Sunday, August 01, 2010

What does Strib poll portend for primary?

I'm reluctant to put too much stock in this morning's release of a Minnesota Poll (Star Tribune / Princeton Survey Research) on the MN-Gov race. Star Tribune polls are not held in high esteem by meta-pollster Nate Silver, and held in even lower esteem by the GOP. However, one of the primary GOP criticisms of the Star Tribune's polling has been what they view as a DFL skew in the party identification numbers, and even Tony Sutton had to admit that this poll's ID numbers weren't out of line.

I'm also very skeptical of drawing any conclusions about the DFL primary, since the sample size is so small that the margin of error for DFL primary voters is 7.8%. Of course, the campaigns are spinning the heck out of this poll. Well, I say "the campaigns," but it would be more accurate to say "Kelliher and Entenza's campaigns," since Dayton's perfectly willing to let a ten point lead speak for itself.


In potential general election matchups, the DFL candidates (save Entenza) led Tom Emmer and Tom Horner by statistically significant margins (overall MOE 4.3%)

In the DFL Primary race, Dayton leads Kelliher, but the huge MOE (7.8% for this sample) means you should take these results with a big grain of salt because these results are based on a subset of 175 or so (my estimate) of the 900 respondents.

Entenza's Spin

But here's where it get interesting. Entenza's campaign's spin is that the Minnesota Poll completely botches the primary turnout number. From MinnPost:

“The Star Tribune poll released today assumes an unimaginably large voter turnout of 61%. ... In addition it has a margin of error of nearly 8%, which means Entenza could be at 25%, Kelliher at 22% and Dayton at 32%.

Since the poll suggests a turnout of 61% of 3.9 million voters, that means it expects that 2,379,000 people will vote in the primary."

This spin is a good example of abusing poll numbers and playing fast and loose with math. Entenza's right that the MOE is very high, but the "could be" scenario is an absolute best case for Entenza, and even then he trails by a significant margin. It is just as likely that Entenza's true support number is 9%. We could more accurately express the results of this survey as a band of possible results like this:

The charge of "61% turnout projected" is one that puzzled me for a moment. I presume this is the math of the Entenza campaign (and I welcome corrections if I am wrong.)
Total universe: 3,900,000 voters
Total sample: 902
DFL portion of sample: 902 x .32 = 289
Likely DFL primary voters: 289 x .61 = 176
The Minnesota Poll's 7.8% MOE for DFL primary voters indicates a sample of under 200. For example, if you use American Research Group's margin of error calculator with a 3.9 million voter universe, the 176 person sample returns a MOE of about 7.4%.

A 61% turnout for a general election would be a bit low. For a primary it is very high, historically gubernatorial primary election turnout has ranged from 14-28% in the last couple of decades. But even if poll respondents overidentify as likely DFL primary voters, that doesn't mean that the poll "assumes a 61% turnout." Most observers believe that we're looking at a low turnout election because of the earlier date, but it's hard to know exactly until you've had one. Does this mean the poll understates Entenza's support or overstates Dayton and Kelliher's support? Probably not.

For one thing, Entenza's potential success rides on turning out less likely primary voters. For examples, consider his targeting of minority voters, an endorsement by Slug, etc. Additionally, he's sailing into the breeze when courting the core DFL'ers who will nearly always show up to the polls. Kelliher has the endorsement which means something to a significant slice of the DFL activist core. Dayton appears to have the strongest support among senior voters who are reliable primary voters. If anything, a huge turnout would probably be better for Entenza than the other two candidates.

And lastly, if Entenza's campaign could point to even one poll that showed that he didn't trail both Dayton and Kelliher, they might have a point. As it is, Entenza risks suffering the fate of a fading third candidate where soft supporters start looking at more viable alternatives. Especially since Entenza consistently does worse in the general election poll matchups, electability-focused DFL primary voters may start to look elsewhere.

Kelliher's Spin

There's one line in the Star Tribune coverage that has been retweeted by MAK's campaign all over the place this morning:
The poll also suggests that come November, Independence Party endorsee Tom Horner -- who trails the field -- could draw equally from Democrats and Republicans in a general election unless Kelliher is the nominee. In a Kelliher-Emmer match-up, Horner would draw far more from Republicans than from Democrats. That would be a change from the dynamic of previous elections, when Democrats have been more vulnerable to third-party candidates.
Without access to the crosstabs it's really hard to evaluate this claim. It's seems strange considering that Horner's topline numbers hardly budge in matchups with Dayton and Kelliher. It makes some sense to me, since Kelliher's choice of John Gunyou seems to be designed to appeal to potential Horner voters, and she's not been nearly as definitive in her revenue raising as Dayton's unambiguous "tax the rich" message.

Expect tonight's debate to feature a hefty dose of Kelliher making this electability argument centering around her ability to appeal to moderates. Since Kelliher's poll results are in the Minnesota Poll's margin of error, and her head-to-head general election results are comparable to Dayton's, this is a credible line of argument. Not proven, by any means, but credible.

In the end, I agree with Mark Dayton's assessment as reported by the Star Tribune - "I think this will be a close election." But the further we go, the more it looks like a two person primary race.

Follow me on Twitter @aaronklemz

No comments: