It appears we will be plagued with confusing, dueling poll results right to the end of this Senate election. On Friday, SurveyUSA released a poll showing Rossi and Murray tied at 47 percent. An hour or so later, the final UW/KPLU/KCTS Washington Poll came out, showing Patty Murray ahead 51 to 45 percent.
They can’t both be right, can they?
Maybe. The Washington Poll shows Sen. Murray winning with 51 percent of the vote. That seems entirely possible. Conventional wisdom in the polling world holds that most if not all undecideds will break for the challenger, so a 51 to 49 percent Murray victory is certainly plausible. But is Murray really ahead by 6 percent among likely voters heading into the weekend, or is the race tied, as SurveyUSA maintains? I think there is good reason to put more stock in the SurveyUSA finding than in the UW's poll.
This was SurveyUSA's final election poll, the one they will be judged on regarding accuracy, and they made some changes in methodology. Cell phone only voters were included. Some voters were called by live interviewers. And most importantly, their likely voter sample is significantly different. SurveyUSA's previous poll was drawn from a sample made up of 36 percent Democrats, 27 percent Republicans, a 9 percent D advantage. This poll shows only a 4 percent advantage for Democrats, 33 to 29 percent.
As I have argued before, I think there is strong evidence that the partisan gap in Washington state has narrowed to less than 5 percent. This subtle shift in the makeup of the sample produces a dead-even race, rather than a Murray advantage. That is consistent with Thursday's Rasmussen poll, and is, I believe, on the money.
There are, on the other hand, good reasons to question the Washington Poll. First is the length of time the poll is in the field. The UW polled in two "waves," the first wave from Oct. 5 to 14, the second Oct. 18-28. Then they created a likely voter sample comprised of responses from both waves. That means some of the people included in this sample were polled three weeks ago. Polls are meant to be a snap shot in time, and three days is about the maximum amount of time a poll is normally in the field. The Washington Poll’s methodology is highly unusual.
Second is how the UW created its likely voter screen. It appears that they simply used vote history to determine likely voters. If so, they are missing the “enthusiasm gap” that other pollsters are finding. In fact, Matt Barreto, the UW professor who supervises the Washington Poll, says his data shows that unlike in other states, Democrats are more enthusiastic about voting than are conservatives.
So the results of the Washington Poll are based on a likely voter sample that assumes that Democrats are more motivated to vote than Republicans. That would be contrary to all other available data. The UW did not provide cross tabs for their likely voter sample, but it appears to include about an 8 percent Democratic advantage. Again, out of step with other polls.
I love the University of Washington, but I think SurveyUSA and Rasmussen have given us a more credible snap shot of where this race is heading into the final weekend.