuesday’s turnout, this is the kind of reporting that sheds some light on where the race is going. Quinn pointed out a recent David Broder column as an example of the kind of work that he has been doing for weeks.
Quinn’s posts are a mix of reporting and opinionated blogging (though not nearly as opinionated as his posts on Daily Kos). In a post from Raleigh, N.C., about a rally and a visit to an early voting center, he confessed that he was overcome with emotion after observing an elderly black man interacting with a black child in line.
So while he acknowledges being an Obama supporter, he said he is doing everything he can to tell the McCain side of the ground war. He said he hasn’t broken through much, though he did with a McCain organizer in Charlottesville, Va. “I wanted that guy to feel like, when he read it, that I had clearly been interested, respected what he was saying and really wanted to get his voice out there,” Quinn said.
Field organizers generally can’t speak on the record and are too busy to deal with reporters anyway. But Quinn said he’s been able to use his knowledge of campaigns to strike up relationships, and he often ends up interviewing volunteers.
That’s how he learned that on Election Day, some volunteers in northern Virginia will go out between midnight and 3 a.m. to drop Obama literature at houses of those who get up as early as 4 a.m. to commute to work. (The Obama campaign later called him and asked him to remove some of the details from that post, but he said no.)
That’s the kind of detail that you probably won’t get from a campaign spokesperson, and definitely not from a poll.