Univ. of Washington for KPLU (10/18-28, likely voters, 10/5-14 in parentheses):
Patty Murray (D): 51 (50)
Dino Rossi (R): 45 (42)
Undecided: 6 (8)
(MoE: ±4%)
SurveyUSA for KING (10/24-27, likely voters, 10/11-14 in parentheses):
Patty Murray (D): 47 (50)
Dino Rossi (R): 47 (47)
Undecided: 6 (3)
(MoE: ±3.8%)
Two new polls out today both show tightening, of either two or three points in the last two weeks, in the Washington Senate race. This race has become increasingly pivotal in the last few weeks as well… quite literally, in that Nate Silver calls it the Senate’s entire pivot point for control — where, in the simulations he runs where the GOP gets a 10-seat pickup, Washington is most frequently the last race across the line. (Of course, it’s worth considering that as West Virginia seems to be getting better for Joe Manchin, there are increasing chances that Dems could lose Washington and keep the majority regardless of the Washington outcome, which is why overall odds of keeping the Senate are still hovering near 90%.)
What the closing of the gap means is quite debatable, though, depending on what method you use. For the Univ. of Washington, it’s still a Murray victory, as she’s already over 50 (at a point when, presumably, most people have sent in their ballots); her 8-point lead is down to 6. (They also find the reverse-enthusiasm-gap that seems unique in Washington, finding only a 4-pt edge among RVs.)
For SurveyUSA, it drops her into a tie, though. There are a couple odd things with the SurveyUSA poll, though; first, it’s strange that undecideds would shoot up in the closing stages of the race, particularly since the state SoS office reports that a majority of ballots have been sent in. Maybe those who haven’t sent ballots yet are still trying to decide; it’s hard to gauge, SUSA doesn’t include the vote breakdown among people who have already voted, which is an odd choice since they’ve done that in some other races, and it’s even more relevant in the (almost) all-mail-in Washington. Also (h/t to Taniel for pointing this out), there’s a steep dropoff in the Dem/GOP makeup of this sample from the last sample from two weeks ago: 33D-29R today, versus 36D-27R before. There’s no party registration in Washington so this is just self-identification, but it’s an abrupt switch.
One other consideration is the cellphone user gap, which has seemed particularly pronounced in Washington of all states (as seen by the wide split between live-calling Elway, UW, and CNN/Time, vs. auto-dialing SurveyUSA, Rasmussen, and PPP). I’m not a fan of mindlessly applying corrective formulas to poll data, but Nate’s most recent post on the “house effects” generated by the different categories of pollsters may be instructive here: all robocallers have an R+2.0 lean, while live polls have a D+0.7 lean (although that may have to do simply with the sheer weight on the R-side of the spectrum brought by Rasmussen’s massive volume of polling). In particular, SUSA has the most pronounced house effect this year, of R+4.0, even more than Rasmussen at R+2.1.
One major caveat, though: SurveyUSA used a live-caller overlay on this particular poll, and find that while the cellphone users they reached did tend to lean Democratic, it doesn’t matter much for the final totals. They’d done this once before on a Washington poll over summer, and come to much the same conclusion. That seemed odd at the time and still does, as it would tend to contrast with recent Pew research that showed a 5-point difference between cellphone-inclusive and non-cellphone samples in the generic ballot. With that in mind, I’ll leave it to you just how much special sauce you want to add to make sense of the results… or you can just average them out to 3, which is pretty close to where Murray’s leaked internals (+4) from a few days ago put the race.