Nate Cohn

Nate Cohn @Nate_Cohn

Biden's lead over Trump has shrunk a bit over the last few months, but Biden hasn't really lost much ground https://t.co/O9zvc7YYUH — PolitiTweet.org

Nate Cohn @Nate_Cohn

@NateSilver538 But if you accept the critique of economic fundamentals in a plainly out-of-sample year, which I think you do, then it's hard for me to justify leaning on a prior other than 50/50 — PolitiTweet.org

Nate Cohn @Nate_Cohn

@NateSilver538 I think it's a perfectly fine economic-fundamentals based prior, and I *suspect* a non-Trump president would be favored to win reelection — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @NateSilver538 again, i think the model pages are quite clear. but the EC/PV split helps explain a bit of the gap here, as ~tied nationally is not 50/50, like you said — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @NateSilver538 dk what he said elsewhere. i think their forecast pages are quite clear — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris a tied election may be exactly the right prior, IMO. my point is that their prior is not a tied election. — PolitiTweet.org

Nate Cohn @Nate_Cohn

I've seen a lot of people badgering @NateSilver538 for what the 2016 model would say today. This makes me more interested in the question: what would the "polls only" version of this year's model say? — PolitiTweet.org

Nate Cohn @Nate_Cohn

If you think this is basically an out of sample election, I'd think your prior would be non-existent/ 50/50. But if you say the fundamentals make Trump a 5 pt fav in tipping points, and that's 18% of the estimate, then I think you're making a claim you do know the fundamentals — PolitiTweet.org

Nate Cohn @Nate_Cohn

I thought I was sympathetic to FiveThirtyEight on the fundamentals, since I think we largely agree that we have zero clue about the fundamentals in this election. I absolutely do not share the view, for ex, that we can be confident that the economy made Trump a 10 pt underdog — PolitiTweet.org

Nate Cohn @Nate_Cohn

So far, uncertainty has been the key front in the 2020 model wars. And I have to say that I'm pretty sympathetic to the 538 side of the case there. But the fundamentals play a big role in these models as well, and I'm a little farther from 538 than I thought I was on this — PolitiTweet.org

Nate Cohn @Nate_Cohn

So far, uncertainty debates has been the key front of the 2020 model wars. And I have to say that I'm pretty sympathetic to the 538 side of the case there. But the fundamentals play a big role in the models as well, and I'm a little farther from 538 than I thought I was on this — PolitiTweet.org

Nate Cohn @Nate_Cohn

FiveThirtyEight forecast pages now include a fairly detailed account of their estimates. The most interesting thing is that their incumbency+econ-fundamentals make Trump a more sizable favorite than I had assumed, with a prior of Trump nearly up 5 in the tipping pts — PolitiTweet.org

Nate Cohn @Nate_Cohn

If you have a limited budget I suppose it might make some sense, though in many states (but not AZ) the early voters are pretty partisan and include relatively few undecided voters https://t.co/bLITN6xeKA — PolitiTweet.org

Nate Cohn @Nate_Cohn

Well we're going to need some nonpartisan NE02 polling at some point, but on paper Biden could easily be favored to win it in a scenario where he had won WI/MI https://t.co/ukZaLTUSsk — PolitiTweet.org

Nate Cohn @Nate_Cohn

Also Change Research was just Biden+2, though at N=344 I'm not sure it's getting a ton of weight exactly. — PolitiTweet.org

Nate Cohn @Nate_Cohn

In the last month, we've got Biden+10 in MC, Biden+8 in Fox, Biden+7 in Emerson (which has a GOP house effect at the moment). The last real "good" news for Trump, as far as I can tell, are the Biden+4/5 polls from late July in CNN and Marist — PolitiTweet.org

Nate Cohn @Nate_Cohn

I was already pretty intrigued by the polling situation in Arizona. There have been quite a few good numbers for Biden this month, and there are reasonable ways of calculating a polling average that now puts AZ left of the tipping point https://t.co/8SvFuw5ZVk — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Redistrict coulda saved u some work https://t.co/cdyBLNSOxg — PolitiTweet.org

Nate Cohn @Nate_Cohn

Anyway, those are just a few theories. I can make theories in the opposite direction, too. But if, hypothetically, polls show PA is the closest of the bunch and Biden leads MI/WI more comfortably, I don't think we'd have a shortage of plausible explanations — PolitiTweet.org

Nate Cohn @Nate_Cohn

If Trump's losing ground among white voters, it makes sense that the distribution of those losses may not be uniform--and could manifest in the places where we know that a certain number of GOPers have had relatively large, longstanding reservations about Trump — PolitiTweet.org

Nate Cohn @Nate_Cohn

During the Trump era, we've become accustomed to treating white voters without a degree as a homogenous group. And that's basically how they acted on election day in 2016. But that's not how it worked for basically all of American history, including in the GOP primary! — PolitiTweet.org

Nate Cohn @Nate_Cohn

PA is also the only state of the bunch where Trump has a really solid GOP base (just check out the 2016 primary results). At a time when Trump struggles, it makes sense that he might lose a lot more ground in west Michigan or the MIL burbs than he would in the PA GOP heartland — PolitiTweet.org

Nate Cohn @Nate_Cohn

Another kind of theory is demographic: WI is the whitest state of the bunch. Biden's gains are among white voters--and, yes, regardless of education/urbanicity. Therefore, maybe Biden should be making the largest gains in WI — PolitiTweet.org

Nate Cohn @Nate_Cohn

As a result, it's harder to correct for various kinds of partisan bias in WI/MI and especially WI. If, hypothetically, it was more challenging to reach GOPers in this region, pollsters can do more to fix it in PA than in WI/MI, and you'd get a similar pattern — PolitiTweet.org

Nate Cohn @Nate_Cohn

There's one hard, data-driven reason why polling in PA might be easier than MI and particularly WI: party registration. PA has it. MI/WI don't. And WI doesn't even have primary vote history. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Here we can transition to another sort of theory: that this is an artifact of state-by-state poll *bias*. Here, the argument is that the pattern of '16 poll bias wasn't due to IRL distribution of undecided voters, but just where state polls struggled more or less — PolitiTweet.org

Nate Cohn @Nate_Cohn

After all, the pattern in the polling today is... the same pattern as the pre-election 2016 polling! Trump may win back these undecided voters in the final account, as he did in '16, and outperform the polls in WI/IA and to a far lesser extent --if at all -- in PA. Or not. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Relatedly, one could argue that the same fact would affect the distribution of undecided voters in a way that yielded the same pattern. The places with more '16 swing would have more undecided voters, mainly at Trump expense, leading to a more '12 like pattern. — PolitiTweet.org

Nate Cohn @Nate_Cohn

One theory is that Biden gains in the northern battlegrounds involve some mean-reversion to the Obama-era map, which might involved outsized Biden gains in a place like IA or OH, but few in PA. That doesn't seem insane to imagine, esp with Biden doing so well among white voters — PolitiTweet.org

Nate Cohn @Nate_Cohn

Here's one basic observation to consider: PA closer than MI/WI would something of a return to the Obama-era alignment. PA was quite a bit closer in '08/'12 than WI/MI. From there, we can work to a number of plausible theories — PolitiTweet.org