The polls are coming in fast and furious — and there are signs of late momentum for Hillary Clinton, whose lead has increased to roughly 3.5 percentage points over Donald Trump. Her chances of winning the Electoral College have ticked up to 69 percent in both the polls-only and polls-plusmodels, recovering some of the ground she lost after FBI director James Comey’s letter to Congress on Oct. 28. We’ll continue to update our forecast through early Tuesday morning.
But even if Clinton’s win probability inches up by another percentage point or two, she’ll still be the probable but far-from-certain winner. That means tomorrow is going to be very exciting — not only because the result is uncertain but because an unusually large number of states will potentially have a say in the outcome. Consider, for instance, that party ideologies and demographic coalitions are changing to the point that Clinton is an overall favorite in the race despite being several points behind in Iowa and even a modest underdog in Ohio, and also consider that Trump still has a shot despite sometimes having trailed in polls in red states ranging from Georgiato Utah.
I think of the swing states as belonging to eight groups of two to four states each. These groups correspond pretty closely to the states’ relative order of importance according to our tipping-point index, which reflects each state’s chance of casting the decisive 270th vote in the Electoral College. But these groups also reflect a sort of flowchart by which the winner of the Electoral College might be revealed. Depending on what happens in the first group of states, we could be all but certain of a Clinton victory — rendering the groups below it less relevant. Or an increasing number of contingencies could come into play where the race comes down to Omaha, Nebraska. But let’s start from the top (all the figures below come from our polls-only forecast unless otherwise noted):
Group 1: The trenches
- Florida (Clinton is a 51 percent “favorite”; 17 percent chance of tipping election)
- North Carolina (Clinton is a 51 percent “favorite”; 11 percent chance of tipping election)
I think of Florida and North Carolina as being the protagonists of this election. They have a bit of everything: early voting, conflicting polling, changing demographics. And they’ve always played a role in the drama of the campaign, since neither candidate has ever really been able to pull away in either state. (No candidate has ever been better than a 2-1 favorite in North Carolina in our polls-plus forecast, for instance.) Furthermore, both states’ polls close relatively early and they count their vote relatively quickly, so they’ll be some of the first states we’ll be checking for clues as to how tomorrow will proceed.
Clinton would seem to be playing offense in these states, in the sense that winning either one would make it very hard for Trump to come back — that’s especially the case for Florida, with its 29 electoral votes, which could cover for her losing two key states elsewhere on the map. But there’s a higher chance than you might think that they prove to be a necessity rather than a luxury for Clinton, in case something goes wrong for Clinton in the Midwest.
Group 2: Clinton’s outer firewall
- Pennsylvania (Clinton is a 76 percent favorite; 11 percent chance of tipping election)
- Michigan (Clinton is a 79 percent favorite, 11 percent chance of tipping election)
Throughout the Rust Belt and the Upper Midwest, Clinton is running 3 or 4 percentage points worse than President Obama did in our final 2012 forecast, which is a big reason why Trump has a better chance of winningthan Mitt Romney did. And the scenario by which Clinton loses a state like Michigan or Pennsylvania isn’t hard to fathom: First, enough white voters without college degrees turn out for Trump, and second, African-American turnout is depressed as compared with four years ago (which isn’t crazy to think could happen). Think of Clinton as flipping a coin twice, once to determine whether Trump turns out his base and once to determine whether she turns out her own base. If neither candidate brings his or her base to the polls or both do, Clinton wins Michigan and Pennsylvania by virtue of their being slightly blue-leaning states. But if the coin comes up tails both times, and Clinton doesn’t turn out her voters while Trump turns out his, she could lose. That corresponds to her roughly 25 percent chance of losing each state in our forecast.
It’s also the case that Trump has almost never been ahead in high-quality polls of Pennsylvania or Michigan. But Clinton’s leads have narrowed, and Pennsylvania is one of the few swing states without early voting, potentially allowing a campaign to make up ground late. Michigan, meanwhile, has an unusually large number of undecideds and a history of bad polling.
Group 3: The pivot points
- Nevada (Clinton is a 54 percent favorite; 4 percent chance of tipping election)
- New Hampshire (Clinton is a 66 percent favorite; 3 percent chance of tipping election)
Speaking of bad polling, no one’s quite sure what’s going on in these states. In Nevada, there’s a conflict between polls that continue to show a tossup and early voting data that seems to portend favorable things for Clinton. In New Hampshire, there’s a conflict among the polls themselves, with recent surveys showing everything from a 5-point lead for Trump to an 11-point lead for Clinton. Both states are also a bit hard to place demographically in the context of a Trump-Clinton election. New Hampshire has a reputation for being working-class, but it actually has fairly high incomes. In Nevada, the Latino and Asian-American populations are growing, but there are also lots of white voters without college degrees. Still, Democrats outnumber Republicans there, and if I had to bet against the polls anywhere, it would be on them having lowballed Clinton’s numbers in Nevada.
These states are also linked for another reason: They’re either going to be super important or totally irrelevant, without a lot of room in between. If Trump loses Florida or North Carolina, Nevada and New Hampshire’s small electoral vote totals (six and four, respectively) won’t be enough to compensate. The same idea applies for Clinton if Trump wins Pennsylvania or Michigan. But if Trump holds the Group 1 states and Clinton holds Group 2 along with the rest of her firewall, Nevada or New Hampshire (though not both) will be critical to produce a winning map.
Group 4: Clinton’s inner firewall
- Colorado (Clinton is a 75 percent favorite; 7 percent chance of tipping election)
- Virginia (Clinton is an 83 percent favorite; 7 percent chance of tipping election)
- Wisconsin (Clinton is an 82 percent favorite; 4 percent chance of tipping election)
These are secondary targets for Trump. Colorado was the tipping-point state in both 2008 and 2012, and Clinton’s campaign can perhaps be accused of having taken it for granted. Still, its highly-educated demographics are not a good match for Trump. The same goes for Virginia, but the race has tightened there more than most other states — Clinton’s lead has fallen from 10 points to 5 points in two weeks — possibly as parts of the Republican base have come home to Trump. Meanwhile, Trump did poorly in Wisconsin in the Republican primary, and Obama won it easily four years ago despite the presence of House Speaker Paul Ryan, who represents Wisconsin’s 1st Congressional District, on the Republican ticket. Still, our model says not to be totally dismissive of these states, where Clinton’s lead is only a point or so larger than in the Group 2 states.
Group 5: The old guard
- Ohio (Trump is a 65 percent favorite; 7 percent chance of tipping election)
- Iowa (Trump is a 73 percent favorite; 1 percent chance of tipping election)
Obama was leading in the polls in Ohio and Iowa four years ago. Clinton isn’t, which is a big reason her Electoral College map is less robust than his was. While the changes are easy enough to understand based on demographic patterns — both states have lots of white, working-class voters — they’re a sign that the map isn’t necessarily going back to normal even if Clinton wins the election.
But it’s not so much that Clinton’s out of the running in these states — she’s more likely to win Iowa than to lose Pennsylvania, according to our forecast — as that it’s hard to see her winning them unless she’s doing well elsewhere in the Midwest. And if she’s doing well elsewhere in the Midwest, she’ll find it hard to lose the Electoral College. Thus, Ohio and Iowa rank poorly according to our tipping-point index, which accounts for these correlations.
Still, Ohio might have one or two things that make it a good target for Clinton, such as Trump’s frosty relationship with the state’s Republican Party and slightly tighter polls in the past week. Iowa, by contrast, might go the way of Missouri, a state that went from bellwether to off-the-swing-state-map without really pausing in between.
Group 6: The four-years-too-soon states
- Arizona (Trump is a 71 percent favorite; 3 percent chance of tipping election)
- Georgia (Trump is an 81 percent favorite; 2 percent chance of tipping election)
These states, on the southwestern and southeastern corners of the map, could be the cornerstone of Democrats’ 2024 coalition. But what about this year? As the race has evolved, Clinton’s numbers have held up better in Arizona than in Georgia — which is consistent with her potentially beating Obama among Hispanics while underperforming with African-Americans. Arizona’s vote is also potentially more elastic than Georgia’s, meaning that there are more swing voters, whereas in Georgia Democrats have a path to 48 percent of the vote but maybe not to 50-percent-plus-one.
But again, the question is not whether Clinton can win these states, but whether they could plausibly be the tipping-point in this election. Arizona, for instance, could play a role in maps like the one below, in which Clinton sweeps the western swing states but struggles in the Midwest:
Unlikely? Yeah, unlikely. But when the electoral map changes — as it did in 1992, for instance — the changes sometimes come all at once. If the polls don’t necessarily have an overall bias, but have substantial errors among certain demographic groups, you could wind up with some funky outcomes like these.
Group 7: The “wait, really?” states
- Minnesota (Clinton is an 84 percent favorite; 3 percent chance of tipping election)
- New Mexico (Clinton is an 80 percent favorite; 2 percent chance of tipping election)
Trump has made late visits to both states, earning a lot of mockery as a result — but the fact is that our model sort of agrees with his decision. Our polling average in New Mexico shows Clinton up by only 5 percentage points. It’s hard to fathom that result given New Mexico’s demographics, but Libertarian Gary Johnson, the state’s former governor, may be claiming most of his votes from Clinton.
Minnesota would probably have required more of a concerted effort from Trump to really put it into play, but the model marks it as somewhat uncertain because there’s been relatively little polling there. If Clinton is having so many problems in Iowa, is it totally implausible that she could be struggling in the state just on its northern border?
Group 8: The wild cards
- Alaska (Trump is a 77 percent favorite; less than a 1 percent chance of tipping election)
- Maine’s 2nd Congressional District (Trump is a 52 percent favorite; less than a 1 percent chance of tipping election)
- Maine statewide (Clinton is a 79 percent favorite; 1 percent chance of tipping election)
- Nebraska’s 2nd Congressional District (Trump is a 62 percent favorite; less than 1 percent chance of tipping election)
- Utah (Trump is an 84 percent favorite; less than a 1 percent chance of tipping election)
Finally, we come to the states that would matter only in the event of a very close election — particularly, in those cases where Clinton loses Nevada and New Hampshire but wins the rest of her firewall, leaving her stuck on 268 electoral votes. Note that it’s 268 electoral votes instead of 269 because Clinton’s firewall doesn’t include the northern, rural, 2nd Congressional District of Maine, whose demographics are almost perfect for Trump — although recent polls there have shown the race tightening from a previous Trump lead. The two statewide Maine electoral votes could also be in play, although Clinton will almost certainly win the 1st District, which includes Portland.
Nebraska also splits its electoral votes by congressional district, and the Omaha-based 2nd District (which went for Obama in 2008) leans toward Trump according to the handful of polling conducted there. But the 2nd district has a ton of college-educated voters and it could be a good target for Clinton’s analytics team. In Utah — and yes, it’s remarkable that we’re even talking about it at as a potential swing state — independent Evan McMullin’s support has stalled out in the high 20s after once threatening Trump, while Trump leads Clinton by around 10 percentage point in recent polls. Of the two candidates, McMullin is more likely to upset Trump than Clinton, since third-party candidates’ vote shares are more unpredictable. If a lot of Clinton supporters decided to take a chance on McMullin, he could surge at the last minute, for instance.
But my favorite wild-card state is Alaska, where Clinton actually led the most recent telephone poll. The state has been competitive all year in a variety of partisan and nonpartisan polls. The demographic component of our model is skeptical, projecting that Trump “should” win Alaska by 10 percentage points given where the polls are in other states. Still, if Trump is at 268 electoral votes and Alaska hasn’t been called yet, you can’t quite take anything for granted.
Nate Silver is the founder and editor in chief of FiveThirtyEight.