I want to thank Lynn Bartels of the Rocky and Lori Weichel of Public Opinion Strategies for being so helpful with this post. Lynn tried valiently to answer my questions. When they exceeded the data she had in front of her, she passed me over to Lori, who spent way too much time on the phone explaining POS's methodology to me.
They both know I'm a blogger, so they both expect this information to show up on the sites. (Lori didn't know this at first, and assumed from the forwarded emails that I had some background in polling. So I guess the Pajamahudeen is making great strides in self-education.)
The answer to question #1 is: 39% Republican, 35% Democrat, 24% Independent, 2% Not Telling. This appears to oversample Republicans and especially Democrats, at the expense of Unaffiliateds. (It's worth noting that re-adjusting doesn't help Coors much, since he's still down 55-38 among unaffiliated voters.)
Here's what Lori had to say about that. I'll let you be the judge.
- Independents tend to shrink in number as you approach the election. They consider this to be a result rather than a demographic to be adjusted for, since it's a self-identification. (Dick Morris uses this method himself. -ed.)
- Democrats seem to be more motivated than Republicans. This is consistent with their new registrations, and with a higher-than-normal level of interest since February.
- This suggests that Democrats will be more likely to vote than Republicans. Also, going back to 2000, exit polling showed only a 1-point edge to Republicans in actual turnout.
- POS uses a random-digit phone call, rather than voter lists. They believe this lets them capture new voters, and voters in places like Weld County which don't require phone numbers to register. The latter helps preserve geographic balance. If other polls are using voter rolls, that could account for some of the difference.
- Lori also suggested that the presence of Nader may help Bush, but hurt Coors, since there's no Green running in the Senate race, but there is a Libertarian.
- Coloradoans are sophisticated voters. Sophisticated enough to feel comfortable splitting their ballot not only for elected office, but also for ballot initiatives and taxes.
So, we learn that POS interprets the oversampling of Democrats as a real phenomenon, indicative of Likely Voter status, rather than a quirk to be corrected, at this point in the election cycle. Even if we do correct for it, it doesn't help Coors much. He's getting killed among unaffiliateds. And POS has usually been identified as a Republican polling firm.
Looks, it's a small sample, it's at odds with four recent polls. The Republicans have tended to run ahead of the polls here for a few election cycles. Their GOTV effort now matches the Democrats'. If you assume that Coors has to make up four points rather than 11, it's very doable.
Going back to 2000, the Republican turnout exit-polled at one point better than the Dems', but Bush carried the state by 5 points. Perhaps that exit-poll followed a pattern where Republicans were less likely to answer the pollsters. Lori did say that most of the "Refused to Tell Party Affiliation" were Republicans. If so, the demographics in this poll may not be as reflective of actual turnout as Lori suggested.
We report. You decide.
Cross-Posted at Salazar v. Coors.