SurveyUSA
DEM PRIMARY (D voters only)
04/27 04/06 03/30 03/01 02/01
Pollster SUSA R2K PPP PPP PPP
McAuliffe 38 19 18 21 18
Moran 22 24 22 19 18
Deeds 22 16 15 14 11
Undec. 18 41 45 46 53
MORAN-MCDONNELL
04/27 04/06 02/05 12/04
Pollster SUSA R2K Ras Ras
McDonnell 46 37 39 37
Moran 34 36 36 41
Net R+12 R+1 R+3 D+4
MCAULIFFE-MCDONNELL
04/27 04/06 02/05 12/04
Pollster SUSA R2K Ras Ras
McDonnell 46 40 42 41
McAuliffe 39 33 35 35
Net R+7 R+7 R+7 R+6
DEEDS-MCDONNELL
04/27 04/06 02/05 12/04
Pollster SUSA R2K Ras Ras
McDonnell 44 38 39 39
Deeds 39 31 33 39
Net R+5 R+7 R+6 Tie
It is with great reluctance that I call an unfavorable poll an outlier, but there are several major problems with the latest SurveyUSA poll that purports to show Terry McAuliffe blowing open the race and Brian Moran collapsing against Bob McDonnell.
1. No likely voter screen. For the head-to-head general election matchups, SurveyUSA polled registered voters and ran no likely voter filter. That alone should indicate that the results are worthless, but...
2. Of those registered voters, a full third are supposedly likely primary voters. In the Research 2000 poll, they accomplished polling both the general electorate and primary electorate by oversampling likely primary voters and then re-weighting results. Here, they ran a pass for registered voters only, and determined that a full 30% of them would vote in the primary. Democratic primary turnout in 2005 was less than 3%, the Obama-Clinton primary of 2008 attracted only 15%. Essentially, what they did was find 1396 registered voters, and deduced that 409 of them were going to vote in the primary, whereas if a real likely voter filter was run, of those 1396 registered voters, perhaps 200 tops would have made it through.
3. Severe Northern Virginia undersample. In 2006, Northern Virginia made up 41% of the primary electorate. SurveyUSA assumes that it will make up only 31%. |