The latest Public Policy Polling poll for the Virginia Governor’s race was just released and draws the conclusion:
Creigh Deeds is now within five points of Bob McDonnell in the race to be Virginia’s next Governor, and most of the remaining undecided voters are Democrats.
Their new poll shows Republican Bob McDonnell with a 48% – 43% lead.
Their analysis draws the conclusion that the race is tightening. While that may be true, their data does not support that conclusion due to the political affiliation of the sample.
On the Surface, it would appear true by looking at the most recent three polls.
The 9/29/2009 Poll shows McDonnell up 48%-43%
The 9/01/2009 Poll shows McDonnell up 49%-42%
The 8/04/2009 Poll shows McDonnell up 51%-37%
It does appear that Deeds is climbing and McDonnell is sinking.
However, if we look at the political makeup of the people polled, it invalidates any comparison of these polls.
In the 8/4 poll, 52% voted for McCain and 41% voted for Obama. And the results of that split show McDonnell 51% to Deeds 37%. My conclusion is the voters that voted for McCain are planning to vote for McDonnell and the ones that voted for Obama plan to vote for Deeds.
In the 9/1 poll, more Obama voters were sampled and fewer McCain voters were included. 49% of the sample voted for McCain and 42% for Obama. Exactly the same split for McDonnell (49%) and Deeds (42%).
The latest poll actually sampled more Obama voters (48%) than McCain voters (45%), yet McDonnell still leads by 48% to 43% for Deeds. This is the best indication yet that Obama voters must be moving to McDonnell.
The only thing these three polls prove is that if you survey a progressively fewer number of McCain voters and a progressively increasing number of Obama voters, the number of votes for Republicans declines and the number of votes for Democrats increases.
But the big news is that if you survey more Obama voters than McCain voters, McDonnell still leads by 5%. The only way that can happen is if Obama voters are moving to McDonnell.
And for the mathematically inclined, you will notice that when you compare the 8/4 McCain voter count to the 9/29 McCain voter count there is a decline of 12%.
Conversely, you will notice when comparing the number of Obama voters on 8/4 to the number of Obama voters on 9/29, there is an increase of 23%.
This is a DELTA of 35%!
So, using these figures to compare statistical movement between the 8/4 poll and the 9/29 poll, we must add 12% to McDonnell’s vote total and deduct 23% from Deeds’ totals.
This results in a corrected poll result of McDonnell 54% – Deeds 33%.
The most disheartening news for the Deed’s camp is that even with a more favorable statistical base, leaning 35% more towards him, his negative numbers continued to increase.
And the thesis questions show 34% less likely to vote for McDonnell. It’s a pretty fair assumption most of these also polled for Deeds. And the 7% that were more favorable towards McDonnell due to the thesis were most likely McDonnell’s anyway. The fact that 59% see it as a non issue, combined with Deeds growing unfavorable numbers is simply not good at all for Democrats.
This is a trend in most of the other polls released lately. Earlier polls sampled McCain voters or Republicans more heavily. Now, pretty much every poll is sampling more Democrats / Obama voters more.
Going by just the PPP polls, this is looking like November 3 is going to be a bad night for Democrats.
Can’t we get any polls to use samples with approximately the same demographic mix? It is totally irresponsible for these pollsters to ignore a 35% swing in sample bias and draw a conclusion as if we were comparing apples to apples. We most certainly are not.