by Datechguy | October 8th, 2012
There is a ton of crowing on the right and worry on the left based on the new PEW poll showing President Obama’s +8 lead turning into a Mitt Romney 4 point lead in just 30 days.
Granted the debate was devastating but even no, especially with polls one finds encouraging, it is VITAL to see if there is some hanky panky going on with the sample.
Last time I hit the PEW poll because the sample was D+9 when I saw the D vs R sample number on this poll I saw something I have never NEVER seen before in a national poll:
A +4 R sample in a national poll? Say What!
That is a 13 point sample swing, 13 points. if you compare this to the Rasmussen numbers this poll slight oversamples the GOP by 1.4 pts. I have never EVER seen a national poll do his and neither have any of you.
So what happens when you skew the sample 13 points in the other direction? Lo and behold the results…move in the other direction.
“But DaTechGuy” you say, How do we know both samples aren’t right? Can’t this isn’t just more people saying they are Republican because they NOW support Romney, you know following the crowd and all that?”
Glad you asked my boy, glad you asked.
Remember last month when I looked at this poll I looked at the numbers concerning who the sampled voted for in 2008, saw the ridiculous gap in the figures and said this:
That’s a 15 point Obama edge in an election that Obama won by 7.2 but it’s even worse because those numbers are based on 86. if you do the full percentages based on 100% of voters it comes out Obama 54.6% (+1.7 over 2008) and McCain 37.2% (-8.6 under 2008) in this sample.
Now you might think Obama is ahead but does anyone ANYONE believe he will do 1.7 point better than 2008. Does anyone ANYONE believe Romney will underperform McCain by 8.6?
Maybe that will fool Talking points Memo, Firedoglake and the like but it won’t fool anyone who can, you know do MATH.
And lets do the math again to turn those who voted into the proper percentages
(pct of Sample that voted Obama)=42 (obama 08 voters)/85(number who voted)
(pct of Sample that voted McCain)=37 (obama 0 8 voters)/85(number who voted)
Ok in this sample the Obama sample matches his 2008 figure EXACTLY, while the McCain sample is under by a mere 2.3 points. Unless all of these people were lying about who they were voting for last time and suddenly got religion we have a sample that is within 2.3% of the actual vote in 2008
So what does this mean?
It means that when you have a poll that closely reflects the actual GOP vs Dem split (within 1.4 points of it if you believe Rasmussen) you get a result that accurately reflects what is going on rather than a Democrat fantasy.
Much as I hate to give solace to Andrew Sullivan, public opinion didn’t swing 12 points in a week, Pew simply took a sample that almost matches the American electorate and we are seeing what was already there, only more so.
Postscript: I follow memeorandum on twitter and as I was finishing this post I saw this jaw dropping tweet:
— memeorandum (@memeorandum) October 9, 2012
I followed that tweet to this story at the Washington Post:
If either Pew or Gallup kept their “sample balance” constant, there would be far less of a shift in their headline numbers. Poll watchers need to closely assess both trend-lines as the election draws near.
That’s exactly right but wasn’t that the same argument I’ve been making right along? The argument that got all of us called Poll “Truther” and poll “Deniers” and “Conspiracy Theorists”?
All I have to say is Pew, welcome to Realville!
Exit Question: Is this sample Pew’s attempt to correct itself pre-election or will see a new poll just before the day re-skewed to try and create the Obama comeback?
Update: Twitchy notices
After screening out respondents who are unlikely to vote, Pew was left with a sample of likely voters that was R+3, according to Chuck Todd.
If the sample in the Pew poll is skewed toward Republicans, it probably is not off by much. A month ago, Rasmussen reported a 2-point edge for Republicans in party ID. In Wisconsin’s recall election a few months ago, the partisan breakdown was R+1, according to exit polls. Party ID was tied 35-35 in 2010.
But we’re glad to see liberals acknowledge that polls can be skewed and that it is possible for a sample to over-represent members of one party or the other. Maybe Obama supporters won’t call us insane the next time we question a D+13 sample.
Chuck Todd? Chuck Todd is now a poll truther? I didn’t see him at the meeting, does he know the secret handshake?
Update 2: Althouse watches Sully’s meltdown and figures it out:
Sorry, but it’s hard not to see this as a lot of posing. A set up for the big announcement that Obama is back. If Obama is any good at all in the next debate or the one after that, we’ll be told the man is a miracle.
Update 3: That new Pennsylvania Poll that has Romney down 2 pts has some interesting splits:
If you include leaners then it’s one point closer.
As for registration
I don’t have a lot of faith in Pennsylvania, there will be plenty of walking around money in the cities and the suspension of the ID law for this election means the left has one last chance to do what they do best. 2016 will be different but I don’t have high hopes here.
Update 4: Michigan internals are now out here is the sample:
So the sample is D+5 has Barack Obama up 3 48-45. Last month when the numbers showed Obama up 10 47-37 the sample was….D+5
be afraid Democrats, be very afraid.
Update Bottom: Oh and a reminder, the Washington Post is a multi-million dollar corporation and the people they pay just figured out poll samples matter.
I am a guy who since the Obama years makes his entire living from Ads and Tip Jar hitters and have been saying this for months. If you think that my analysis of polls that the Washington Post finally has figured out, is worth something I would be most grateful if you were to hit the DaTipJar to help keep this stuff coming
And of course Tip Jar hitter get the codes for my Subscriber only videos the newest once comes out tomorrow.