It’s Election Day, and that brings the Daily Kos Elections predictive model to a close. We’re going out on a pretty high note! As our chief data guru Drew Linzer wrote on Monday, our final projection is for 323 electoral votes for Hillary Clinton and 215 for Donald Trump. And our model is very confident in that figure; our model doesn’t even have any states in the “Tossup” category at this point. The most competitive state on the red side of the ledger is Ohio, where our model says Clinton has only a 13 percent chance of winning (based on a 45-43 aggregate polling deficit), and the most competitive state on the blue side is Nevada, where we say Clinton has a 71 percent chance (thanks to a 45-44 polling lead). In the real world, her odds are probably much higher than that in Nevada, given Democratic performance in the early vote there … but our model can’t account for early vote details.
Currently, our model gives Clinton 92 percent odds of victory; in other words, it’s not mathematically impossible for Trump to win, but it would require a remarkable run of luck. He’d need to flip not only Iowa and Ohio, but also Florida (88 percent Clinton odds, based on a 46-45 aggregate lead), North Carolina (75 percent Clinton odds, based on a 46-45 aggregate), Nevada (where he’s probably already lost), and finally, one of the many states where Clinton has 99 percent odds and aggregate polling leads of 4 points or more (Colorado? Michigan? New Hampshire? Pennsylvania? doesn’t matter; they’re all the same dead end).
So, unless we’re looking down the barrel of a categorical, industry-wide polling error on the magnitude of Truman vs. Dewey, Hillary Clinton will be the next president. (And keep in mind that in 1948, there were only a few pollsters operating, using terrible methods like quota sampling, and they all stopped polling weeks before the election.) The real question at this point, instead, is: what kind of legislative allies can Clinton count on being elected today, to enact her agenda?
Unfortunately, the odds are near zero of flipping the House right now. The Daily Kos Elections model doesn’t offer specific House predictions (there aren’t enough polls of individual races to do that), but the Huffington Post Pollster average of generic House ballot polling gives the Democrats a 3-point edge on that question, which, according to Alan Abramowitz’s model for converting generic ballot into pickup size, is good for a gain of 12 or 13 House seats. That’s short of the 30 they’ll need to take control, though it’s enough to erase the GOP gains from the 2014 election.
Even without the House, though, the Senate is still critically important, because Senate control would allow Clinton’s SCOTUS appointments (as well as cabinet and lower-level judiciary) appointments to be approved without obstruction. So where does the model leave us standing on the last day, regarding the Senate?
The Democrats still have better-than-even odds of taking control back in the Senate; our model puts the odds of ending up with 50 seats or more (which would give them control if Tim Kaine is vice-president) at 66 percent. That’s down a few ticks from a peak of 72 percent last week, but, if nothing else, one would rather be in the Dems’ position than not.
The median projection out of all our simulation runs is for the Democrats to end up with 50 seats. However, if you look at the states where our odds are 50 percent or greater and total them up, you actually wind up with 51 Democratic seats instead, for a net gain of five! The explanation for that disparity is that once you get past the races where we’re in reasonably good shape (pickups in Illinois, Pennsylvania, and Wisconsin, plus a hold in Nevada), two of those seats have only a very narrow Dem advantage (Indiana and New Hampshire), so there are a lot of simulation runs where we lose one or the other of those two seats, leaving 50 as the median.
Even though the bottom line in the Senate has barely changed over the weeks, the rank order among those seats has gotten very scrambled. You probably remember that for most of the last two months, I framed the Senate problem as there being three tiers of three seats each: likely pickups (greater than 90 percent odds) in Illinois, Indiana, and Wisconsin, pure tossups in Nevada, New Hampshire, and Pennsylvania, and long-shots (less than 45 percent odds, but more than 10 percent) in Florida, Missouri, and North Carolina.
However, today, the “likely pickup” tier has broken apart. Illinois is now a pickup with 99 percent certainty (ahead of even Colorado at 97 percent, where Michael Bennet’s underfunded challenger has occasionally polled within high single digits), but Indiana and Wisconsin have suddenly become close races. In Wisconsin, we’ve only ever seen one poll with Russ Feingold trailing, but the aggregate has dwindled to a low-single-digits race as the GOP has poured in a lot of late money to salvage Ron Johnson; we currently give Feingold 62 percent odds, based on a 47-45 aggregate lead. And in Indiana, the trendline is much worse; Evan Bayh started out with a large lead but (mostly thanks to a lot of recently-reported unflattering stories about his transition from the Senate to the private sector) has dwindled to only 51 percent odds and a 44-43 aggregate lead, with the undecideds possibly stacked against him in this red state.
In the middle bucket, Nevada and Pennsylvania have broken easily in the Dems’ direction, to the extent that they’re actually in much better shape now than Indiana and Wisconsin. The odds for Catherine Cortez Masto (running to replace the retiring Harry Reid) in Nevada are now 71 percent, based on a 45-43 aggregate lead … and in the real world, her chances are probably better than that, given crushing Dem turnout in early voting. And in Pennsylvania, Katie McGinty’s odds are 79 percent, based on a 46-43 aggregate lead. New Hampshire’s Senate race, however, has bounced all around thanks to that state’s volatile local pollsters and is still stuck in tossup territory; currently, Maggie Hassan has a narrow 46-45 polling lead, good for 55 percent odds.
And in the last bucket, Missouri and especially North Carolina have edged closer to tossups while Florida has languished. Thanks to the last wave of Survey Monkey state polling which put both of them in the lead, Jason Kander (in Missouri) and Deborah Ross (in North Carolina) gained ground but not enough to get them up to 50-50. Kander’s odds are 39 percent (based on a 45-44 deficit) and Ross’s odds are 47 percent (also based on a 45-44 deficit), good enough to bump North Carolina back up to a greyish hue on our map. While a number of polls have seen Patrick Murphy closing the gap against Marco Rubio in Florida, he’s still trailing by an average 47-43, which is only good for 17 percent odds.
So, while the overall Senate odds and the expected number of seats (50 or 51) really haven’t changed, the names and faces that we’ll see for the next six years definitely have changed. Most notably, Evan Bayh’s once-sure-thing stock has plunged, while Catherine Cortez Masto, who looked DOA not that long ago, now looks very likely to fill Harry Reid’s shoes.
Finally, let’s look at this year’s gubernatorial races, which always seem to get short shrift (even from me). Part of that is because we don’t have a separate governors’ races page, the way we do for the presidential and Senate races—which in turn is due to there being only a dozen gubernatorial races this year; most of the big state races fall in midterm years.
The median number of Dem-held gubernatorial seats after the election has bobbled back and forth between 19 (which would be a net gain of one) and 20 (a net gain of two). For the last week, it’s been more settled on 19 as a median. And, in fact, if you switch over to the method of seeing which side has better-than-50 percent odds in each states and totaling those up, the result this week has been only 18 seats, meaning no net gain at all.
The problem here is that we had a cluster of dodgy polls last week in New Hampshire that, in addition to showing a near-tied presidential race, also showed Republican Chris Sununu pulling back ahead of Dem Colin Van Ostern. That couple days of terrible New Hampshire polling (from New Hampshire’s stable of terrible pollsters), however, was bracketed both before and after with better-looking polls, and so the aggregate in the Granite State is currently a dead even 43.3 to 43.3. However, the model’s simulation runs gave a faint edge to Sununu, with him winning 53 percent of the time (which may simply be random noise given the exact polling tie; another day’s run might give a small edge to Van Ostern). That’s on top of a much likelier Republican pickup in Vermont, where Democrat Sue Minter trails Republican Phil Scott by a 43-40 margin, giving her only 11 percent odds.
On the plus side, even if the Dems lose those two seats, they’re likely to pick up two prizes in much bigger and redder states: Indiana (where John Gregg now has a nearly-insurmountable 45-33 polling lead for the open seat left behind by Mike Pence, and 99 percent odds) and North Carolina (where Roy Cooper hasn’t quite put away incumbent Pat McCrory, though he still has a 48-44 polling lead and 78 percent odds). On top of that, the Democratic candidates are ahead in three of their tricky red-state retentions: Missouri, Montana, and West Virginia. The least secure of those three is Missouri, where Chris Koster has a 46-42 lead over Eric Greitens and has an 85 percent chance of winning.
And with that, the Daily Kos Elections 2016 model is going into the books. Luckily, we’re closing on a much rosier note than we did in our inaugural year of 2014. On the plus side, in 2014 we wound up having the most accurate model of any major forecaster, on both the Senate and gubernatorial sides. I would’ve gladly traded that, at the time, for being hilariously wrong but being able to maintain control of the Senate. Hopefully, though, in 2016 we can be both accurate and victorious.
