Track and Field

Some people have asked me about national polls in the Presidential Election as opposed to state polling. I tell them don't give yourself the headache, it's not worth it.

One thing that is becoming more prominent in national elections though are tracking polls. These are polls taken usually over a 3-day period, sometimes longer, with a rolling average that is updated on a consistent basis.

There are currently four major national, reputable tracking polls (IMO):

- Gallup Daily (+5)
- Rasmussen Daily (+6)
- Research 2000 (+10)
- Diageo/Hotline (+5)


Diageo is the only one of these that does their tracking samples at night. All of them are updated on a daily basis. The current Obama lead totals are in the parenthesis, with R2K being a bit out of an outlier, perhaps. But what's important to note is that all of these polls are using different ideas of who they believe "likely voters" are. R2K has more independents, Gallup and Rasmussen typically put more people in the Democratic and Republican categories, with less independents.

And, it took me a little bit of time to stop worrying about the actual numbers. Because the numbers matter a little, but what you're looking for is trends in the numbers. If all the polls I mentioned above have a 1 point loss for Obama, and some kind of a gain for McCain all in the same day, they'll still all be different numbers. You won't know which one is most accurate, but you WILL know that there's some kind of a drop in support for Obama, and some kind of gain for McCain. That much is clear, and it makes it easier to pinpoint what kind of an event within the campaign is causing voters to lean one way or the other. Only in the final week of a campaign would I say start to really pay attention to the numbers. If McCain is still 5 points behind a week from November 4th, then his camp is in trouble.

The problem with tracking polls is that it's been proven before, and again this election cycle, is that companies change their demographic samples at will. Gallup's done it as recently as about a month ago, so you never know exactly how consistent their idea of "likely voters" is.

An even bigger problem with this election cycle is that there's bound to be increased turnout among minorities and the youth vote. But by how much? That's where pollsters can guess, but in reality, no one will know until election day. New voters favor Obama, but how many of them will vote?

Another problem that can come up in national polls is the concentration of support for one candidate in a particular region. For example, Obama in the Northeast, and McCain in the Southeast, that may tend to overinflate one's national numbers.

This can be verified by state polling. Obama's national numbers shot up this week, and evetually, the state polling followed suit. What we're seeing now are polls where Obama is competitive in North Carolina, Florida, and Virginia -- all Bush-voting states in 2000 and 2004. If McCain starts to crawl back in national polling and manages to tie it, expect numbers in state polling to head back McCain's way for places like Florida and North Carolina.