Wednesday, November 28, 2007

A poll by any other name

AZAce over at Arizona 8th had a post up yesterday about a new Zogby poll showing Clinton trailing all major Republican candidates. Of course, AZ8th isn't the only site to post this, as numerous other bloggers and news outlets have jumped all over the data.

At first glance this looks to be significant news, deserving of prominent display. After all, it would represent a roughly 10-point slide by the presumed Democratic front-runner in a very short span of time.

However, what AZAce didn't mention (and I suspect simply missed, because it isn't noted until you get to the small text all the way at the bottom - I missed in on my first fast read-through) is that this latest poll is not a traditional phone poll, as all the prior surveys Zogby cites were, but instead an online interactive poll. This changes things considerably.

First, there are clearly some self-selection issues with online polling (old folks, poor, etc., are less likely to vote in it). This can be adjusted for with proper weighting, but it does take more care.

Second, Zogby's interactive polls do not, shall we say, have the best reputation for accuracy. In general, they have been shown to be far less accurate in predicting actual outcome margins than traditional phone polls.

Third, a more traditional poll done by Gallup about 10 days earlier showed Clinton with a small, but steady, lead on all Republican candidates.

Given the result of the Gallup poll is entirely consistent with across-the-board trends over the last several months, while the Zogby interactive poll represents a dramatic break from those trends, it might be best to wait for a few more sets of polling data to come in and confirm the Zogby result before concluding the entire dynamic of the race has suddenly shifted.

If I were to place a bet, it would be that the Gallup poll will be found to be consistent, and the Zogby results an outlier.

1 comment:

x4mr said...

I consider online polls ridiculous and entirely unreliable. The self-selection (as opposed to established selection methodologies) makes the results inherently absurd.

While not as ridiculous as say, Rush Limbaugh surveying his listeners, it is flawed to where its claims are unsound.