NJ: Christie 47, Corzine 37 (Quinnipiac 8/25-30)
Emily Swanson | September 1, 2009
Quinnipiac
8/25-30/09; 1,612 likely voters, 2.4% margin of error
Mode: Live telephone interviews
(Quinnipiac release)
New Jersey
Favorable / Unfavorable
Chris Christie (R): 41 / 30
Jon Corzine (D): 34 / 57 (chart)
Chris Daggett (i): 8 / 4
Job Approval / Disapproval
Gov. Corzine: 34 / 60 (chart)
Sen. Lautenberg: 43 / 41 (chart)
Sen. Menendez: 40 / 37 (chart)
Pres. Obama: 51 / 43 (chart)
2009 Governor
Christie 47%, Corzine 37%, Daggett 9% (chart)
By Emily Swanson | September 1, 2009 8:05 AM | Permalink | Comments (3) | TrackBacks (0)
Comments
Hi Mark,
I've got a question for you.
Playing around with the tools for the governor's race here I saw that excluding Zogby brings Christie's numbers down by 2 whole points and Corzine's by half a point.
However, the last Zogby poll in the database was conducted in AUGUST 08, more than one year ago.
Do you really think that a result from a poll that was conducted more than one year ago with dozens of more recent results available should have so much impact on the numbers, or is something wrong with the weighting algorithm?
Maybe include (or increase if you have that) something like a decay function?
Posted on September 1, 2009 10:25 AM
Good point. Polls should drop off after 6 months. Things change so quickly these days that older polls like that are worthless.
Posted on September 1, 2009 11:21 AM
@Rasmus:
Short answer: No. You're right, a single poll that old should not make any difference for the "nose" of a trend line based on 41 polls.
Longer answer: As you surmise, something does not appear to be working in the algorithm that re-draws the trend line once you filter out a pollster. It is supposed to keep the smoothing consistent, but if you notice, removing that one old poll makes the line significantly flatter (i.e. less sensitive). We need to do some debugging, and I'll report back.
Posted on September 1, 2009 11:53 AM
Post a comment