Tuesday, December 18, 2007

Misreading the Gallup Poll

Greg Mitchell at Pressing Matters has a discussion of the new Gallup poll. He focuses especially on the fact that Ron Paul has dropped from 5% to 3%, in a tie with Alan Keyes, since the last poll. He points out (correctly, I think) that the media is not missing a grassroots surge in Paul's support over the last few weeks. He also says that the poll shows Huckabee peaking nationally and Giuliani still in good shape.

However, I think he reads a bit too much into this poll. While Paul's support is certainly nowhere near what you would call a grassroots surge, there are a few key factors Mitchell misses here:

1. The problem of small numbers: one of the big problems with relying on any polls involving Ron Paul is that his support level is on the whole relatively small. Thus, a poll showing a rise or fall of one, two, or three points is not statistically significant, but it can appear significant since it shows a big change in his relative performance (ie, a 2% difference in one poll can amount to 40% of his support from the previous poll). If this poll were one of several showing a downward trend in Paul's support, Mitchell's conclusion that Paul was at Keyes' level would make sense. We should probably also add that Keyes' 3% has this same statistical problem. Paul's change in support amounts to 8 fewer people out of 399 polled. The sample size here is just not large enough to draw much of a conclusion.

2. This poll seems to be an outlier in general. Giuliani's support in this poll increased by 2 points, while Huckabee remained steady at 16 points over the last Gallup poll 2 weeks ago. In that same time period, though, the overall polling trend for Giuliani has been sharply down, while the overall trend for Huckabee has been sharply up.

Don't get me wrong- the overall trend for Paul seems to have been going slightly down the last few weeks (since about the time of his Thanksgiving appearance on Alex Jones). But it's tough to say that this particular Gallup poll, standing alone, proves anything about anyone. Especially the Giuliani-Huckabee trends. Since this poll is clearly an outlier, I would not make any conclusions about the state of the horse race based on this poll without some supporting evidence from other polls.

UPDATE: Ron Chusid largely comes to the same conclusions as Mitchell. While I agree with Ron on a lot of things when it comes to the Paul campaign (though he is slightly more critical than I), I do think he's overreacting to this one poll a little bit for the reasons I stated above. Still, as I said, the overall polling trend for Paul has either leveled off or is going slightly downhill since the Alex Jones appearance.

(via memeorandum)