So I did a little tinkering with the IBD stats, and there's something reall off with the weights

[Godhumor]

So, we already know the IBD poll is odd with the youth vote going to McCain by 50 points, but I noticed something else. Take a look at these numbers:

Conservatives: Voting for O 21% Voting for M 69% Don't Know 10%
Moderates: Voting for O 57% Voting for M 29% Don't Know 14%
Liberals: Voting for O 87% Voting for M 7% Don't Know 6%

Obama is winning two out of the 3 groups by a wide margin (even conservatives are breaking more that you would think for Obama), so how does the final result end up at 44.8% voting for Obama, 43.7% voting for McCain and 11.6% Don't know? It looks pretty fishy.

So I went ahead and set up a spreadsheet to figure out how these results could possibly happen. Based on the info given (i.e. they rounded to the nearest percent) it looks like there's only one possible sample split of the 1060 voters:

Approximately 443 conservatives (Again approximately due to the rounding of the percentages)
Approximately 518 moderates
Approximately 99 liberals

That seem off to anybody else?

[Blaze Diem]

1. interesting..I'll be waiting for greater minds than mine for a discussion here.

But I will back to read what they have to say about this.
tks.

[TornadoTN]

2. 443 Conservatives vs. 99 Liberals - well now that's not a loaded poll

I wouldn't put much stock into this poll, especially if it refuses to weight the poll properly.

[mohc]

3. In 2004 the split was

21% Lib
45% Mod
34% Con

Nevermind that the split will probably be more favorably Lib vs Con this time, the IBDD split looks to be:
9% Lib
49% Mod
42% Con

They are so far off base it isn't even funny.

[Godhumor]

5. Yeah, I just looked at the percentages for this sample

My only guess is that there must have been a handful of people who didn't answer the question--otherwise that's such an inaccurate sampling it is almost criminal that the company went ahead and released this poll.

[lessthanjake]

4. Theres always more conservatives than liberals though

However, its generally more like 18% liberal, 30-35% conservative, so this is still off.

[Godhumor]

6. There may always be more conservatives but do you think 42% of the country would be conservative?

Because that's what this poll is showing. (edit: not going after you if it seems like it; I agree with you that the split is much closer than what this poll shows)

[Abacus]

7. I did a similar thing with party ID:

Solve simultaneously:

84d + 44i + 5r = 44.8
8d + 39i + 86r = 43.7
8d + 17i + 9r = 11.6

result:

Dem = 32.4%
Rep = 31.3%
Ind = 36.4%


Did I miscalculate or are those numbers really as laughable as they appear to be?

Edit: Now with COLORS!

[Godhumor]

8. The math certain seems to show you're right

What an odd poll.

[Roland99]

10. Geez...and we all thought Gallup skewed Republican? Wow! What a FARCE!

[nickn777]

9. Nate Silver on 538 has already debunked this poll...

www.fivethirtyeight.com/

[Godhumor]

11. Well aware, as I mentioned in my OP. I wanted to see some of the other weights

To put it simply--everything is off.

[pnorman]

12. Good interview of Nate Silver on "Democracy Now!":

http://www.democracynow.org/2008/10/23/the_spreadsheet_psychic_with_fivethreeeightcom_nate

pnorman