Our busiest polling day in several weeks:
Let's get the easy ones out of the way first.
Kansas and New Jersey were two states that the campaigns talked early
on about being competitive, but they have since retreated into their
respective red and blue corners. Each of these polls have moved
slightly toward Obama from their previous edition, but otherwise
there's nothing to see here.
Wisconsin may have tightened some; McCain's 5-point deficit in the Strategic Vision
poll is the closest he's been in any individual survey since April.
Still, the race has consistently polled outside the margin of error,
and Strategic Vision's polls are notoriously Republican-leaning --
their only prior poll of the state, taken at about the time of the
Democratic primary in February, had Obama ahead by just one.
We're listing a different result from the Franklin & Marshall
poll of Pennsylvania than some other outlets, using Obama's +8 among
registered voters rather than his +5 among likely voters. A couple of
weeks ago, in the brouhaha over the Gallup-USA Today poll that showed a huge likely/registered voter split, we established a policy
of deferring to the registered voter version of a poll when given the
choice, until the time of the first Presidential Debate on September
26, after which we will switch to the likely voter version. There is
some evidence that likely voter models are more accurate near to the
election -- but very little that they work this far out, and in fact there is some evidence to the contrary.
Of
course, it is relatively unusual for a pollster to list both registered
voter and likely voter numbers -- most just pick one or the other. But
when in doubt, we're going to defer to allowing voters to speak for
themselves -- which means using the registered voter version until
there is a good empirical argument not to.
That notwithstanding,
the difference between a +5 and a +8 hardly matters, considering that
Pennsylvania has been polled extensively, and not really shown any
daylight for John McCain in a couple of months now.
Nevada has
been one of the most vexing states to us poll junkies, with only
Rasmussen surveying the state on a regular basis. Their latest result
has McCain up by 3 -- a reversal from last month, when Obama had pulled
into a 2-point lead. As I said last month, I think these numbers may be
lowballing Obama somewhat, considering that the Democrats have made significant gains in party registration, something which might be hard to detect in a poll (like Rasmussen's) that weights based on party ID.
Lastly, two polls
out of Virginia show the race essentially tied: McCain leads by one
point in the Rasmussen poll, the same margin he held last month, and by
a fraction of a point in the first InsiderAdvantage survey of the
state. The last six Virginia polls have all been in a tightly-banded
range between McCain +1 and Obama +2, and it looks like the state may
come down to turnout operations.
--Nate Silver