Monday, May 14, 2012

The Associated Press Is Up To It's 
Old Tricks In It's Polls


The oldest trick in the book when slanting your polls to create the results you want is to oversample one party over another and then report the results as 'news'.  


Party ID between Democrats & Republicans is currently pretty close.  In 2008 Obama won by 7 points.  
Note that out of 7 national polls, only 2 give Obama a lead outside the margin of error of 3 - 3.9% (MOE).  Romney leads, ties or is within the MOE in five of them. 


So what was the sampling/weighting in the AP poll that presently claims Obama has a +8 advantage? 


http://surveys.ap.org/data/GfK/AP-GfK%20Poll%20May%202012%20Topline%20final_2012eln.pdf


How about a whopping 12+ point advantage to Democrats?  


Remove the leaners and the results look like this: 

So with leaners a +12 pt advantage for Dems, and without leaners a +9 pt advantage.  

To get an idea of how much Party ID fluctuates between elections, here's Gallup's results on that the past few years: 

2008:

Dems hold a +8 advantage, their biggest since 1983. 

2009: 

Poll taken in Sept. 2009 & Dem advantage has shrunk to +6.

2010:


From that high in 2008 to a 22 year low in 2010. Lead shrinks from +8 to just +2, thanks to Obamacare, the Stimulus Bill, etc. Thanks Barack, Pelosi & Reid! Awesome job there! With Indies swinging to the Republicans, Dems get handed a historic shellacking in the 2010 mid-terms. 

2011: 
According to Gallup, Dems +4 in late 2011

So - quick recap.  Democrat's Party ID advantage:

2008: +8
2009: +6
2010: +2
2011: +4

So the AP expects people to believe that the Democrat Party advantage has surged to over double what it was last year to +9 or more?  

Laughable.  Note as you read the end of the PDF the AP mentions they also weighted their sample but does not say HOW or in which direction.  How much did they swing the weighting of the sample to get a +9 Dem advantage & how much did that affect the +8 lead the poll gives Obama over Romney?

To get an accurate poll they need a sample that mirrors the current electorate, which would give Dem's a +4 Party ID advantage at best.  Instead they more than double that to an absurd +9.  

To get Obama above the 3.9 MOE, they have to resort to this?  Can't wait to see how big the skewing of the sample gets the next few months.  

No comments:

Post a Comment