I generally agree with
this analysis by Bob Holsworth:
But most of the explanation can be found in the diametrically conflicting models of the likely electorate found inside the poll numbers.
Public Policy Polling has an implicit turnout model heavily tilted toward the Democrats...
Survey USA has a likely voter model that is tilted in the exact opposite direction...
Here’s the difficulty.
In a gubernatorial election, less than 50% of registered voters will actually show up.
The fundamental issue for pollsters, then, is to determine who among this “less than 50%” is really going to vote.
So there you have the bottom line question -- who is going to show up on election day to cast their ballot? And no matter what poll you're looking at, they almost certainly don't know the answer right now. Having said that, it's not good news - as Bob Holsworth points out - that Democrats are "even trailing in [the poll] that has turnout tilting heavily toward the D’s." So, even if SurveyUSA is a complete outlier, we Democrats still have our work cut out for us in letting Virginians know why Bob McDonnell would be a disaster as governor, and why they should support Creigh Deeds. And we only have 33 days to go...
