With less than a week left to go until Election Day, the odds of the Democrats controlling 50 seats or more in the Senate are as high as they’ve been: 72 percent. That’s still far from a sure thing, but the trendline looks very good. There’s one other first that we’re seeing: for the first time, the median number of Democrat-held seats is now 51, not 50.
Fifty-one seats has two advantages. One, it gives the Democrats a bit of a cushion going into 2018, a midterm election where they’re defending a number of red-state seats that they won in 2012. And two, it keeps the Democrats from having to depend on their least-liberal member to reach 50 on non-cloture votes, so it keeps Joe Manchin from being perpetually in the driver’s seat. (Of course, any 5-seat gain includes Evan Bayh, so that would still leave Evan Bayh in the driver’s seat.)
The main reason that we’re seeing our Senate odds increase is the race in Pennsylvania, where Katie McGinty now has an 89 percent chance of winning, based on a polling average of 45-41. Her race with Pat Toomey has been very back-and-forth all year, but she’s led the last eight polls; she hasn’t trailed since an Emerson poll from mid-October, and a Franklin & Marshall poll from a few days ago giving her a 12-point lead is the coup de grâce.
If you’ve been following our Senate model closely, you’ve probably seen the pattern where there are three races (Illinois, Indiana, and Wisconsin) that were all pretty solid in the Democrats’ corner, and then three more races (Nevada, New Hampshire, and Pennsylvania) hanging around in the “tossup” realm. However, what’s surprising here is that Pennsylvania is now tied for the second-likeliest Democrat pickup, as Wisconsin and Indiana have both become more competitive in the last few weeks. (In fact, just in the last few days we’ve seen a Marquette poll of Wisconsin giving Russ Feingold only a 1-point lead, and a Monmouth poll of Indiana finding Evan Bayh in a tie. Those have pushed Wisconsin down to 89 percent odds, and Indiana down to 79 percent odds.)
That may sound alarming, but those odds are still high enough that the Democrats still win both Wisconsin and Indiana in most of our simulations. Here’s an illustration; you might remember that a few weeks ago, we showed you what the 10 most common permutations were on all our simulations of the presidential race. Along those same lines, we can also show you what the 10 most common Senate permutations are.
(We aren’t going to list every race, just the competitive ones. It shouldn’t be a surprise that the Democrats win, say, the Maryland and Oregon races, and the GOP wins the Idaho and Kansas races, in nearly every simulation.)
1. Dems win Illinois, Indiana, Nevada, New Hampshire, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, Missouri, North Carolina, Ohio (51 seats; 4.8 percent of all simulations)
2. Dems win Illinois, Indiana, Nevada, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, Missouri, New Hampshire, North Carolina, Ohio (50 seats; 4.7 percent of all simulations)
3. Dems win Illinois, Indiana, Missouri, Nevada, New Hampshire, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, North Carolina, Ohio (52 seats; 3.8 percent of all simulations)
4. Dems win Illinois, Indiana, Missouri, Nevada, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, New Hampshire, North Carolina, Ohio (51 seats; 3.6 percent of all simulations)
5. Dems win Illinois, Indiana, New Hampshire, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, Missouri, Nevada, North Carolina, Ohio (50 seats; 2.8 percent of all simulations)
6. Dems win Illinois, Indiana, Nevada, New Hampshire, North Carolina, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, Missouri, Ohio (52 seats; 2.7 percent of all simulations)
7. Dems win Illinois, Indiana, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, Missouri, Nevada, New Hampshire, North Carolina, Ohio (49 seats; 2.7 percent of all simulations)
8. Dems win Illinois, Indiana, Nevada, North Carolina, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, Missouri, New Hampshire, Ohio (51 seats; 2.6 percent of all simulations)
9. Dems win Illinois, Indiana, Missouri, Nevada, New Hampshire, North Carolina, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, Ohio (53 seats; 2.1 percent of all simulations)
10. Dems win Illinois, Indiana, Missouri, Nevada, North Carolina, Pennsylvania, Wisconsin; GOP wins Arizona, Florida, Iowa, New Hampshire, Ohio (52 seats; 2.0 percent of all simulations)
Nine of those 10 permutations are 50 seats or more, which is a “win” for our purposes. Note that in the top 10 permutations, the Democrats win Illinois, Indiana, Pennsylvania, and Wisconsin in each one; all the variation happens within Missouri (where the Democrat odds of victory are 43 percent), Nevada (64 percent), New Hampshire (exactly 50-50), and North Carolina (28 percent). You have to drop all the way down to the 15th most common permutation to find one where the Democrats lose one of the IL-IN-PA-WI quartet (and in that permutation, the Democrats manage to hit 50 seats anyway despite losing Indiana, by still winning both Nevada and New Hampshire).
Of the 30 most common permutations, only 6 involve the Democrats finishing under 50 seats! That’s only 20 percent of the time, so you might think that even the 71 percent win probability that our model gives you is still too low. However, there’s more downside for the Democrats once you get into the very rare permutations; in other words, when you’re getting into the scenarios that rarely show up, there are more that involve losing Pennsylvania or Wisconsin than involve picking up Arizona or Florida.
While the Senate model is good news, in that Hillary Clinton is likely to be able to get her Supreme Court and cabinet nominees approved easily with a Democrat-controlled Senate, let’s not get ahead of ourselves; there’s still also the wee matter of getting Clinton herself elected, a ride which got noticeably bumpier in the last week. That’s still likely to happen, though; our model sees her likelihood of winning at 92 percent, which is down from her peak of 96 percent last week (though up a tick from Tuesday’s 91 percent).
What’s driving that decline is her odds dropping in some of the key states: North Carolina is now at 72 percent odds, down from a peak of 87; Florida is now at 80 percent odds, down from a peak of 91; and Nevada is now at 68 percent odds, down from a peak of 80. North Carolina and Nevada, in particular, each had a surprisingly bad poll pop up in the last few days; in North Carolina, it was a Survey USA poll putting Donald Trump ahead by 7, while in Nevada, it was a CNN poll putting Trump up by 6.
I’d be more worried about those numbers, though, if they weren’t contradicted by early voting performance, which seems good for the Democrats in North Carolina and even better in Nevada. (In addition, that North Carolina poll got crowded out by a number of other polls in the last few days with entirely mundane results; Quinnipiac, for instance, put Clinton ahead by 3 in the Tar Heel State on Wednesday.)
What’s not happening, though, is Clinton’s odds dropping in the must-win states: Colorado, New Hampshire, Pennsylvania, and Virginia (and Michigan and Wisconsin). Clinton’s odds remain at 99 percent in all of those states, with the exception of Colorado, which has fallen to a perilous 98. Clinton has at least a 5-point lead in our polling averages of all those states (again, except for Colorado, where the lead is closer to 4). That’s not a holdover from the pre-Comey memo days; there were three different polls on Wednesday of Pennsylvania (from CNN, Monmouth, and Quinnipiac), all giving Clinton a lead of either 4 or 5. (Polls that produce numbers right in line with the averages instead of being SHOCK POLLS!! tend to get overlooked in the hubbub, but they exert just as much weight on the averages as the shiny objects do.)
And, as I’ve incessantly reminded you for the last several months, those six states, plus the solid blue states, take her past 270 electoral votes as is. Florida, Nevada, and North Carolina are all icing on the cake. That firewall is what’s keeping her overall odds as high as they are, even if the icing is looking a little thinner this week.
That thinning of the icing may also, in fact, not even really be happening, in terms of actual voter intent, despite what the polls are telling you. As I’ve mentioned several times before, “differential response bias” is a problem that plagues pollsters and is one that they’re only beginning to understand. “Differential response” occurs when the supporters of one side in an election get a piece of bad news (and the Comey memo probably qualifies as that, though it’s more cryptic than bad; most likely what agitated Democratic loyalists was the media’s breathless speculation on Friday afternoon about the implications, until it gradually dawned on the media that they’d been played by the House Republicans’ spin); the result is that they’re less likely to respond to polls when they’re worried or upset about that news.
It tends to fade even in the polls after a week or two … as we’ve seen many times before, whether it was Barack Obama’s poor performance in the first debate in 2012, or Hillary Clinton’s bout of pneumonia this September. And new polling methods are actually helping us capture snapshots of the phenomenon. YouGov, for instance, uses a large pool of potential respondents that it contacts for its online polls; sometimes it recontacts them to participate again. In the wake of the recurrence of the email flap, YouGov drilled down into the response rates for their recontacts.
They found that, indeed, people who had previously participated in their polls and expressed support for Clinton were less likely to participate in their most recent survey, than people they had contacted before and had expressed support for Trump. It wasn’t a huge difference (3 percent), but that’s enough to swing the results somewhat for a pollster who isn’t looking for that problem and controlling for it. (Survey Monkey, another pollster that uses online non-probability sampling from a huge pool of potential respondents, also remained remarkably stable from before and after Comey-gate, much like YouGov.)
This, of course, is not an invitation to rest easy for the remaining days of the campaign; this weekend is prime time to get out and do some GOTV to make sure that the victories that are model is seeing actual happen. But, overall, the polling numbers say that you should be pretty confident, at this point.