Pennsylvania may be a problem for Harris

The Electoral College/popular vote gap is increasing in post-DNC data.

We’re aiming to get back on schedule for Silver Bulletin Subscriber Questions by actually running the “August” edition in August — which means by Saturday. There’s still time for paid subscribers to submit questions.

---

I’m going to keep this brief, both because I’m jetlagged in two different directions¹ and because the model is really hungry for post-DNC/post-RFK-dropout data and may bounce around a fair bit until it gets more. But Kamala Harris had one of her worst days in some time in our forecast on Thursday despite gaining in our national polling average.

Harris is ahead by 3.8 points in our national poll tracker — up from 2.3 points the day before the Democratic Convention began — which does suggest some sort of convention bounce. However, she’s fallen to a 47.3 percent chance of winning the Electoral College versus 52.4 percent for Donald Trump. (The numbers don’t add to 100 because of the possibility of an Electoral College deadlock.)  Some of this is because of the convention bounce adjustment that the model applies to polls that were conducted during or after the DNC. It assumes Harris’s polls are somewhat inflated right now, in other words — just as it assumed Trump’s numbers were inflated after the RNC. While there’s a solid basis for this empirically, you could argue we’re under unusual circumstances because of her late entry into the race. So if you want to treat all of this as a little fuzzier than usual, I don’t really mind that. The good news for Harris is that if she merely holds her current numbers for a couple more weeks, she’ll begin to track up again in our forecast as the model will become more confident that she’s out of the convention bounce period.

There’s another, longer-term concern for Harris, though: it’s been a while since we’ve seen a poll showing her ahead in Pennsylvania, which is the tipping-point state more than a third of time in our model. Today, in fact, we added one post-DNC poll showing Pennsylvania as a tie, and another (conducted during the DNC) showing either a tie or Trump +1, depending on what version you prefer.

The model puts a lot of weight on this recent data because of all the changes in the race. And you can see why it thinks this is a problem for her: if she’s only tied in Pennsylvania now, during what should be one of her stronger polling periods, that implies being a slight underdog in November. 

Now, all of this could change quickly with one or two high-quality polls showing Harris ahead in the Keystone State. The model is relying more heavily than I’d prefer on the Emerson College poll as it’s the only fully post-DNC/RFK data point — although note, again, that the immediate pre-DNC polls hadn’t been great for Harris in Pennsylvania either.

But for now, we now show a 17 percent chance that Harris wins the popular vote but not the Electoral College, a big concern for her campaign all along. If she won the popular vote nationally by between 1 and 2 points, for instance, the model estimates that she’d still be a 70/30 underdog in the Electoral College:

It’s possible that RFK’s dropout and endorsement of Trump is having more effect in Pennsylvania and the other Rust Belt states than elsewhere, which have older, whiter and more disaffected electorates. And as I said, it’s also possible that all of this is noise and/or that the model is overdoing the convention bounce adjustment. But while Tim Walz has had a strong rollout as Harris’s VP, I can’t help but wonder what her numbers would look like with Josh Shapiro instead.

1. West Coast last week; London this week.