About the polls...

[Southern Patriot]

Does anybody understand what "Margin of Error" means?

People seem to get their panties in a wad about polls that really don't mean anything because the results fall within the MoE.

Can anybody tell me why we should pay the slightest attention to, for instance, Kerry supposedly leading Bush by ONE point when the MoE is 4.5%? Even if it's a key swing state poll, I still say that it's essentially meaningless.

[On the Road]

1. A Lead Within the MOE is Still a Lead

there's just more than a 5% chance the lead is due to a sampling error.

If a poll has an MOE of 3% and one candidate leads by 3%, there's a 95% chance that's a real lead and not due to too few people being polled.

There are bigger problems with polls, particularly (1) whether people's responses reflect what their actual vote will be, and especially (2) whether the poll sample matches the profile of people who will actually go to vote on Nov 2.

[papau]

3. Polls that ask if you'll vote for "Bush-Cheney" get fewer yes than if just

Will you vote for Bush, which in turn gets fewer yes than "will you vote for President Bush"

Then there are the pre-questions that set up a mood that affect the response on the money question.

And, with the "GOP" controlled polls especially, the tone of the person asking the question can bias the result.

[On the Road]

4. Absolutely -- The Response Bias is Very Real

I did a major project in experimental psychology in college in which response bias was a very major problem. There are all kind of assumptions people make about these polls. They do show the general picture and major trends, but no matter how precise the computations, they are far from accurate.

[cspiguy]

2. it's the trend that matters

it aint over yet.

[The_Counsel]

5. It's Not the Polls, But the Way We READ Them...

You're right: if it's within the margin of error, it could swing in any direction of that given margin (usually it's about 3%, but could be as much as 5.5%).

The purpose of these polls is supposed to show a trend. If you look at it that way, they can be quite useful. If you think it's an exact predictor, as most folks seem to, the poll will have less value than used toilet paper. Take Colorado, for an example: the latest polls seem to show Bush ahead, but I believe in my heart -- because of long- and short-term trends -- that Kerry will take it. He won't even need the "Electoral Vote Division" referendum to pass -- which likely won't anyway...

Another example: Ohio. A couple of days ago, no less than THREE polls showed Kerry slightly ahead, but within the MOE. It was still heartening to Kerry supporters because he had been behind there for most of the general election campaign. Today, another poll is out showing Kerry ahead 50-44. That's OUTSIDE the MOE, which suggests Kerry's strength there is VERY real.

That actually leads me to my next point. Kerry's lead in states like Ohio seems to have been built gradually, which is something a person like me can trust more. But look at the national polls that recently had Bush jumping out to 8-13 point leads for NO APPARENT REASON AT ALL! I just can't trust unresonable bounces like this. Directly after the GOP Convention, yes; but NOT directly after the President took CLEAR ass-whoopings in the debates that millions of Americans watched with our own eyes. We can't be THAT blind AND stupid, can we? So now you can get an idea why so many DU-ers were pissed off with pollsters like Gallup: it seems they have an agenda in their polling, and it isn't "to inform the people..."