The presidential election is a toss-up

Kamala Harris is giving Democrats the race that Joe Biden couldn’t.

I don’t want to get in the habit of making a huge ruckus every time the election model passes a certain arbitrary probabilistic threshold. There are still a very long 96 days to go until the election, and although polls are more stable than they used to be, they’re still likely to gyrate back and forth a number of times between now and Nov. 5.

But I’m going to make an exception here, because when we launched the presidential model on June 26 — in the lifetime ago when Joe Biden was the Democratic nominee — the headline in the post that introduced the model was that the election wasn’t a toss-up. Instead, Biden had persistently been behind in the states that were most likely to decide the Electoral College, enough so that he was about a 2:1 underdog in the election despite the uncertainties in the race. His situation wasn’t unrecoverable, or at least it wasn’t until the debate. But you’d rather have had Donald Trump’s hand to play every day of the week and twice on Sundays.

Now that the election is in kamala_mode, however, it’s far from clear whose position you’d rather be in, and I wouldn’t blame you if you wanted to bet either on Harris or on Trump. At FiveThirtyEight, we actually had a formal definition of a “toss-up”, which is an election where each candidate had at least a 40 percent chance of winning. We’re now quite comfortably into that territory. As of this afternoon’s model run, Harris’s odds had improved to 44.6 percent, as compared to 54.9 percent for Trump and a 0.5 percent chance of an Electoral College deadlock. It’s not exactly 50/50, but close enough that a poker player would call it a “flip”: Democrats have ace-king suited, and Republicans have pocket jacks.

It’s also a toss-up in the states that are most likely to decide the election. Let me share one of the charts that we usually reserve for paying subscribers:

Harris has a 54 percent chance of winning Michigan, a 50 percent chance of winning Wisconsin and 47 percent chance of winning Pennsylvania, states that would suffice to net her 270 electoral votes, one more than she needs to win (assuming she also holds lean-blue states like New Hampshire). She also has a 40 percent chance of winning Nevada, where her polling has been much better than Biden’s so far, and roughly a one-in-three chance in Georgia and North Carolina, which gives her some backup options that Biden lacked.

Democrats shouldn’t get too out over their skis about this. In principle, the model is designed in such a way that there’s no momentum (or autocorrelation, to use the more technical term). That is, just because Harris has risen in the forecast over the past couple of days (from 38 percent on Tuesday to 45 percent now) doesn’t imply anything either way about the future direction of the model. Now, you could maybe make an argument that this is an exception because of the unusual circumstance of Harris entering the race so late, but usually the model is pretty smart about this stuff. (After the debate, for instance, Biden’s odds fell quickly in the model, and then stayed at their new lower levels but didn’t fall much further.) The bear case for Harris — particularly that she could win the popular vote but lose the Electoral College — still applies. It’s even possible that this will prove to be Harris’s high-water mark and that the polls reflect a temporary surge of Democratic enthusiasm or even partisan non-response bias as Trump has been knocked back on his heels.

Still, we’re not exactly going out on a limb here. Betting markets agree with the assessment that the race is about 50/50:

And although we can’t tell you who’s going to win, there’s one thing I think we can say with some confidence: Democrats are lucky that they’re getting a second chance in this election with Harris instead of Biden.