Saturday, October 18, 2008

New Poll: Collins+13

From Research 2000:

Rep. Tom Allen (D): 40 (38) (34)

Sen. Susan Collins (R-inc): 53 (57) (56)

The poll, in the field from October 14-15, was commissioned by Daily Kos. (But Research 2000 is a non-partisan pollster.) The numbers in parenthesis refer to polls from the same firm conducted September 8-10, 2008 and October 15-17, 2007, respectively.

UPDATE: As much as I respect Kos' knowledge and judgment when it comes to elections, I don't share his inference that these results come close to sealing the deal for Collins. And I'm pretty confident they're not doing victory dances at Collins HQ. There are three reasons.

1. The previous Research 2000 poll gave Collins a 19 point lead--six points more than the Rasmussen result from the same time frame. It was the largest lead any survey had produced for Collins in months, even taking into account a Survey USA poll in which 18-34 year-old respondents tilted overwhelmingly (and unrealistically) toward Collins.

Now, maybe Research 2000 is dead-on. But their work could also be an outlier. And if that's true, the new poll's trend--a six point smaller advantage than last time for Collins--suggests that her lead may actually be in the high single digits.

Hence the recent Collins +10 results from Rasmussen and the eight point lead found by the Mellman Group.

All things being equal, I'm inclined to trust Research 2000 over the DSCC-sponsored poll from Mellman. But the fact that no one else duplicated the 19 point spread found in Research 2000's September poll raises real doubts.

2. When you dig into the data, you learn that Collins builds her lead by earning the support of 32% of Democrats and 48% of CD-01 voters.

Both of those numbers are plausible. But the 32% figure among Democrats is much higher than what we saw in either the most recent Rasmussen survey or the Collins-friendly result from SurveyUSA.

That Survey USA poll also showed a mere four point spread between the performance of the candidates in CD-01 and CD-02. The Research 2000 poll shows an 18 point spread.

Clearly, someone is wrong here. But it will take more polling to find out who.

3. The poll appears to assume equal turnout in both congressional districts.

I don't have the time to look into this now, but my sense is that, historically, CD-01 has generated higher turnout. And in this election in particular, my assumption has always been that many more voters will go to the polls in CD-01 than in CD-02.

Bottom line: No one who thought this race was up for grabs yesterday has reason to think the result is a foregone conclusion today.

In most years--in a less volatile political climate in which there were fewer question marks--even a narrowing outside-the-margin-of-error lead 19 days before an election would be enough to make an incumbent feel safe.

But this isn't most years.

No comments: