Polling for Dummies

Polling for Dummies: How oversampling, skewing effects the ALL current polling companies #Obama #Romney

Nothing against either candidate, but I wish all their polling companies would get their math right, and if they cannot I thought I'd put together a little simple translator spreadsheet for the rest of us becuase I think all of the polling companies and unskewed polling companies are making one of two mistakes.

First is that over sampling of Dem+6 does not mean that they just added 6% to Obama's lead, it actually means they just added 2%. and Second most polling companies are expecting approximately  Dem+8 to turn out at the polls this fall, a little higher than the 2008 elections, when everything you research tends to make it appear to be more like Dem+3  a true skew of Obama 1.67%

First oversampling of Dem+6 does not mean add 6%

First is that over sampling of Dem+6 does not mean that they just added 6% to Obama's lead, it actually means they just added 2%.and here is my chart to back this up. Assuming Rep Dem and Independents are currently at an exact tie. Here is how a Dem+9 thru a Rep+3 skew would play out. As you can see each oversampling point counts for 0.33%

	Dem	Rep	Ind		Skew Perventage
Dem+9	109	100	100	Obama	3.0%
Dem+8	108	100	100	Obama	2.7%
Dem+7	107	100	100	Obama	2.3%
Dem+6	106	100	100	Obama	2.0%
Dem+5	105	100	100	Obama	1.7%
Dem+4	104	100	100	Obama	1.3%
Dem+3	103	100	100	Obama	1.0%
Dem+2	102	100	100	Obama	0.7%
Dem+1	101	100	100	Obama	0.3%
No Skew	100	100	100	TIE	0.0%
Rep+1	99	100	100	Romney	0.3%
Rep+2	98	100	100	Romney	0.7%
Rep+3	97	100	100	Romney	1.0%

Now if you have stumbled across unskewedpolling.com, they try to do a good job by filtering this skew, but I think they have the math wrong too. he Gallup Dem+5 poll released today shows Mitt Romney behind President Obama by a 50 percent to 45 margin. Unskewedpolling reports with a Rep+3 Romney would now have a 2% lead, but using the chart above Gallup Rep+3 Poll would still have Obama +2%, In general this is easy to correct, don't add 1% for each change add 0.33%.. What is probable for this election is for the turnout to be the same as 2010, which was Dem+3, not the Dem+8 many of the polsters are using, expecially when 2008 was Dem+7 and 2004 was even...

Second most polling companies are expecting approx Dem+8 

Secondly in 2008, there were three polling companies that came really close to predicting the election. They were Pew, Gallup and Rasmussen. In 2008 Pew thought the turnout would be Dem+15, Gallup Dem+7 and Rasmussen Dem+7, and the exit polls showed it was actually Dem+7. In In 2012 Pew is thinking the turnout would be Dem+8, Gallup Dem+5 and Rasmussen Rep+3, A few weeks ago many of the polling companies were using Dem+8 which honestly if Pew was correct is a GREAT number to be following and using in their Math, but is Pew the most correct?

In 2008 Pew thought it would be Dem+15, now they think it is Dem+8, seven points lower. This would seem to infer a number more like Dem+1. In 2004 and 2008 Gallup was dead on perfection, but this year they have a lawsuit from the DoJ, and the threat from the Obama administration to "come show us your numbers" inferring a potential second lawsuit. They have to be freaking out, so I am guessing their current prediction of Dem+5, might be high but we will use it as is... Lastly Rasmussen is showing Rep+3 for it's oversamplings. Take an average of everything and we have Dem+3%

So who is winning using corrected Math Dem+3 Poll

So coming right down the middle assuming Republicans and Democrats are a complete wash with 95% of each party voting for their candidate, and using a Dem+3 which I derived above. I will pick my current prediction assuming this link still shows independents are still Rep+3.5 at this link http://polltracker.talkingpointsmemo.com/contests/president-12-independents

	Dem	Rep	Ind		Skew Perventage
Dem+9	109	100	96.5	Obama	1.8%
Dem+8	108	100	96.5	Obama	1.5%
Dem+7	107	100	96.5	Obama	1.2%
Dem+6	106	100	96.5	Obama	0.8%
Dem+5	105	100	96.5	Obama	0.5%
Dem+4	104	100	96.5	Obama	0.2%
Dem+3	103	100	96.5	Romney	0.2%
Dem+2	102	100	96.5	Romney	0.5%
Dem+1	101	100	96.5	Romney	0.8%
No Skew	100	100	96.5	Romney	1.2%
Rep+1	99	100	96.5	Romney	1.5%
Rep+2	98	100	96.5	Romney	1.8%
Rep+3	97	100	96.5	Romney	2.2%

Now you can figure all of this out in your head if this data seems confusing. Just add 1% for every three points you see listed. If you see Romney has +9% of the Independents vote and you think its really even (9/3=3) subtract 3% from Romney for whatever poll you are looking at. If you see a Poll that has Dem+12 oversampling and you think its really Dem+3 just (9/3=3) subtract 3% from Obama for whatever poll you are looking at. Make sense? Good luck and hope this helps, and maybe one day soon all polling companies will get it right...