Kausfiles Special

Why I Hate Exit Polls

Suppose two candidates are running for office. Candidate Jones is charismatic, youthful, dynamic, and relatively liberal. His campaign is built around the economy. Candidate Smith is mature, unattractive, reserved, and conservative. His campaign is built around foreign policy.

You turn on your TV Election Night to learn that Jones has won. The broadcast then cuts to a correspondent who reports at length, in very scientific tones, on what the “exit polls” show. He says that “Jones won because he convinced voters he was best equipped to deal with economic issues. Some 66 percent of the voters said the economy was their major concern, and Jones captured 75 percent of those voters.”

But this is obvious crap. There is no reason to think that voters voted for Jones because they cared about the economy. It’s just as likely–maybe more likely–that they said they cared about the economy because they liked Jones, who said the economy was important. They might have thought Jones was handsome, smart, honest, or whatever. The economy might have had little to do with it–they would have voted for him if he’d run on a platform of creamed spinach. Exit polls would then have shown that creamed spinach was a major issue for voters.

This fallacy pervades the networks’ Election Night coverage–the tedious hours given over to analyzing exit poll data that really tell you very little about why candidate Jones beat Smith. A similar fallacy seems to be at work in John Ellis’s recent New York Press column arguing, on the basis of exit polls, that “Bradley’s heart problem has doomed his campaign.”

Ellis’ reasoning? Exit polls showed that 76 percent of primary voters in New Hampshire said they were not concerned that “Bradley’s health would interfere with his ability to serve effectively as president.” Bradley carried these voters by a good margin–56 to 42 percent.

But “roughly 20 percent” of New Hampshire Democrats told pollsters they thought Bradley’s health would interfere with his ability to serve effectively. And Gore carried these voters 86 to 13. Put the two groups together, and Gore won by a nose. But, concludes Ellis, “[a]bsent the Bradley health issue, Gore was a goner.”

I don’t get it. It seems to me that the polls Ellis cites don’t come close to proving this. Isn’t it likely that if you liked Gore, or disliked Bradley, you were looking for reasons to buttress your vote, and the Bradley health scare filled the bill? If you thought Bradley was a sanctimonious dissembler, say, then you’d probably think he was sanctimoniously dissembling when he said the heart flutter was no problem. But if there had been no heart flutter, you would have voted against Bradley anyway.

Sure, it was possible for a voter to honestly say he hated Bradley’s guts but didn’t think Bradley’s health was a concern, and some did, but that’s not how the human mind tends to work, is it? If you’re pro-Gore, you don’t analyze the “health issue” in a discrete, detached, analytic fashion; you fit it into a pro-Gore framework. So of course Gore supporters will be disproportionately represented in the 20 percent who said they had Bradley heart worries. That doesn’t mean that if there had been no heart worries they wouldn’t have still voted for Gore.

Ellis, who runs the Decision Desk for the Fox News Channel on election nights, is a genuine politics maven who, as they say, has forgotten more than I’ll ever know on the subject. I called him, and he stuck by his guns. His major argument: “The population of people who didn’t care about Bradley’s heart problem didn’t look all that different from the people who did care about Bradley’s heart problem.” They were just as educated, with similar incomes and religions, etc. So, the argument presumably goes, they differed only in whether they cared about the heart problem.

I still don’t get it, though. Why isn’t it just as possible that the two groups (non-worriers/worriers) differed mainly on whether they liked or didn’t like Bradley, regardless of his heart problem–and that this difference of opinion, it turns out, didn’t vary much among different demographic groups? That seems just as plausible as saying that there was a generalized worry about heart problems, having nothing to do with whether you liked or disliked Bradley, that didn’t vary much among different demographic groups (and that this detached, neutral worry caused this representative cross-section of voters to turn on Bradley when otherwise they would have broken down 56-42 for him).

Perhaps Ellis, or some readers, can set me straight.

Anyway, as regular kausfiles readers know, people couldn’t have voted against Bradley because of his heart problem. That’s because they voted against him because of his position on welfare reform! I’m sure the exit polls that prove this are lying around here somewhere …