What would be our sense of the midterm landscape if Research 2000 and Quinnipiac released as many polls as Rasmussen does?
This question is not meant to denigrate Rasmussen. I am not dismissing his results by pointing out that his polls represent a set of assumptions that are on the GOP-friendly end of the spectrum of possible turnout patterns and partisan breakdown, whereas Research 2000 appears to be using assumptions that result in more favorable results for Democrats and whereas the model used by a group like Quinnipiac makes its results fall somewhere in the middle. Case in point: Quinnipiac’s new Ohio poll finds Governor Ted Strickland leading 44% to 39% while Rasmussen has him trailing by the sort of decisive margins an incumbent rarely recovers from.
Another case-in-point is Illinois. A new Research 2000 poll finds very positive numbers for Democrats, with Alexi Giannoulias leading Mark Kirk 43% to 36% and Governor Pat Quinn up by double-digits against the two Republicans who are fighting over who won the February 5th primary (46-35 against Kirk Dillard, 47-32 against Bill Brady). These numbers are almost hard to believe, but it’s unclear why that would be: Three independent polls of the Kirk-Giannoulias match-up have been released over the past month. One had Kirk leading 46-40 (Rasmussen) while the two others had Giannoulias up outside of the margin of error (Research 2000 and PPP). In short: The conventional wisdom that has emerged of a front-running Kirk and a struggling Giannoulias is certainly not backed by a polling consensus.
The problem arises not from different pollsters’ differing assumptions but from Rasmussen amazingly prolific rhythm: The pollster typically tests half-a-dozen states a week, whereas no more than two for PPP and at most one for Research 2000, Quinnipiac, Mason-Dixon or Suffolk. This has led to a situation in which the Rasmussen model of what the fall electorate will look like is dominating our understanding of the political landscape.
This is especially true in states that few if any other polling firms test. Indeed, in states that are often polled by a variety of pollsters (say New York, Connecticut, Nevada, Florida and North Carolina), we can confront Rasmussen’s findings to those of other surveys and thus avoid relying on a single poll. For instance, a new Rasmussen poll released today has Senator Richard Burr in a very solid position. Without dismissing his take on the race, we are well-served by being able to compare his numbers to those PPP released earlier this week, which had Burr looking more vulnerable.
The situation is very different in a state like Wisconsin: As far as I can see, the only non-Rasmussen poll that has been released over the past 6 months testing the Governor’s race and Russ Feingold’s vulnerability is a PPP survey dating back from November which showed Democrats in a strong position. That makes our view of Wisconsin’s political situation far more dependent on Rasmussen than is healthy. The same is true of Colorado, which only Rasmussen and Research 2000 have tested since September; the two have found very contradictory results as to Michael Bennett’s vulnerability, but since R2000 visited the state only once whereas Rasmussen has released four polls, we have grown more used to seeing Bennet in a catastrophic situation. This phenomenon is perhaps most consequential in Missouri, which no pollster but Rasmussen has tested since mid-November.
Once again, none of this is meant to suggest Rasmussen is distorting its numbers or that his polls are unreliable; after all, in many of these states (starting in Missouri) not only Rasmussen’s raw numbers but also his trendline have shown bad news for Democrats. The point is that no individual poll provides a reliable snap-shot of the electorate, Rasmussen no more than others, and that we should keep this in mind when commenting on the landscape in places like Wisconsin and Missouri.
The usual full polling round-up will come tomorrow morning.