Monday, April 16, 2007

Robots make better pollsters

Slate's Mickey Kaus offers a theory about that outlying Gallup poll that showed no progress for Obama:
The Gallup Poll appears to be a conventional telephone survey conducted by human interviewers. Why might Hillary do worse in a robo-poll, like Rasmussen's, where the pollee doesn't have to talk to an actual person but simply presses buttons? There's an obvious possible answer: Men don't like Hillary but they're reluctant to say so in public. They'll tell a robot. But they chicken out when they'd have to tell a human interviewer--especially, maybe, a female interviewer. They're scared of looking like sexist pigs.
Clinton is less popular with men, but the robo-pollster explanation wouldn't have to be gender-based. Clinton's aura-of-inevitability thing could work on anyone: maybe people talking to human pollsters fear that expressing support for Obama will make them sound naive, unrealistic, even disloyal! Either way, the robo-poll probably is a closer approximation of actual voting behavior...

My own theory (link above) was/is that Gallup's problem is its tedious list of 11 candidates, which artificially inflates the value of name recognition. It appears that the Rasmussen poll is only asking voters their opinions of Clinton/Obama/Edwards. Obama's stronger showing under these circumstances would seem to be a point in my favor.

No comments: