Shock Poll: Kerry/Edwards down 15 pts in NC

[bfp]

[Here]

[Ryan]

OMFG!!!!!!!!

Actually, bfp, NC was never close. The article, which you have completely misrepresented, states that adding Edwards to the ticket kept the numbers the same among likely voters, but actually narrowed the margin among registered voters in that particular state. I am not sure how this is a "shock" to anyone - the poll and the article describing it certainly state nothing of that nature. Oh, and this is an important tidbit that you conveniently excluded from your post:

Nationwide, however, Edwards gives Kerry a six-point bounce. The Democrats now lead Bush-Cheney by 50% to 45% among likely voters, with independent candidate Ralph Nader at 2%. Three weeks earlier, before Kerry announced his vice presidential choice, Bush led Kerry by a single point.

You should go work for Michael Moore. You just stole a page right out of his playbook.

Now, for the rest of you, if you'd like to see the entire article and poll, then here you go.

http://www.usatoday.com/news/politic...election_x.htm

[bfp]

hehehe - I love watching you get worked up. I took the subject line directly from Drudge along with the link. And the bounce that Kerry/Edwards received after the nomination was much lower than expected. Any pollster will tell you that.

[Ryan]

Originally posted by bfp
hehehe - I love watching you get worked up. I took the subject line directly from Drudge along with the link. And the bounce that Kerry/Edwards received after the nomination was much lower than expected. Any pollster will tell you that.

I'm not worked up at all. I'm just telling the truth.

[bfp]

What matters is that this poll could have an effect throughout the country about what the state thinks about Edwards, if this poll gets any time within the mainstream. I doubt it will.

[MysteryMeat]

*cracks knuckles*

Aiight, you're about to get owned by a Statistician.

As usual with reporters and those that don't understand statistics, 'cherry picking' and 'fact searching' (rather than fact finding) is rampant. USA Today is generally pretty good at printing the fine print, however most people don't care (or don't understand it) and they go straight for the pretty tables and graphs.

This poll, however, is shit.

First off, the poll is conducted in only a single state (North Carolina), which leans toward Bush. There is no mention in the fine print about skew or normalization (which would be there if they put any fine print at all), so let's get something straight off the bat: what you're really looking at is how the residents in North Carolina might vote and how they consider the candidates, particularly Kerry. [my note: who the fuck cares other than N. Carolinians?]

Secondly, the fine print itself is manipulated somewhat. Note how is says that the Margin of Sampling Error for the 848 Registered voters is +/-4%. However, two paragraphs down, it fills out the rest of this statistic (which a vast majority of the people won't read), so really the Margin of Sampling Error (MSE) is +/-4% up to +/-6%. That leaves a HUGE swing area of 8%-12%; you could drive a truck through that.

Third, the second paragraph in the fine print tells you that of the 848 registered voters polled, only 680 are likely to vote. As you can read in that paragraph, the move from a sample size of 848 to 680 automatically bumps the MSE from +/-4% to +/-5%. But wait -- there's more. Of those 680 likely voters, historical voting records for that state show that only 50% of those likely will actually vote. So guess what? The sample base just dropped from 848 (+/-4%MSE) to 680 (+/-5%MSE) to [second-level statistics excluded by the reporter] 340, which has a minimum calculated MSE of +/-6.27% at 95% confidence. Note how the minimum +/-6.27% is higher than the reported maximum 6%.

Fourth, considering subjective statistics, note that the last paragraph (as most good Gallup polls will give you) says "In addition to sampling error, question wording and practical difficulties in conducting surveys can introduce error or bias into the findings of public opinion polls." Without a list of questions, we have no way of knowing whether the questions were biased in their wording in the first place.



If you want to know anymore about it, let me know, but I think this pretty well tears that journalistic turd apart.

[bfp]

That was a good summary M Meat. Great analysis. A+

So, in contrast, you're saying that the majority of all political polls are flawed, right? I've done some research and have found that more Rep's or Dem's were sampled in past polls to achieve a certain outcome.

[MysteryMeat]

Originally posted by bfp
That was a good summary M Meat. Great analysis. A+

So, in contrast, you're saying that the majority of all political polls are flawed, right? I've done some research and have found that more Rep's or Dem's were sampled in past polls to achieve a certain outcome.

You really have to look for the right polls. That's why its always important to read the fine print, and never, ever trust a poll that doesn't list a margin of error, a sample size, and a description of the sampling method (in this case, registerred voters in North Carolina).

The problem with reading journalistic reports containing statistics is that they are written to take something extremely boring to most (statistics), and make them interesting. By doing so, they have to trim a lot of fat, and in the process, a lot of the meat gets trimmed as well.

Take all journalistic statistical reports with a grain of salt, do a little more research on the ones that matter to you.