Douthat’s main points are that if the polls continue to tighten, then Trump will be in a position to possibly win; that the most-likely scenario for polls continuing to tighten is for third-party candidates not to fade; and that Trump has a bit of an advantage in the electoral college in a very close outcome. I tend to agree with all of those points.
Larison takes issue with what he sees as Douthat’s implicit assumption that Trump will be able to mobilize voters that show up as supporters, particularly if they are disproportionately infrequent voters. It’s entirely fair to question whether these people will actually show up (since they usually don’t), but I’m not sure that a strong ground game (which Larison correctly points out Trump doesn’t have) actually makes as much of a difference for these voters specifically. The thing about a ground game is that it’s primarily about getting voters you’ve identified to the polls. If you don’t know who your voters are, because they’re historically disconnected from the political process, then that’s harder to organize around. If any plausible scenario for a Trump victory revolves around him generating enthusiasm among unlikely voters who likely aren’t on anybody’s list, then the ground game may not be Trump’s most important game — compared to whatever strategy does motivate these people to come out and vote for him.
Of course, you can still lose if you generate higher turnout among unlikely voters but also fail to deliver your more traditional voters, the sorts of people a good ground game is designed to bring out. So I still think it’s reasonable to assume that Trump’s poor ground game will cost him — just not necessarily for the reason Larison gives.
The rest of Douthat’s argument, though, strikes me as pretty solid. Nate Silver has been saying for some time that Trump may have a slight edge in the electoral college if the race is extremely close, for the very reasons Douthat highlights: Trump lags a typical Republican nominee in a bunch of red states more than he does in some of the traditional swing states. In other words, Trump’s vote is arguably more efficiently distributed than a typical Republican nominee’s.
That doesn’t mean that Trump is more likely to win than a typical Republican nominee. It means that if the election is close, then Trump has a bit more of an edge. But it could simultaneously be true that, with Trump on the ticket, it’s less-likely to be close in the first place.
Meanwhile, third-party candidates are very hard to model, both because there’s little history and because the history that we do have is highly idiosyncratic. (How much do George Wallace’s, John Anderson’s and Ross Perot’s campaigns really have in common?) Which means that many different scenarios are plausible once you make the assumption (as Douthat does) that Johnson and Stein don’t fade as the election gets closer. And one very plausible scenario could be that Johnson and Stein wind up hurting Clinton more in the swing states, while they hurt Trump more in the safe red states. And in the extreme case, yes, that could lead to the sort of scenario Douthat outlines at the end where Trump wins without a plurality in the popular vote.
In theory, it’s also plausible that the opposite could be true, and third-party nominees could hurt Trump more. Anderson initially looked like he was going to hurt Reagan, but wound up hurting Carter. This year, Stein could fade, and Johnson could become the vehicle for libertarians, Mormons and other folks who would never vote Clinton but don’t want to assent to Trump, and could thereby tip Nevada and New Hampshire to Clinton to win a race where she marginally loses the popular vote. Right now, though, it feels like Clinton hasn’t held on to the younger voters who went overwhelmingly for Obama, and hasn’t closed the deal with moderate Republican-leaning suburbanites who she needs to offset losses among working-class whites who are swinging Trumpward. Some of the former are going to Stein and some of both are going to Johnson.
Finally, the electoral college needle that Trump needs to thread isn’t nearly as narrow as Larison suggests, because the swing states tend to swing together. Yes, it’s true that Clinton has a larger electoral college base (states where the projected margin has been larger than the national margin). But arguably so did John Kerry in 2004: Bush won only 254 electoral votes by a margin at least equal to his national margin of 2.4%, while Kerry won 237 electoral votes by at least 2.4% in an election where he lost the national vote by that margin.
But all that means is that a large Clinton victory is more likely than a large Trump victory. If Clinton wins every state where she has at least a 40% chance of victory according to Nate Silver, she’d win 348 electoral votes. If Trump won all the same states, he’d be at 266 — just shy of a majority, needing New Hampshire to put him over the top. If those states were all independent contests, than a sweep like that would be vanishingly unlikely. But they aren’t — a victory in Ohio means a victory in Florida is more likely as well. Which is why Trump has a real shot at victory — about 1 shot in 3 according to Silver.
I still would bet on Trump to lose — and I think his odds are worse than 1-in-3 to win. But that’s because of my personal read on him as a candidate, and Clinton as a candidate. I think the numbers are somewhat cyclical, so Clinton’s last big boost has been followed by a drop; I think Trump is going to be poorly served by the debates; and I think the window for events to significantly transform the campaign in Trump’s direction is closing fast. By the numbers, though, he’s totally got a shot.