If you were hoping for a nice, politics-free weekend, well, guess again; instead you got two days of constant Sturm und Drang, in increasingly absurd ways. If you’re more interested in the polls than the latest pundit derp, though, there wasn’t much change at all in the big picture. Hillary Clinton’s current odds of winning the presidential race are currently at 72 percent, which is a bit lower than where it was last Thursday (75 percent) but also a bit higher than last Monday (70 percent).
As for whether this weekend’s derp explosion starts to seep into the polls, we won’t really have a good idea until another week or so. National polls, which are more numerous and tend to have shorter periods in the field, might tell us something earlier, but our model is only based on state polls, which are fewer in number and thus are slower to show broad trends. It seems likelier, though (but not guaranteed, of course), that we’ve settled into a groove after Clinton’s strangely long-lasting convention bounce gradually wore off in late August, and things will just go sideways for a few weeks until the next big event when the non-diehards focus on the race, which is the first debate.
(If you’re wondering about the weird-looking map above, no, the weekend’s crazy events didn’t include the nuclear annihilation of the Mountain West states. Look closely, and you’ll just see that they're now very small. As we first discussed last Thursday, we’ve unveiled some cool new maps that go with our model. This includes your choice of a conventional-looking map, which we used to headline our Thursday article, or, as you see above, a cartogram, which is a map that weights the size of the states according to the electoral votes. Doing so requires some choices about how the shapes get distorted, and our answer is to try to preserve the overall shape of the country while leaving America’s Empty Quarter, well, empty.)
Probably the two most momentous polls of the last half a week were a CBS/YouGov poll of Ohio that gave unusually good results for Clinton (showing her up 7, compared with our average, which shows Donald Trump leading by a fraction of a point) and an NBC/Marist poll of New Hampshire that gave her unusually poor results (though still showing her up 1, which contrasts with our aggregate, which puts her up by an average of 6). The simplest explanation would be simply to say “they just sort of cancelled each other out,” but the more nuanced explanation is that neither of these polls even moved the needle within their respective states’ averages, seeing as how we have a lot of other polls in each state that tell us a different story.
If anything, Clinton’s odds in New Hampshire ticked down a bit more than they went up in Ohio, because we have somewhat fewer polls in New Hampshire. (There were, for instance, six polls released in August in New Hampshire, with Clinton leads of 5, 6, 9, 9, 10, and 15.) New Hampshire also has a bit more of a magnified effect on the overall odds because it's potentially the pivotal point in the quest for 270 (though Pennsylvania is probably the best answer for that, because our average is Clinton +5 in Pennsylvania while it’s +6 in Colorado, New Hampshire, and Virginia). Ohio, by contrast, is one of the “frosting on the cake” states along with Florida and North Carolina.
Where that NBC/Marist poll really mattered, though, was on the Senate side of the equation. Until today, our Senate forecast has been narrowly in favor of the Democrats; we projected out to a 50-50 Senate (which is Democratic control so long as Tim Kaine is vice president), with odds of Democratic control in the 50s. However, today sees those odds fall to 46 percent, and the projected Senate composition falls to 49-51 (which is where things started in the first few weeks where we were running the model in early August but hadn’t publicly unveiled it yet).
NBC/Marist’s New Hampshire poll found Republican incumbent Kelly Ayotte leading her Democratic challenger, Gov. Maggie Hassan, by an 8-point margin. That’s entirely inconsistent with, again, the six polls released in August in New Hampshire, with Hassan leads of 1, 2, 2, 3, 3, and 10. Or more accurately, it is consistent in terms of the usual disparity by which Ayotte overperforms Trump (or Hassan underperforms Clinton, if you prefer) … but then, you would have to assume that New Hampshire residents either shifted hard against both Clinton and Hassan at the same time in the last week, or else that Marist stumbled upon a particularly Republican-friendly slice of the electorate in their particular sample.
While I’d urge not getting too bent-out-shape about this or any other individual poll (the best thing to do, as always, is to throw it on the pile and average everything out), it's not hard to see this poll is very much out of step with everything else we've seen in New Hampshire. I don’t like to throw around the term “outlier” (which basically has come to mean “any poll I don’t like”), but even if you concede that the Ayotte/Hassan race is currently 50-50 in reality (and I don’t know if you should assume that, given that the August polls are better than that), a poll with a 2.9 percent margin of error (like this one) shouldn't give you results outside of a range of 53 Ayotte-47 Hassan to 53 Hassan-47 Ayotte. The danger, though, is that any poll that looks weird at first glance isn’t an outlier but rather the herald of a new trend (think back to that first PPP poll of the 2010 Massachusetts special election that showed Scott Brown in the lead; how crazy did that seem?), which is why we don’t discard or seek to correct polls that have that outlying whiff about them, but, again, throw them on the pile and let the model sort them out.
The other Sunday polls from Marist (generally a very good as well as prolific pollster, one that I don’t have any particular concerns about) didn’t help matters much, either, showing two potential sleeper races, in Arizona and Georgia to be pretty uncompetitive even while showing Clinton trailing only by a point or two in those states’ presidential contests. Those two Senate races are potential other routes to 50, but the Dem odds in both those races have now fallen into the single digits, which makes those unusual “lose Pennsylvania and New Hampshire, but win Illinois, Wisconsin, Indiana, and Arizona" permutations even rarer. Marist had Democratic Rep. Ann Kirkpatrick trailing GOP incumbent John McCain by 19 in Arizona, and Jim Barksdale trailing GOP incumbent Johnny Isakson by 15 in Georgia.
There was one other less-than-ideal Senate poll from late last week that’s worth a mention, even though it didn’t seem to affect the overall odds too much. It was a Howey Politics poll of Indiana, where Democratic ex-Sen. Evan Bayh leads Republican Rep. Todd Young by only 4 points in this open seat, after a series of internal polls that gave Bayh double-digit leads. The downward pressure of this one poll, however, only budged his odds down a smidge from 99 percent.
The presidential topline (Trump leading only by 7) wasn't such that it would call the whole poll into question. It’s not clear whether future polls will also see this as a single-digit race, but this reflects why we’ve been cautious about moving this race past “Tossup” in our qualitative ratings: Bayh is running in a red state where he hasn’t been elected since 2004, and has some residency-related optics problems. This poll could be a blip, or it could reflect political gravity starting to kick in despite Bayh’s initial name rec-fueled advantage.