Larry Thornberry reports that the Giuliani camp is responding to Rasmussen’s poll showing a Huckabee lead in Florida by questioning Rasmussen’s methodology. This is deeply unconvincing. The perception that robo-pollsters are less accurate isn’t backed by sound data; automated pollsters have performed very well in the past. I’m not even sure there’s an intuitively sensible theory as to why automated polls should be more prone to error. (Thornberry’s comment that he’d hang up on a recording is inapposite. People hang up on live pollsters, too, and it shouldn’t matter unless there’s a specific demographic or ideological trait that corresponds more to hanging up on one than on the other.)
There’s a very good chance that the poll in question is an outlier. To be precise: The margin of error is +/- 4 points with a 95% confidence level, which means that we can expect that, in one out of twenty polls taken with this same method, the margin of error will not hold — that is, the error will be greater than +/- 4 points. But that’s true whether a poll is automated or not. After all, the Giuliani camp surely isn’t prepared to dismiss the (also automated) SurveyUSA poll, which showed them way ahead in Florida a couple weeks ago.