About | RSS

TPMDC

Poll: Deeds Gaining On McDonnell In Virginia

Eric Kleefeld | September 1, 2009, 11:28AM

VA-GOV

Send to a friend!

The new survey of the Virginia gubernatorial race by Public Policy Polling (D) suggests that Democratic candidate Creigh Deeds could be seriously catching up with Republican Bob McDonnell, as Democrats become more motivated and McDonnell takes heat over his controversial right-wing law school thesis.

In the top-line, McDonnell still leads with 49% to Deeds' 42%. However, this is a big shift from the 51%-37% McDonnell lead from a month ago. There has been a significant shift in the make-up of the likely voter pool: A month ago, respondents had voted for McCain by a 52%-41% margin, while the new pool is at McCain 49%-45%. This is still a long way from the actual result last fall, when Obama carried the state 53%-47%.

One big change, according to PPP communications director Tom Jensen, is that the likely voter pool shifted from 37% Democratic to 47% Democratic in the days of sampling after the story broke about McDonnell's thesis, and Deeds then led by eight points in that sample within the sample. However, the margin of error in the post-thesis sample is ±10.8%, and Jensen told me we'll need more post-thesis polling to really see whether this is a fluke, or a real sign of Democrats becoming energized.

Comments (9) | Join the Conversation!

Recommend Recommend (1)

The Commenter Formerly Known as NCSteve

September 1, 2009 11:42 AM

Thus do we see how pollsters use their likely voter models to generate marketing buzz from the "free" polls they run in the months when no one's paying them to do election polling.

A sophisticated likely voter model can do a great job of identifying people who will vote in the two to four weeks immediately before the election. Thus, the poll gets a reputation for accuracy. Before that, however, all it does is identify who's sufficiently pissed off about something to pay attention to politics before a month before the election.

Most of the time, the answer to the question of "who's pissed off?" is "Republicans" because they're besotted with psuedo-news and thus always pissed off about something. Weight a poll up with a likely voter model way out ahead of an election and it tells you nothing about who's actually going to vote come Election Day. What it does do, however, is generate headline grabbing results that are badly skewed toward people who are unusually pissed off about something thereby getting the pollster's name in the article and enhancing its brand awareness.

This is why if you bring up the presidential approval numbers at pollster.com and eliminate the Rasmussen Daily from the regression, you'll see a marked improvement in Obama's number. Rasmussen's likely voter model skews the weighting heavily Republican right up until a week or three before the eleciton.

Rasmussen, incidentally, is the only poll in the Pollster regression whose omission makes a noticiable difference to the numbers. (Because of the skewing and because, as a daily, it provides a lot more datapoints than the non-dailies.)

Reply | Flag Abuse