What's your opinion on the accuracy of the polls?

[Muenkel 0]

Basically if you average them together, we have a dead heat. But I wonder how accurate they are. With all the cellphones and caller ID, I wonder if they are getting an accurate sample even with the margin of error.

_________________________________________
Chris

[kallend 2,174]

Biggest problem is that what people say they will do and what they actually do are not necessarily the same thing. (Insert gratuitous comment about Bush and fixing Social Security here).

In 2000 after the election I recall they interviewed some people who had been sampled who said they deliberately gave false information to the pollsters...

...

The only sure way to survive a canopy collision is not to have one.

[Muenkel 0]

I agree with that Kallend. I think some folks definitely lie to make the other candidate look better in the polls and motivate their candidate's supporters. But aren't those folks considered in the margin of error?

_________________________________________
Chris

[kallend 2,174]

Quote

I agree with that Kallend. I think some folks definitely lie to make the other candidate look better in the polls and motivate their candidate's supporters. But aren't those folks considered in the margin of error?

Well, the motives might be that, and might just be that they find polls and pollsters annoying and they lie as a passive form of retaliation.

How is the margin of error determined anyway? What is the margin of error in the margin of error?

...

The only sure way to survive a canopy collision is not to have one.

[Jimbo 0]

Quote

Well, the motives might be that, and might just be that they find polls and pollsters annoying and they lie as a passive form of retaliation.

I think that this is much more likely than folks lying to motivate thier candidates supporters.

-
Jim

"Like" - The modern day comma
Good bye, my friends. You are missed.

[quade 4]

Quote

How is the margin of error determined anyway? What is the margin of error in the margin of error?

Is that a rhetorical question? Is this?

quade -
The World's Most Boring Skydiver

[CanuckInUSA 0]

The only polls I believe are the official result polls after election day (hopefully 2004 will not be a repeat of 2000 and things won't drag on for weeks after election night). GWB may very well be re-elected prez and if he is, so be it, the American people have spoken. But I'm spectulating that there are tons of non-vocal, non-traditional, non-political types of people who will be voting this year for the first time and I'm speculating that many of them will NOT be voting for GWB (because of what he did in Iraq). So it'll be interesting to see who actually wins.

Try not to worry about the things you have no control over

[quade 4]

Polls can be as accurate as whoever is putting up the cash to have them done. They can be as inaccurate as a biased question or sample.

If I phone 1000 people in Santa Ana, California and ask them a question about immigration, I'm going to get wildly different answers than if I asked the same question of the same amount of people in Beverly Hill.

If I phone 1000 people in Fayetteville and ask them questions about abortion, The 10 Commandments, AWB and Kerry I'll get different answers than had I called people in San Francisco.

So, you have to look at the methods of the pollsters and make sure they are independant. Polls sponsored by politicians are notoriously biased.

You also have to understand that not all polls are really polls but sophisticated campaigning. A Push Poll sounds like a poll, but isn't.

On the whole . . . I'd say that if you looked at PollingReport.com and saw all the different polls together you'd get a much better read on things than simply looking at one poll. Sometimes it's pretty obvious which polls have been skewed for some reason or another.

quade -
The World's Most Boring Skydiver

[crwtom 0]

Quote

Polls can be as accurate as whoever is putting up the cash to have them done. They can be as inaccurate as a biased question or sample.

most generate random phone numers all over the US. Of
course there's the problem with cell phones - but by-and-large
it's pretty random.

The main problems seem to be

- 80-90% of people being called up hang up the phone. You
can easily speculate that these are undecided who want their
f*** peace of mind when they make up their mind. The pool of
undeicdeds might be much larger than we think it is.

- everyone has a different definition of what a "likely voter" is.
Patterns of behavior used to determine that may have a bias.

- even with the registered voters (a much larger pool) it
depends how you phrase things .. from "vote for definitely"
to "tend towards".

Here an interesting article from the same pollingreport site.

*******************************************************************
Fear causes hesitation, and hesitation will cause your worst fears to come true

[billvon 3,131]

>But I wonder how accurate they are.

Like Quade said, they can be as accurate or inaccurate as you choose. I remember doing an analysis in college of what sort of sample size you need to have 99% confidence that you are not off by more than 1% on a popular vote, and it was absurdly small - something like 100 people. But (and these are big buts) you have to pick them perfectly randomly and then make sure that the people you pick answer and answer honestly.

Since that's rare, people often use methods to compensate; that's mainly increasing the sample size. You can do other things (like artificially decimating the samples to even out the population that actually answered the poll) but they are not well understood by most people and thus open to charges of bias.

Of course, the big problem with a presidential election that the result will not be based on popular vote but on electoral vote. In such cases it's possible for 75% of the country to vote for candidate A and have candidate B win, which renders nationwide polls less useful. There are some good sites out there that go state by state.

[kelpdiver 2]

Quote

Of course, the big problem with a presidential election that the result will not be based on popular vote but on electoral vote. In such cases it's possible for 75% of the country to vote for candidate A and have candidate B win, which renders nationwide polls less useful. There are some good sites out there that go state by state.

The distribution you'd need to make that happen is so absurd that we can write it off as impossible. From a practical standpoint, you might get a winner with a vote deficit up to 5 points.

The big elements for polling variation is that uncertainty in sampling the youngest voting cohort, and the determination of likely voters. The results of this election may alter how some of them deal with these two issues.