Remember a few weeks ago when people were all “OH NO DONALD TRUMP IS GOING TO WIN HOW IS THIS HAPPENING TO US AAAAAAGGGGHGH?” No, really, it was literally just a few weeks ago! The peak of that sentiment was right around September 21, which not coincidentally was when the Daily Kos Elections model hit its low-water mark, with Hillary Clinton having “only” a 63 percent chance of winning in November. That was when the polls taken with field periods right after Pneumonia-ghazi on Sept. 11, with the subsequently demoralized Democratic respondents, were at their fullest effect.
(Strangely enough, both September 2008 and September 2012 also had a couple weeks of dark journeys into the Democrats’ souls, after the Republican convention bounce in ‘08 and after the first debate in ‘12. But when we were all crying “OH NO JOHN MCCAIN/MITT ROMNEY IS GOING TO WIN AAAAAAGHGH,” McCain/Romney actually took the polling leads during those brief periods, unlike this time! Obviously, there’s no political law of gravity that says the pattern has to happen that way, but I suspect we’ll still be surprised when it happens again in September of 2020.)
Well, that sentiment seems to have vaporized. That’s what a successful first debate, the leak of an opponent’s video where he brags about sexual assault, a successful second debate, and now multiple harassment victims coming forward, will tend to do for you. With that in mind, it looks like the focus in the remaining few weeks of the campaign will be moving toward control of the Senate and the House.
As much as the Republican Party is currently lying upside down in a ditch with flames pouring out of it, we’re still at a point where they could retain control of the Senate. Our model estimates the Democrats have a 54 percent chance of ending with 50 seats or more (including independent caucus members Angus King and Bernie Sanders), which is hardly the slam dunk that people have started assuming. The good news, however, is that Senate polling tends to lag presidential polling a bit. State-level polling, in general, is less quick to pick up on emerging trends than national polling, and, in addition, it seems to take a while for trends in the presidential race to turn into coattails (or the lack thereof) for races down the ballot. Whether it’s a good year (like 2006) or a bad year (like 2010), waves tend to materialize very late in the year. You definitely have a sense of which way the wind is blowing by October, but it’s hard to gauge the size of the wave until you’re right next to it, looking up at it.
For now, the main problem seems to be that Republican Senate candidates are still able to show some separation from Donald Trump. Perhaps the most startling example was a poll of Ohio that came out from Baldwin Wallace University on Wednesday. A 22-point disparity between a Democratic presidential candidate and a Senate candidate is unheard of, but that’s what we got. The poll found Hillary Clinton winning the 2-way presidential race in Ohio by a 10-point margin (the best result we’ve gotten all year in a Ohio), but Democratic Senate candidate Ted Strickland losing by 12. (Needless to say, our model gives GOP incumbent Rob Portman a 99 percent chance of winning that race, no different from the Republican odds in Oklahoma and Utah.)
That’s happening in New Hampshire, where Kelly Ayotte maintains a small lead against Maggie Hassan despite Trump’s struggles there. It may even be happening in Wisconsin; case in point, the Marquette University poll that came out on Wednesday found Clinton increasing her lead in Wisconsin to 7 points (up from 3 in their previous poll), but the same sample had Russ Feingold leading Republican incumbent Ron Johnson by only 3 points. While our poll average still gives Feingold a 6-point lead, his odds fell noticeably thanks to this poll, down to 89 percent now after nearly maxing out at 99 percent recently. That downtick in his odds is probably what’s most responsible for the overall Senate odds falling a bit to 54 percent.
Nevertheless, our model still sees 50 seats as the “median” result in all of our simulations. That, as I’ve discussed before, means that even though the seats where the Democratic candidate has over 50 percent odds only add up to 49 total, there are enough individual races where the Democratic odds are in the 30 or 40 percent range (Missouri, New Hampshire, Nevada, North Carolina) that the Dems manage to win at least one of them (in addition to Illinois, Indiana, Pennsylvania, and Wisconsin) in a majority of our simulations in order to get to 50 or better.
Okay, you’re now hopefully saying, “even if we flip the Senate, that’s only half the battle; taking the Senate restores our appointment power for judiciary and executive branch positions, but in order to actually pass new legislation, we’ve got to flip the House too. Are we going to be able to do that?” As much as I’d like to see that, the odds are steeply against that. Note, however, that I’m not assigning a specific number to that, either in terms of a percentage likelihood of flipping the chamber, or a likely number of seats that flip.
That’s because there really isn’t a satisfactory way to model the House, at least not race-by-race the way we do for the Senate (or even for the presidency, using the Electoral College). You might think that it’s not that complicated because you only need to model a few dozen races; most House races are in districts that are too blue or too red to ever flip. However, even among the competitive races, many never get polled at all, and even if they do, they often get publicly polled only once or twice, usually not from an independent pollster, but rather from a partisan pollster affiliated with one of the campaigns (who tend to leak results only if they represent a best- or worst-case scenario, in order to signal to donors that they need to pile on to a particular race).
And you can’t just apply uniform swing to congressional districts; as you’ve probably noticed, entrenched incumbents, by virtue of a moderate record, charisma, good constituent service, or simple entropy, are often able to hold on in districts that lean toward the opposite party at the presidential level. The quality of the opposing candidate and the amount of money they’re able to raise also makes a huge difference.
In other words, if you just thought one-dimensionally about uniform swing, you’d probably assume that NJ-02, a district that went 54 percent for Obama in 2012, should flip blue amidst even the slightest breeze, while MI-01, a district that gave Obama 45 percent in 2012, is unassailable. However, the precise opposite is true; Michigan’s 1st is one of the Dems’ top pickup opportunities, thanks to the combination of a strong, well-funded Democratic candidate (Lon Johnson, a former state party chair in Michigan), an open seat being vacated by an unpopular Republican who won only narrowly in previous elections, and a historic Democratic tradition, especially in downballot races. Meanwhile, New Jersey’s 2nd has Frank LoBiondo, an uncontroversial and well-liked moderate Republican who, by virtue of his popularity, draws underfunded nobodies as opponents who can’t even win in wave years.
The Daily Kos Elections model does have a House page, which I’d urge you to take a look at. It has a cartogram map (shown at the top of the article) showing which races are competitive. (The gray races are the Tossups; races leaning one way or the other are red or blue. The vast majority of races, however, aren’t competitive at all, and they’re the ones that aren’t filled in with any color.) It also has a list of the competitive races, and whether they’re delineated as Tossups, Lean Dem, and so on, using our qualitative rating system.
For that, we still do ratings the old-fashioned way, like old-school prognosticators like Charlie Cook and Stu Rothenberg still do … looking at what few polls we see, but filling that in with how much money the candidates have, what they’re doing with it, and maybe most importantly, which races are seeing the investments from outside groups like the DCCC and House Majority PAC (who do poll the races privately, and make advertising dollar decisions based on that). We move a few races to a different column each week, trying to stay on top of the changing battlefield, which is maybe the most you can do when you don’t have the same volume of polling data that you need in order to model the House the way we do with the presidential and Senate races.
Unfortunately, that still doesn’t give you any level of numeric probability. A handy rule of thumb is that the parties tend to split the races that are in the Tossup column, while usually winning most if not all of the races that lean their way. Of the 18 races that are currently filed as Tossups, only 4 of those are Democrat-held seats (open seats in AZ-01 and FL-18, plus endangered incumbents in MN-08 and NE-02). So the Democrats are on their way to picking up a dozen or so seats out of the Tossup column … but obviously that’s not enough to flip the House, which requires 30 seats (or 29, really, since they’re almost certain to pick up FL-10 and VA-04 but lose FL-02, thanks to the un-gerrymandering of those states). To do so, they’ll either need to mount a sneak attack that lets them pick up all the Lean Republican races too … or else for all the races to shift one column to the left in the coming weeks, if things visibly start falling apart for the Republicans.
There is, however, one weird trick for making a quantitative assessment of how big the swing in the House will be, and that’s to use generic ballot polling … in other words, national presidential polls that also ask something like “do you want a Democrat or a Republican to win your particular House race this year?” Currently, Generic D is leading Generic R by a 46 to 41 margin, according to Huffington Post Pollster’s aggregate of all generic House ballot polls. (We don’t keep track of that particular number for the Daily Kos Elections model.)
That, in itself, doesn’t tell you much. However, political scientists can, by observing patterns from previous election cycles, gauge how many seats are likely to swing based on generic House ballots and other factors (like how many seats are up for grabs and whether it’s a midterm or presidential year). One of the most prominent modelers, in fact, is Alan Abramowitz, who’s also the creator of the “Time for a Change” model that fills in the “fundamentals” aspect of our presidential model.
Abramowitz’s model for 2016 projects that with a 5-point Democratic lead on the generic ballot, that would pencil out to a 15 or 16 seat Democratic gain … a nice increase, but only half of what we need to take control. (Unless you can envision a farfetched scenario where the Freedom Caucus has enough dissenters to prevent Paul Ryan from becoming speaker and after repeated deadlocked votes, the most moderate dozen or so Republicans throw in with the Democrats to elect Nancy Pelosi … or more likely, a moderate Republican as a compromise … as speaker.) Scroll down to table 2 of the link to see the relevant chart.
Instead, in order to truly gain 30 seats, the Democrats would need to be sporting a lead of around 13 on the generic House ballot, according to Abramowitz’s model. That’s a truly heavy lift, though, if Trump’s death spiral not only continues to get worse but starts to rub off on Congress members across the boards, it’s at least mathematically possible. Keep watching our future model posts on Mondays and Thursdays, though, and I’ll try to keep you updated about what’s happening with both the generic House ballot numbers and with our own in-house qualitative ratings, as the focus keeps shifting from the presidential race to the fight for both chambers in Congress.
Finally, speaking of that presidential race … Hillary Clinton’s odds of victory have climbed to 96 percent as of Thursday morning, the highest they’ve been, leaving a “black swan” event really the only thing standing in her way at this point. Her odds in key states include 93 percent in Colorado, 98 percent in Pennsylvania, and even 84 percent in Florida. Ohio remains the only state in Tossup territory right now, with Clinton holding 52 percent odds of winning. The main questions at this point are whether she can salvage Iowa and/or run up the score with additional states like Arizona or Georgia … but how many Democrats are in Congress come next January, rather than how many states she colors blue on the presidential map, is what will really matter when it comes to determining how far she can go with progressive legislation in the coming years.