Knowing How Polls Can Be Used Well And Misused

[Leonidas]

Trusting any one polling organization is a mistake. Not trusting polls at all to help you understand voting trends and patterns is also a mistake. This election cycle has seen a couple of wild swings of opinion 'measured' by a single polling organization that should warrant skepticism about any one 'snapshot' poll. The Newsweek polls from a few weeks ago and the most recent Gallup polls are glaring examples.

But a couple of things are clear about how to use poll results reasonably.

National polls are most useful if averaged with each other.

Local...state level...polling on the presidential election is far more informative that national polling results since, as we all know, you win by winning states and not by winning the popular vote nationally.

AND

News organizations tend to report polling results differently one to the next because they have contracts with different polling organizations.

They also tend to report polls that fit their political POV if they have a clear POV. But if their POV is simply "Let's make this race look very close no matter what so that we keep people watching and reading," then that may lead to the most misleading reporting of poll results of all.

Here's an interesting blog article discussing how CNN may be intentionally misreporting poll results in order to exploit the 'horse race' aspects of the campaign. Basically it reveals how the much viewed CNN national election map is plain unreliable.

The blog posted below is followed by a national map that is much more accurate in relation to the best polling results presently available.

Why report news when creating it is so much more fulfilling? TheZoo

Electoral-vote.com: President, Senate, House Updated Daily

[dvgtr]

I wish every poll included the exact sample size, method of communication (telephone, in-person, etc.) and the exact questions asked of the respondents.

Without that info it's just, as Mark Twain said, "Lies, Damn Lies, and Statistics".

Numbers don't lie but a**holes with agendas will lie using numbers.

[FishMonger]

You can make up stat to prove anything. 73% of all people know that.

[Leonidas]

Latest PEW TRUST national poll on the presidential race. Pew is a particularly reliable and non-partisan polling org. Still, it's a national, not state by state poll and so it can give a misleading view unless carefully scrutinized.

Seems to me some of the major hints from this poll suggest that in the last month:

McCain ads designed to create doubts about Obama based on "experience" and race have worked to solidify and broaden the Republican base support for McCain by making Obama easier for conservatives to hate. McCain's support has grown particularly among southern whites and evangelicals and that doesn't particularly help McCain since they live in large numbers in very red states he was already far ahead in. However, overall the poll seems to be a bit of good news for McCain. Be interesting to see swing state polls after the two conventions and after (if) Obama's camp starts running some tough commercials against McCain.

Overview: Presidential Race Draws Even