Thursday, January 28, 2010

No Exit on Exit Poll Misinformation

On the day of last week's Massachusetts special election for the U.S. Senate seat, I was disappointed to read that there would be no exit polls. The mainstream media that normally organize exit polling had decided against them for this election back when the race looked to be a cakewalk for Massachusetts Attorney General Martha Coakley. When State Senator Scott Brown surged ahead of Coakley in polls during the last one-and-a-half weeks before election day, there was simply not enough time left for the media to organize a reliable exit poll.

A week after the election, I was surprised and intrigued to read an op-ed in the Boston Globe by Democratic media consultant Dan Payne describing the results of an "exit poll" conducted by Hart Research Associates, a Washington, D.C. based survey research firm.

Did somebody actually take an exit poll despite election-day stories to the contrary?

No, as is clear from two organizations which did the polling on election day.

David Drembus, director of operations at Hart Research Associates, confirmed in an email that the firm took a telephone poll on election night of 810 Massachusetts voters. Nothing wrong with that, just that it's not an exit poll; Payne stepped in it by mischaracterizing the poll. (Payne could not be reached for comment.)

Exit polls are a gold standard for election analysis, because they track the reasons behind actual voters' decisions at the ballot box. They are also difficult to organize, since they require interviewing actual voters [image at right] as they leave representative polling places scattered across the entire election area. As Mike Allen wrote at (and was also reported by the Wall Street Journal):
No exit polls from today’s Senate special in Massachusetts, where the polls are open from 7 a.m. to 8 p.m. The consortium scrambled to put something together — for the “why,” more than for the call — but wasn’t confident a reliable system could be built so fast.
Facts like these didn't stop WBZ's curmudgeon political analyst Jon Keller from proclaiming on his blog and Twitter about the "eye-popping exit poll numbers"*** in the election. The problem with Keller's political analysis is that no such exit polling data exists: in Keller's case, he linked to a Rasmussen summary of a telephone poll taken on election day (see bottom of this page). A curmudgeonly comment left on his blog, pointing out that there were no exit polls taken, seems to have gone missing in the message moderation queue... (Keller did not immediately respond to a request for comment.)

Last Friday on the show Beat the Press, Boston Phoenix journalist Adam Reilly and WGBH host Emily Rooney bemoaned the fact that we wouldn't really know why voters voted the way they did because there weren't any exit polls taken:
REILLY: I wish we had exit poll data so we could...

ROONEY [interrupting]: I've been lamenting that.
Sadly, the studio's cameras didn't show the look on the face of suddenly-silent Jon Keller -- seated in-between Reilly and Rooney -- who was presumably either ready to pounce with his "eye-popping exit poll numbers" or dropping his jaw that his colleagues could be so clueless as to think there weren't any exit polls. Or maybe, just maybe, he kept his mouth shut so that he would not step in it again.

That Keller so readily accepted the outcome of his Rasmussen "exit poll," in any case, is a reflection of his sagging political analysis of Massachusetts politics. He should easily have recognized that Rasmussen's polling results were not representative of Massachusetts voters:
Our polling shows that 53% of voters in Massachusetts are Democrats, 21% Republican and 26% not affiliated with either party.
These numbers agree with neither the voter registration in Massachusetts, which have 51% of the electorate unenrolled, nor the 2008 presidential election exit polling data, which have 43% of actual voters identifying as unenrolled. Those poll internals are red flags that were picked up by Nate Silver of, who is currently the best independent poll analyst on the national scene; Keller should have immediately seen the poll problems, too.

Let's hope that the media don't sell short the next election and put together a set of reliable exit polls. Otherwise we are likely to see another round of misrepresentations of the nature and reliability of the polling data come November.

*** Update 1: Keller has, as of 1/29/09, now corrected the text on his blog entry to read "these election-day poll numbers."

Update 2: It looks like WGBH's Beat the Press will be tackling the issue of no exit polls in this week's broadcast -- without Keller.

No comments: