What a difference a week makes! When we looked at the model on Monday, September 26, (the day of the first presidential debate), Hillary Clinton’s odds of winning were 64 percent. There had been some subtle improvements in the previous week in Clinton’s national polling numbers (as Pneumonia-ghazi started to fade from view) but that hadn’t really trickled down into the state-level polls, which is what our model is based on. By Thursday, September 29, that improvement was starting to filter into the state polls, and our model ticked up to 68 percent … but that was still based only on polls with a pre-debate field period.
On the morning Monday, October 3, we had a few post-debate polls under our belt … and Clinton’s overall odds were up to 72 percent … but we were still left wondering why everything was so quiet on the polling front. By the end of Monday, though, the deluge had arrived, and with one exception (Quinnipiac’s Ohio poll), everything was very good news for Clinton: among others, a Clinton +11 poll from Monmouth in Colorado, another Clinton +11 poll in Colorado from Keating Research, polls from Quinnipiac with Clinton +5 in Florida and +3 in North Carolina, a Clinton +9 poll in Pennsylvania from Franklin & Marshall, a Clinton +3 poll in Nevada from Hart Research, and a Clinton +7 poll in Virginia from Christopher Newport Univ.
It may well have been her single best polling day of the cycle, and by Tuesday her odds had jumped to 82 percent, a one-day gain of 10. That matches the largest single-day gain our model has seen since we started running. That other gain of 10 happened between August 8 and 9; in case you’re wondering what was happening then, that was the Monday after the Democratic convention ended, when the post-DNC polls started to show up. So you could say that the debate was one of the most momentous events of the campaign: if your metric is the effect it had on our model, she got a convention-sized bounce out of it.
The subsequent days have seen even more strong poll results, most notably two different polls on Wednesday (from Monmouth and Anzalone Liszt) giving Clinton a 2-point lead in Ohio, which isn’t a lot but serves to counteract the Quinnipiac poll that had her down 5. The subsequent polls weren’t enough to really move the model much higher; it currently places Clinton’s odds at 83 percent. But they do continue an impressive little winning streak: out of the several dozen polls of swing states released since the debate, only one of that entire stack had Donald Trump leading (again, that Quinnipiac Ohio poll). And that stack covers every major swing state except Iowa (and Wisconsin, if you even consider that a swing state in the first place).
If you look under the hood of the model at what the various moving parts are doing, you’ll see that the biggest impact of the new polls is on the state-level odds in Colorado, Florida, Nevada, and North Carolina. Just since the weekend, Clinton’s odds in Colorado shot up from the mid-70s to 88 percent (thanks to those two 11-point leads on Monday). That’s actually enough to push Colorado … which previously had been the most troublesome of the “firewall” states … up above Pennsylvania on the totem pole, leaving Pennsylvania as the likeliest “pivot point” state, i.e. the one that puts Clinton over the top with her 270th electoral vote. Either way, the odds of the pivot point state tend to have the biggest impact on the overall odds, and if the odds in the pivot point state are in the 80s (Pennsylvania is currently at 87), then the overall odds are likely to be in the 80s as well.
Similarly, Clinton’s odds are in Florida are up to 66 percent, up from around 50-50 over the weekend. Her odds in Nevada are 49 percent, up from around 30 percent over the weekend, and her odds in North Carolina are now exactly 50-50, also up from around 30 percent previously. That all has moved Florida from gray to light blue on our map, and Nevada and North Carolina from light red to gray. (There’s also some subtler stuff going on, like Maine, Michigan, and New Mexico all moving above 95 percent odds, changing them to darkest blue on the map.)
There are a couple different processes going on here with these rapid gains. One, the tables have likely turned, on the issue of “differential response.” The nervous Democrats who were in hiding after Pneumonia-ghazi had, post-debate, resumed feeling energized and were more likely to respond to a pollster’s inquiries; instead, it was the Trump supporters’ turn to hide under the blankets. (It’s also quite possible that the debate genuinely changed some Trump supporters’ minds, if not switching them to outright Clinton support, then at least bumping them to undecided or not-voting. Trump’s debate performance may truly have been that bad. However, there’s really not that much mind-changing at this stage of the process. Week-to-week poll fluctuations, most of the time, have more to do with changes in the patterns of who responds to polls and who says they’re likely to vote.)
And two, the third-party vote is starting to shrink. That’s not specific to this election cycle, or a direct consequence of the debate; this is always something that happens in the polls as we get closer to the election. (What may be different this year is that we’re starting from a position of higher-than-usual support for third parties to begin with.) Traditionally, when the actual election date approaches, people tend to drop their idealistic impulses and fall in line with one of the candidates who has a shot at winning. In addition, at this point this year, Gary Johnson may be finally getting more scrutiny from the media, which may be leading people to realize that he doesn’t actually seem to know anything about world affairs.
Drew Linzer compiled a graph (see below) showing this decline over the month of September in national polls (meaning the effect is now starting to trickle down into state polls as well). Johnson’s support starts from a higher baseline than third-party candidates had in 2008 or 2012, but it’s taking the same downward trajectory.
Another way of looking at it is to check out the national poll aggregate at Huffington Post Pollster. For much of the race, when one candidate or the other’s bounce would wear off, and the race reverted to the mean, it’d be something like 45-40. Now, the race is up to 48-42. You can see how undecided voters and third-party voters are coming off the fence, and backing one of the two plausible candidates.
Will Clinton’s numbers continue to go up? Well, we can’t really make any promises, because of there’s still a degree of uncertainty about what happens in the coming month. Perhaps Donald Trump will be the very model of wisdom and civility at the next debate, resetting the race yet again. (Perhaps monkeys will also fly out of my butt.)
What I can tell you is that, if there aren’t any external events that shake up the race and the polls continue in the exact same place they’re in now for the coming weeks, then, yes, her odds will continue to go up slowly anyway, because with each passing day, there’s less uncertainty. We use economic and historic “fundamentals” in our model to substitute for that uncertainty, and each passing day means the model will be less fundamentals (which don’t especially favor Clinton) and more polls (which, currently, very much do favor Clinton).
Finally, let’s look at the Senate side of the model. The Democrats’ odds of hitting 50 seats is currently at 60 percent, after a bit more upward movement in the last few days. This isn’t the absolutely best result we’ve seen from the model (it hit 62 percent in mid-August during peak convention bounce), but it’s close to that. Currently, the model shows the Democrats with four seats likely flipping (Illinois, Indiana, Pennsylvania, and Wisconsin).
However, they need one more seat to seal the deal. There aren’t any more seats where the Democratic candidate has better than 50 percent odds of winning, but there are at least four possibilities where the Democratic candidate is still within striking distance: New Hampshire (where the odds are 48 percent), North Carolina (46 percent), Missouri (33 percent), and Nevada (24 percent). There are increasingly loud anecdotal reports that, in particular, Jason Kander in Missouri may well have much better odds than that … but until someone publicly coughs up one of the rumored internal polls that are rumbling around, our model can’t really do anything to account for that.
There’s also some even better news on the Senate front … just as we unveiled individual pages for each state in the presidential race last week, now we’ve gone live with individual pages for each state’s Senate race! Here’s the North Carolina page as an example; we show you not just the current odds there, but also the polling aggregate (currently 42.3 for GOP incumbent Richard Burr, 41.8 for Deborah Ross, which is why the odds are so close here) and the list of the most recent polls of the Tar Heel State. To get to a particular state, either click on that state under the overall Senate odds on our landing page (if it’s a competitive election), or else click on the “Senate” page from the bar at the top of our landing page, which will take you to the full list of all states with Senate elections this year.