The polls are saying everything from "dead heat" to "Obama has a 5-point lead." How much stock do you put in pre-election polls?

I've been reading a lot about the polling this year recently, not just the data itself but some rumored behind the scenes stuff.

Generally, I've always followed the RCP averages that take into account both conservative and liberal poll data as well as independent sources. I think that usually gives a good picture.

However, I read something the other day that interested me. Two things actually. Once source indicated that a lot of polling data is being skewed. Typiaclly you have a roughly 30/30/30 poll. Thats evenly breaking down your surveys between liberals/conservatives/indepe... However, it sounds like a number of sources are skewing heavy on liberal samples and only using 6 percent to 10 percent of independent samples. Thats going to skew any poll.

The other thing I read is about actually modeling. Pretty much every source, even Rasmussen, is using 2008 modeling. Meaning that the demographic turnout that occured in 2008 is a baseline for the poll modeling. 2008 was a historic year. There are a lot of questions if minorities and college students are going to turnout anywhere near where they did in 2008. For the polling data to be accurate that has to happen and a lot of pundits don't see that happening. That historic turnout universally favored Obama, which means that modeling based on 2008 will skew in that direction.

When you see a 3 point lead for Obama, it likely means the contest is a tie. If its a point in Obamas direction or a tie it likely means a slight Romney lead.

But when you take into account the first piece of information, about whos being sampled and how thats getting skewed, and you look at how accurate Rasmussen has been as far as presidential elections are concerned its Rasmussen that you need to look at.