Blue Crab Boulevard

A Cautionary Note About Exit Polls

The networks that pool together to get exit polling data are actually locking people into a room this year to prevent early leaking of exit poll data. These extraordinary measures are an effort to stifle the leak of the data to those evil bloggers. Oh, wait. I am one of those.

Two-by-two, polling specialists from ABC, CBS, NBC, CNN, Fox News and the Associated Press will go into rooms in New York and Washington shortly before noon Tuesday. Their cellphones and BlackBerrys will be confiscated; proctors will monitor the doors; and for the next five hours, these experts will pore over exit-poll data from across the country.

If all goes well, only when they emerge from their cloisters will the legions of ravenous political bloggers have any chance of getting their hands on the earliest indication of which party will end up controlling Congress.

"The demand for info is intense, and if the safeguards aren't steel doors bolting people inside a room, it will get out," says Marc Ambinder, associate editor of National Journal's Hotline OnCall. "The insatiable appetite for this info will overwhelm the ability to keep it secret."

So, unless their measures fail, there will be very little exit poll data until quite late in the day. Now, frankly, the exit polls are pretty well useless these days as the more rational pollsters admit:

"People need to realize those numbers aren't the real results," says David Bohrman, vice president of news and production at CNN's Washington bureau, who urges people to be cautious when interpreting poll results. "They show why people voted today, and what was on their minds. The only real way to figure out who won is to count the votes."

Mark Blumenthal, a Democratic pollster known as the "Mystery Pollster" on Pollster.com, which tries to explain the mechanics behind polls, agrees. Exit polls are "still a sample," he says. "Don't get all excited about it." In 2004, Mr. Blumenthal posted a warning in the morning about the vagaries of exit polling data and says he'll post another Tuesday. "You learn the hard way that a one- or two-point lead on a leaked exit poll is meaningless," he says.

Mr. Ambinder of Hotline OnCall, the free, wonky blog updated by National Journal's political reporters, says editors there have vowed not to post any exiting-polling data. They say that the less they post while polls are open the better. "If a header in my inbox says 'exit poll,' I'm going to try to resist the temptation to open it," he says.

Throughout the day, a lot of unreliable numbers on exit polling will be flying around in emails, predicts Jacob Weisberg, editor of online magazine Slate, which published early exit-polling data in 2004. If Slate receives exit-polling data from a reliable source in the media it will publish them again, Mr. Weisberg says, although he doesn't know if they'll be very meaningful. "When elections are really close, exit polls aren't that reliable."

"I'll post it with the caveat that it was all crap last time and will probably be crap again," says Glenn Reynolds of Instapundit.com, a conservative-leaning blog, who will be doing interviews for CNN via a Webcam from his Knoxville, Tenn., home.

That is the important thing. The left has made much of the exit polls, but they are unreliable and pretty well good for not very much. Why the networks even bother anymore is beyond me.

Politics | Gaius Tuesday, 7 November , 2006

By Robert, Tuesday, 7 November , 2006 @ 1:41 pm

You’re right about exit polls.

Remember the vote in the Ukraine?
By Lex, Tuesday, 7 November , 2006 @ 2:11 pm

Now, frankly, the exit polls are pretty well useless these days as the more rational pollsters admit:

Actually, that’s not what your rational pollster is saying. He’s saying that random sampling is not the same thing as counting all the ballots, which is undeniable from the standpoints of statistics AND politics. But he’s not saying exit polls are useless.

A well-constructed, random sampling of voters in an exit poll does in fact provide useful, reliable data that can, among other things, be used as a kind of tripwire to detect balloting problems, as long as the geographic area you’re sampling isn’t too small and the number of people you poll is sufficiently large. In fact, exit polling has alerted officials to the possibility of fraud and other problems in a number of elections here and abroad.

In a county of, say, 350,000 people, like mine, or anything bigger, if you have a representative sampling, you will have, within a small margin of error, a good indication of what the final totals will be. And the rules for determining a representative sampling are well-known to statisticians.

In short, although well-done exit polls are not ironclad determinants of final vote talleys, suggesting that well-done exit polls are “useless” is only slightly less ignorant than insisting that we must study “intelligent design” rather than evolution.
By Gaius, Tuesday, 7 November , 2006 @ 2:19 pm

Well, aside from dropping by to make a gratuitous insult, you really miss the point. Because the people who allow themselves to be polled on exit are self-selecting, not random. Therefore the usefullness depends entirely on getting a true representative sample. I’d say there is quite enough evidence that this is not happening.
By Diva, Tuesday, 7 November , 2006 @ 2:26 pm

I don’t think the problem is with the exit poll data per se - the problem is how that data is used by biased news sources as it was in 2000 and 2004. In Florida, in 2000 the exit polls showed FL going strongly for Al Gore - before the polls had closed in the pan-handle ( a region favorable to G.W. Bush).

These early and erroneous exit poll calls/reports had the effect of turning voters away from the polls. Why stand in line to vote for your guy when the election has already been called?

Anyhow, I for one am glad that the exit poll data reporting is being held back (because the MSM does tend to be biased)- this will benefit voters of all political parties.
By gil, Tuesday, 7 November , 2006 @ 10:46 pm

< ?xml version="1.0" encoding="utf-8"?> < !DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
Hey, gil, maybe if you could express yourself without obscenity your comment would have been read by a few people. Now it wont. Read the rules on commenting on the "about" page and follow them.
By Lex, Friday, 10 November , 2006 @ 5:33 am

Iâ€™d say there is quite enough evidence that this is not happening.

Well, by all means, let’s see some.