Last Wednesday Daily Kos Elections debuted its 2016 forecast model, where we assess the likelihood of Hillary Clinton’s odds of retaining the presidency and the Democrats’ odds at flipping the Senate. These odds were very good last week, and they continue to be good today! We project that, currently, Clinton has an 88 percent likelihood of winning. We also project that the Democrats have a 61 percent chance of controlling the Senate after the election, which isn’t as much of an lazily contested jump shot (note that, despite the rosy projection, I’m still not describing it as a “slam dunk” or even a “lay-up”) as the presidential election, but should still fill you with some confidence.
This isn’t to say that if the election were held today, Clinton has an 88 percent chance of winning; if it were today, her odds would be considerably higher. Instead, that remaining uncertainty reflects there’s a possibility that, with two-and-a-half months left, there are still opportunities for things to go wrong, and moreover, that the fundamentals of the election (such as mediocre growth in the gross domestic product in the second quarter, a presidential approval that’s slightly above water, and the inherent difficulty in the incumbent party running for a third term), by themselves, point to a much closer race.
We’ll be discussing the model’s conclusions on Mondays and Thursdays each upcoming week; we might also shift to more times per week as we get closer to the election, depending on the volume of new polls and if things get more volatile. Today, we’ll talk a bit more about what changed in the presidential election since our data scientist, Drew Linzer, introduced the model last week, but we’ll also talk more extensively about the Senate races, which we didn’t say much about in our first post.
You might have noticed that last Wednesday, Hillary Clinton’s projected odds in the presidential race were 90 percent. Now they’re down to 88 percent. TIME TO PANIC!!!
OK, just kidding. It’s not time to panic at all. That difference isn’t particularly significant. It may simply be the variation that naturally occurs when you run thousands of Monte Carlo simulations each day using pretty much the same data; some days you might get a few more favorable bounces than other days, but if you have a large enough pool of permutations, the results will fall within the same narrow range.
You might have heard some vague rumblings that “well, Donald Trump is getting closer in the polls again, isn’t he?” That’s not affecting the overall odds in our model. For one thing, no, he isn’t actually getting closer; the Huffington Post Pollster national polling average still gives Clinton an 8-point aggregate lead (47-39), which is basically where the race has been hovering since the bounce from the Democratic convention took hold in early August. Several individual tracking polls have gotten closer, which may be what’s spurring that discussion; but, given how many polls there are in total, each individual tracking poll has only a very small effect on the overall aggregate.
And more importantly, we simply don’t give any weight to what’s happening in national polls. Our model doesn’t use them; we only use state polls, since the election is decided in the Electoral College. (That’s why we don’t average out the national polls, and I had to link to HuffPo for a chart about that.) Some other modelers do place a small amount of weight on the national polls, as something of a “reality check” for the state polls, but we do not. It’s worth considering that the state polls tend to be a lagging indicator compared with the national polls (since there are fewer of them, and they’re more sporadic, so aggregated state polls are likely to be slower to track changes in the election’s overall direction), but, frankly, we’re not in a race to try to show volatile changes that might or might not be happening.
And even when one state poll does show up that provides us with somewhat sub-optimal results, it too is unlikely to really dent the odds in a perceptible way. Take, for instance, the CBS/YouGov poll of Iowa that came out Sunday, which found a 40-40 tie between Clinton and Donald Trump in that state. For starters, that poll barely changed the Iowa aggregate, where the polls have already been close (our polling average of Iowa gives Clinton a 39-37 edge there); while there are many fewer Iowa polls than national polls, there are still enough Iowa polls that one more poll in the same basic range won’t cause a dramatic swing. (The story behind why Iowa seems to be underperforming the other swing states this year is a very interesting one, and we’ll probably talk more about that on Thursday.)
Most significantly, a win or loss in Iowa doesn’t tip the overall victory odds that much, either. While Iowa has tended to be closer to being a “tipping point” state in recent years, this year, it’s more or less frosting on the cake. The reality is that if Clinton wins Colorado, New Hampshire, Pennsylvania, and Virginia … all “swing” states where she’s sporting a near-double-digit aggregate lead and where her odds, individually, are in the 80-90 percent range ... she has already exceeded 270 electoral votes with or without Iowa (or, for that matter, without the much more vote-rich Florida or Ohio).
So, bearing in mind that it’s going to take some very major shifts in state polls, across a number of different states, to dislodge Clinton for her dominant position in our model, let’s shift gears and talk about Democratic odds in the Senate, where we’re well above water but our model isn’t feeling the same level of confidence.
The most basic facts, first: the Democrats hold 46 seats in the Senate, if you count independents Bernie Sanders and Angus King as caucusing with the Dems. To “win” the Senate, they’ll need to take that number up to 50, if there’s a Democratic vice president who’ll cast tie-breaking votes (or 51, if there’s a Republican vice-president). So, our model asks what the percentage odds of picking up those four (or more) seats is. Last Wednesday, those odds were at 61 percent, and today, they’re still at 61 percent.
The likeliest outcome in the Senate, in fact, is that 50-50 split that we mentioned. If we look at it race-by-race, the likeliest way that happens is to gain five seats (defeating Republican incumbents in Illinois, New Hampshire, Pennsylvania, and Wisconsin, and picking up the open seat in Indiana), but then lose one seat (the open seat in Nevada). We don’t have a Senate page currently running that we can link for you (we’ll have one very soon), but for now, here’s a table showing the anticipated outcomes in the key races.
Colorado: 99 percent Democratic odds; 52-37 polling aggregate
Indiana: 99 percent Democratic odds; 53-37 polling aggregate
Wisconsin: 95 percent Democratic odds; 49-40 polling aggregate
Illinois: 65 percent Democratic odds; 42-40 polling aggregate
New Hampshire: 65 percent Democratic odds; 44-43 polling aggregate
Pennsylvania: 62 percent Democratic odds; 43-42 polling aggregate
Florida: 35 percent Democratic odds; 42-45 polling aggregate
Nevada: 35 percent Democratic odds; 39-41 polling aggregate
North Carolina: 33 percent Democratic odds; 39-42 polling aggregate
Arizona: 20 percent Democratic odds; 38-43 polling aggregate
Missouri: 20 percent Democratic odds; 39-45 polling aggregate
Ohio: 18 percent Democratic odds; 39-45 polling aggregate
Georgia: 17 percent Democratic odds; 38-45 polling aggregate
Iowa: 5 percent Democratic odds; 41-50 polling aggregate
Despite being the reddest state to host a competitive Senate race, Indiana actually looks like the most probable Democratic pickup for now, thanks to ex-Senator Evan Bayh’s high name recognition. Bayh sported a 7-point lead over Republican Rep. Todd Young in the only public poll of the race, from Monmouth, but also led by much more in a Democratic internal poll of the race; even after downweighting that poll because of its status as a partisan poll, it still helps Bayh to open up an overwhelming lead. That lead, however, may become less overwhelming as time goes by and more polls are released, as Young’s advertising helps introduce him more to the other 8/9ths of the state that he doesn’t represent (and the NRSC tries to take Bayh down a few pegs).
Although Maggie Hassan’s lead in New Hampshire and Katie McGinty’s lead in Pennsylvania are both quite small, those states are polled consistently enough, and those small leads are replicated consistently in most polls for our model to feel that they have significantly better than 50-50 odds. Tammy Duckworth in Illinois has, on paper, a much easier race (thanks to a much bluer state); however, our model isn’t that much more confident about her race, in large part because we have so few polls to go by. (Unlike the other competitive Senate races, most of which are in swing states or reddish states that could swing with a big Clinton win, nobody’s polling Illinois. What few polls we’re seeing in the Illinois race are mostly leaked internals.)
Finally, there’s the puzzling situation in Nevada, where polls have tended to give Republican Rep. Joe Heck either a tie with Democratic Attorney General Catherine Cortez Masto, or give Heck a small lead. It’s certainly possible that the Democrats are genuinely underperforming in Nevada this year (there are possible demographic reasons for thinking that’s the case, which, again, we’ll discuss on Thursday), and that could be rubbing off on the Senate as well as the presidential race.
On the other hand, it’s also possible that pollsters’ persistent problems in estimating the Democratic vote in Nevada are still happening; you may remember that polling consistently found Harry Reid losing his 2010 Senate race (hint: he won), and also underestimated Barack Obama’s margin in Nevada in the 2012 presidential race. Part of the problem is that there are a lot fewer polls in Nevada this year than in the other swing states, and the most frequent pollster there seems to be Rasmussen Reports, which may have turned the page with new ownership but still tend to put up very Rasmussen-like results this year.
The good news for the Democrats is that there are various other paths in the Senate that aren’t closed off, which could provide insurance if they lose one of the New Hampshire or Pennsylvania duo, or give them the chance to expand their total into the low 50s (which is important as a cushion for trying to maintain control in 2018, when they’ll be defending a number of red-state seats against the backdrop of Hillary Clinton’s probable first midterm). Florida and North Carolina, for instance, are both within striking distance, with other lower-tier opportunities present in Arizona, Missouri, and Ohio.
As more Senate polls continue to roll in, we’ll keep discussing the Senate outlook, as well as the presidential race. (At this rate, the race for the Senate promises to be more suspenseful.) While there are very few gubernatorial seats up in a presidential year, we’ll also offer more discussion about those races in the coming weeks. So please keep tuning into our model write-ups every Monday and Thursday!
The real drama in October may, in fact, turn out to be the Dems’ efforts to take back the House … something that I’d characterize as a long shot (we’d need to flip 30 seats) but more plausible than it was even a few months ago. Unfortunately, we aren’t able to model House races; to put together an effective model, you need polls, and most House races (even the competitive ones) simply don’t see polls that are released to the public. There are models for converting “generic ballot” polls into a possible range of seats that flip, but if you’re really interested in delving into the House, Daily Kos Elections continues to offer qualitative ratings of not just Senate and gubernatorial races, but all competitive House races as well, and we’d urge you to check those out as well (which you can access through our new landing page).