What the Polls Show

So, what do the polls say about the 2008 election? Are they even worth the paper they're written on? Michael Barone thinks that they're somewhat relevant. I'm with DJ Drummond on this. You really can't trust the polls to be accurate about what people are thinking with this or any election. It doesn't matter if it's a poll showing a dead heat or Obama ahead or McCain ahead (several weeks back).

I suspect that the polls say more about the preferences of those polling rather than the likely or registered voters contacted. After all, the pollsters use subtle wording differences to shade the results, and any number of statistical tricks to gussy up the results.

The internals of many polls expose just how the process is done.

Here are the internals of the AP/GfK Roper poll. Just a quick glance at the breakdown of those who self identify as Republican or Democrat shows some interesting details. Here are the numbers (Q10/11):

Democrat – strong 16 19 21
Democrat – moderate 19 15 19
Republican – strong 10 14 13
Republican – moderate 12 14 16
Independent – lean Democratic 13 11 10
[VOL] Independent – don’t
lean
5 4 4
Independent – lean Republican 11 12 11
None – lean Democratic 4 3 1
[VOL] None – don’t lean 6 5 2
None – lean Republican 5 4 2
Don’t know 1 1 *

The first number is total respondents, then likely voters for the current period, and the third number is the breakdown for the prior period (9/27-9/30).

The total number of self identified Republicans was 22, compared with 35 Democrats (13 spread), while the likely voters is much closer at 34-28 (6 spread). It suggests that those GOPers contacted were much more likely to vote than those Democrats contacted. It also shows that the pollster continues to push the idea that Democrats will come out in bigger numbers than Republicans.

Even accepting that premise, this particular poll would suggest that not only are things tight, but that McCain could conceivably be ahead in the polls.

However, as I've stated before, I don't trust these polls nearly as much as others are willing to believe. Are things close? Absolutely. Will it be a blowout? Hardly.

The best indication that it will not be a blowout - look at where the candidates are campaigning and what they're saying. When Pennsylvania is in play, you know that this is far tighter than anyone looking at the polls would be led to believe. Look at what they're saying and their target audiences. They're literally all over the map. The polls are all over the map, which leads to the conclusion that the pollsters are seriously screwed up with their methodologies and assumptions, or both. In Gallup's case, they've completely changed their methodologies, and when the numbers showed so far out of line with every other polling outlet, they were forced to adjust and provide additional statistical analysis and polling (traditional model).

There's only one poll that counts for anything, and it's fast approaching. It's one where you have to go to the polling place or submit a ballot, and your choice is known only to you and the counters who tabulate the results.

It also means don't bother paying attention to the exit polls on election day either, since they too have been shown to be faulty.