So the Democrats’ Research 2000/Daily Kos crew has released a poll today showing incumbent appointed U.S. Senator Michael Bennet one point ahead of leading Republican challenger Jane Norton. But, you say, wasn’t it just Friday — three days ago — that Rasmussen had Norton up 12 points on Bennet?
Rasmussen’s margin of error (MOE) is 4.5 points, which means at the bottom end Norton is ahead of Bennet by 7.5. Research 2000’s MOE is 4 points, which means at the top end Norton is ahead of Bennet by 3. Since the two polls were conducted at virtually the same time, the obvious conclusion is that one (or both) of them simply is wrong. So whom do we trust?
I don’t have access to any of Rasmussen’s internals, but Research 2000/Daily Kos shows that 239 Democrats were sampled, compared to 224 Republicans. We can make even further inferences from their internal numbers — which show Bennet winning among Democrats 74 to 5, Norton winning Republicans 79 to 6 and independents somehow splitting for Bennet 37 to 34.
If you apply the actual percentages of active registered voters from the Colorado Secretary of State (35.2% Republicans, 33.7% Democrats, 31.1% Unaffiliated/Minor Party), then the outcome looks a little different than Research 2000 says: Instead, Norton beats Bennet by 2 points (41-39).
Unless you are willing to project at this point that the national trend will break and Democrats in Colorado will turn out in bigger numbers than Republicans, then there is no doubt the Research 2000/Daily Kos polls are skewed at least 3 points in the Dems’ direction. (Interestingly, the same math applies to their polling of the governor’s race — which shows Scott McInnis and John Hickenlooper deadlocked at 43. Following active voter registrations, McInnis should be ahead 44-41.)
Even including the margin of error then, there’s still a small gap with Rasmussen. That tells me it’s likely Rasmussen projects some more voter energy from the Republican side this fall. But for the sake of argument, we’ll make a conservative projection and argue that the state of the U.S. Senate race now is probably about a 6-point Jane Norton lead over Michael Bennet. (Interestingly, a 6-point Norton win is what takers of our latest survey predicted.) Extrapolating for the other candidates, that would mean at least a modest 4-point lead for Ken Buck and a narrow 2-point edge for Tom Wiens.
At this point, then, I would recommend adding a 3-point fudge factor to anything that comes out of Research 2000/Daily Kos. Of course, giving the poll as much credence as I have in meshing results, it’s also important to note that Rasmussen was the most accurate of 20 major pollsters in forecasting the 2008 presidential election, while Research 2000/Daily Kos came in a respectable seventh.