It’s no secret that the exit polls the networks were setting such stock in were deeply, deeply flawed. Megan says it better than I can.
A number of people have emailed to point out that the samples in individual precincts are small. That’s true, but the overall sample is large, and it went awry on every level: in each state and in the national vote. Sure, it was a close race, but as far as I can tell, the errors all ran one way: towards Kerry. Rumour has it that the reason the networks were so slow to call the Carolinas is that the exit polls showed them going for Kerry, a nonsense result in light of the result, and even in light of previous polling.
I’ve been tilting against biased media for a few years now, but I’d like to point out that this is hardly anything new. We had the exact same problem last election, and even the mid-term elections of 2002. Just in case you think this is a case of American news media falling down on their face, I would like to point out that they have the same problem with exit polls conducted in foreign countries.
So the news organizations should just save some time and money and not bother. But I doubt they’ll listen to me.
This item from the Washington Post mentions that the blogs were experiencing heavy traffic mainly because they were able to post leaked exit poll data in real time.
This troubles me, because it appears that Big Media has figured out how to turn the blogs to their own ends. Simply use them as a faster news outlet. That way they can spin the story they want and still have plausible deniability. (“It wasn’t us. We’re a respected news organization. It was those pajama guys who blabbed crappy data without bothering to verify.”)
This is self correcting, for the most part, since blogs allow instant feedback from readers. But if the source of the data is protected in some way, and we’re operating from a leak, then our credibility can be knocked into the toilet. Just like the blogs managed to do to Big Media’s rep.
So everyone should make sure that they mention when data is not 100% credible. Or else it might (gasp!) bring your readership down.