Sunday, October 26, 2008

Hell, everybody else did it....

So, I'm going to jump in to complete the pool out there.
My absolute favorite in polling is done by the Investor's Business Daily (which foretold the winner of the 2004 Presidential race by 3/10ths of ONE PERCENT). As far as polling goes, it doesn't get a whole lot closer than that, Ladies and Gentlemen. To put this in context, every other polling source (alright, to be fair...most of them) had Kerry up smartly going into the last 4 days.

Today, IBD has McCain and Obama neck and neck....both are just skirting the three percent over/under. 46.5 and 43.3 respectively. This was with 886 likely voters.

And today was the worst of the news in the past three days. The last survey to use 1060 likely voters was Oct. 23, where Obama had 44.8, and McCain showed a tight 43.7 trail. (That's ONE FRIGGIN POINT, ladies and gentlemen.)

Now, if you'll remember, I said Kerry was up going into the 2004 contest's last days.

Presidential Hopeful John Kerry Has a 5.3% Lead Over President George W. Bush According to Newly Released Results in WHYY West Chester University Poll.
Publication: PR NewswirePublication Date: 29-OCT-04

And finally, in past years, Ralph Nader has garnered up to 4% of the final vote. If and when his name appears on the 2008 ballot, will it pull the same support ?

But WAIT!!!!!!! It gets better!!!
Actually, there appears to be a simple methodical error in the polls, and we can actually quantify it and adjust for it. This error is the hang-up rate, i.e. people who refuse to participate in the polls. It is far higher this election than previously. This makes sense, after all, what do you do when you get junk phone calls? You tell them to pi$$ off and hang up. Now suppose, instead, it's a pollster, and you decide to participate. But then they tell you that they have 100 questions, and will you take 45 minutes to participate -- NOW, will you tell them to pi$$ off? Don't laugh, some of these polls are as bad as that. Specifically, the CBS/NYT poll asks 100 questions, and they consistently show a 12%-13% lead for Obama. It's pretty clear that they are measuring only those people who are willing to answer 100 questions. If you look at other polls, the trend is plain -- those polls which ask the fewest questions show the best results for McCain. So this is the systematic bias in today's polls -- they don't count people who won't particiate, and those people are overwhelmingly McCain voters. So, why has Obama's lead been increasing lately? Simply because people are losing their patience with the left-wing biased media, and are hanging up at an increased rate as the campaign progresses. And that's all it is, the hang-up rate, and not a swing to Obama at all. Simple, eh? Lastly, we can remove this bias by treating the hang-up rate as a certain percentage McCain vote, and seeing what hypothetical percentage yields consistency between the polls, if you add them in. My rough rule of thumb calculation shows 80% McCain, and does not include the "excess", that is, those voters who just hang up right away, regardless of how many questions the pollsters want to ask. This all makes sense, as with all the junk phone calls we get, people don't have the patience for telephone-led processes like they used to any more. It's like the old famous Willkie phone poll, which showed only how a systematic bias controlled the result, well, today's sytematic bias is the vastly greater hang-up rate, which are McCain voters overwhelmingly. Look for a decisive McCain win on November 4.

Posted at Ace Of Spades by a genius named "proxy" at October 26, 2008 04:58 PM
I fully think "proxy" is definitely onto something here.

And all this is absolutely lacking in a discussion of the "Bradley Effect" which I believe will actually set some networks back on their heels when they see just how far off their exit polls were from their realtime results.

Results for Oct. 27th indicate that Obama's lead has decreased from 3.2 to 2.8 undecided voters appear to be joinging McCain in larger displays.

October 29- Exactly 4 years to the day of Kerry's 5.3 Lead in the polls in 2004, and with 9.2% still undecided, Obama goes from 4.0 points up yeserday to a mere 3.0 points over McCain today. And the margin of error? +/-3 points...

October 30th... Obama is up today by 4.1%, and his people have already admitted their need for a 7% lead to overcome "Bradley's Effect". I think we're closing in Pennsylvania and leading in early voting in Florida.


PWConservative said...

I Predict the Bradley effect in VA will be at least 4 points.

The Bulletproof Monk said...

That's a good start. And it will predictably lead to a little confusion on Election night if the press actually predicts the race on exit polls that conflict with "real" hard numbers by morning.

David Weintraub said...

That's an interesting theory.

Wrt the IBD poll, I wouldn't have too much faith in their accuracy. Some issues were exposed with their methodology - see here:

As Nate Silver points out, almost everybody got the 2004 race right, but "this election is considerably more difficult to poll, and it's exposing the weaker pollsters."

Are you counting on the "Bradley effect" as a narrative to cover for vote flipping?

The Bulletproof Monk said...

Between the Bradley described by a democrat pollster, a 7% spoiler,
And the hang-up rate (because I've failed to answer the phone three times myself)
And the Bradley Effect in the exit polling that occurred in the Kerry/Bush contest, I feel confident that this ain't over yet.
The IBD was 1/3 of ONE PERCENT within the final count.
The rest of the polls were not uniformly anywhere near close. Most of the other polls had Kerry anywhere from 5.3 points up to 7.7 points up just a day before the election. That's why there was a rolling fog on the networks as to how their exit polls could have been so far off.

The ONLY scary and unpredictable part of this is the rampant and far-slung voter fraud.
The chances of a smooth election were less than favorable WITHOUT having to doublecheck (or not bothering to check at all) numerous voters who can be proven not to have lived in a particular state despite voting there, as well as those three counties in Mississippi where they have more voters than they have living residents.