Gallup The Shark

by Mike on October 8, 2008

in Election 2008,Politics

Some polls are scientifically valid.  Others are not.  At no time is that more obvious when the major polls not only show different numbers, but are also moving in different directions.  Earlier today, Sal gave us an excellent synopsis of both today’s tracking polls and their trends.  After his post, Gallup published its tracking poll showing Obama with an 11 point lead and growing.

So which polls do we believe?  The ones that show the race tightening, or the one with Obama running away with it?  Although I’d love to believe Zogby (Obama by 2 today) and Hotline (Obama by 1 today), I cannot do so honestly given the former’s history and the latter’s bias.  Throughout this election, I’ve stood by Rasmussen (Obama up 6 today) and Battleground (Obama up 4 today) and will continue to do so.

It should be obvious by now that I’m about to discuss Gallup.  I will never put any faith in a Gallup poll (Obama up 11 today) for several reasons.  The most obvious is that  with less than a month to go, it is still a poll of registered voters rather than the more predictive “likely voter” sample.  Less obvious but equally important is that Gallup has been unable to get its party identification story straight for at least the last eight years.  That is not a good sign for a polling company in a country where party ID rarely sees a dramatic shift in a short period of time.

The end product of Gallups’ shenanigans is a poll with wild fluctuations during a campaign, followed up with trends that seem to produce the results Gallup wants, and ending with a “corrected” election day poll designed to protect its reputation for “accuracy.”  In 2004, Gallup took this to the extreme by predicting three mutually exclusive results.  Maybe Gallup covered its bases in 2004 as a result of their final poll in 2000, which was within the margin of error but incorrectly projected the popular vote winner.  Or maybe Gallup didn’t.  Whatever the reason, forgive me for not taking Gallup at face value.

I put stock in polls that sample likely voters, follow party ID samples close to historical turnout, and don’t show wild fluctuations.  Gallup’s horse race number and trend are out of step with every other poll at this time.  The valid pre-second debate polls show Obama leading and McCain slowly gaining.

If you’re interested, here are some links on polling accuracy for 2000 and  2004.

Leave a Comment

You can use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Previous post:

Next post: