September 27, 2004
— Ace Emphasis mine.
The CNN/USAT/Gallup poll shows Bush ahead by 8 with likelies and, oddly enough, ahead by 11 with registered voters. (The ABCNews/WaPo poll also finds Bush doing better with RVs, so maybe this is a real thing going on here.)
CNN's headline: Bush is apparently ahead.
He's on the very edge of the margin of error. The MoE is +/- 4 for either candidate; but that doesn't mean that a 8 point lead is meaningless-- far from it. It is pretty unlikely these men are tied at the moment; furthermore, Kerry can't be ahead, unless this is one of those 1-in-20 polls that is simply unreflective of the greater population.
With the ABCNews/WaPo poll confirming the general picture -- Bush ahead, and significantly so -- the "1-in-20 polls are garbage" theory would seem to be weak.
Does CNN typically say that one candidate is "apparently" ahead by a whopping 8 points (more with registered voters), or do they only do that when the wrong candidate is "apparently" ahead?
Ace of Spades HQ Headline:
Panicked CNN Liberals Apparently Still Believe They Can and Should Cocoon Their Audience
Hey, look, I only said apparently.
Ace of Spades HQ Future Headlines (provided to me by my friend, the time-travelling bologna sandwich called Johnny Coldcuts):
Bush Apparently Wins Re-Election
Apparently Takes 325 Electoral Votes; Apparently Nets 55% of Vote in Reagan-Like Avalanche
Will Take Second Oath of Office in January, Apparently
Nevermind, that one might be real.
Posted by: Joe R. the Unabrewer at September 27, 2004 06:03 PM (5anNT)
Posted by: Joe R. the Unabrewer at September 27, 2004 06:15 PM (5anNT)
Posted by: Charlie at September 27, 2004 06:34 PM (m3S0Y)
Posted by: Tom Alday at September 27, 2004 06:36 PM (c3m2t)
One small correction. The polls use a 95% confidence interval. The odds that Bush is 4 points lower is 2.5% because it's ONE TAIL of the normal distribution. Similarly, the odds that Kerry is 4% higher is 2.5%.
Therefore, the odds that Bush and Kerry are tied is .000625.
That's a 1 out of 1600 chance of Kerry being tied. The odds that he's ahead would have to be somewhere South of that. So the cocoon is much deeper than you imagined.
Posted by: Birkel at September 27, 2004 07:03 PM (1mwIk)
Stick a fork in Kerry, he's french toast. And Bush will be applying the powdered sugar on Thursday in the debates, maybe fruit and whipped cream in October...
Posted by: Brett at September 27, 2004 07:09 PM (7OiLt)
It's much more likely that the poll is accurate to, say, +/- 1 than +/- 4. Half of all polls like this will be accurate within 1; 10% or so will be accurate within 4. (These aren't the actual numbers; they're ferinstance numbers.)
Posted by: ace at September 27, 2004 07:12 PM (RGQgo)
The margin of error is a direct measure of the accuracy of _the_poll_. 95% is the normal setting for the margin of error - and that leads to "There's a 95% chance that Bush _currently_ would win between 48 and 56% of the popular vote". That means (duh) that there's a 5% chance that he's outside that range. half-too-high, and half-too-low.
You're right that if you're given '52% with MOE 4%' you should always pick exactly 52% as the expected outcome - it is the most accurate guess available. The range 51 to 53 is also substantially more likely than the range 49 to 51... but the chance of it being 'lower than 48' is precisely 2.5%. (Assuming 95% confidence levels, which is normal, but not clearly stated here.)
Holy crap. No changes -> toast even including a 2% cheat factor.
Posted by: Al at September 27, 2004 07:30 PM (bXUgf)
They almost always round up. The margin of sampling error is probably about 3.5% -- they always round up for no particular reason, which means it is outside the margin of error. Even if they didn't, then at a 90% level, the the margin of error will be lower, and it will be outside the interval....
Of course, you can throw all that out the window if they fucked up and didn't take a random sample. It wasn't a fluke that they initially called Florida wrong in 2000. It was incompetence by the VNS. Getting a random sample is harder than it seems it should be, but my guess would be that if anything, it might be a little biased in favor of Kerry, but that's pure conjecture based on 2002 polls versus actual outcomes. As in 2002, I think conservatives turn out.
Posted by: Mark at September 27, 2004 07:42 PM (uvpsw)
Posted by: Mark at September 27, 2004 07:51 PM (uvpsw)
Oh my God! That's like the fucking prophicies of Nostradamusor something. Thanks for clearing that up for me, douche Bag!
Posted by: Dacotti at September 27, 2004 08:37 PM (QvMeW)
Posted by: otalps at September 27, 2004 11:16 PM (Y1bWh)
Posted by: mr.pie at September 27, 2004 11:34 PM (tDTe4)
It's got some links and stuff. No big deal but if you're interested check it out.
Oh, and sorry it doesn't have any of that patented Ace of Spades humor. My bad--it's a dry subject.
Posted by: Birkel at September 28, 2004 12:43 AM (JBOkg)
"Does CNN typically say that one candidate is "apparently" ahead by a whopping 8 points (more with registered voters), or do they only do that when the wrong candidate is "apparently" ahead?"
is found here:
I'll give you three guesses...
Posted by: Superstar at September 28, 2004 02:21 AM (T5ThL)
Posted by: Mark at September 28, 2004 03:27 AM (uvpsw)
Posted by: Ralph at September 28, 2004 03:59 AM (AaBEz)
But the question would then become what is the relationship of the two. Certainly it is not 1:1 as Nader and Undecided receive votes. There is no evidence that moving away from Kerry wouldn't just result in a non-voter (or an undecided voter). Ditto for Bush's side.
Also, the previous CNN/Gallup/USAT poll was in line with this one. Certainly that is evidence that the reported MOE for each of the polls overstates the MOE if the polls can be combined. If they can be combined it's even less likely the CNN headline makes any shred of sense.
Posted by: Birkel at September 28, 2004 04:32 AM (JM/Zq)
or "Poll: Bush apparently gets modest bounce"
Posted by: James at September 28, 2004 04:35 AM (rP4OC)
Posted by: Sobek at September 28, 2004 06:31 AM (XwlU1)
Posted by: Nickie Goomba at September 28, 2004 06:53 AM (9B2q6)
Posted by: kelly at September 28, 2004 07:33 AM (c2+Oq)
Posted by: MD at September 28, 2004 07:41 AM (hoo48)
Up next, Bill Schneider will tell us how we can do a better job of shilling for John Kerry.
Posted by: Bladebender at September 28, 2004 07:51 AM (YvgQi)
Posted by: dano at September 28, 2004 09:21 AM (zqLrp)
62 queries taking 0.9983 seconds, 261 records returned.
Powered by Minx 1.1.6c-pink.