Nate Cohn

Nate Cohn @Nate_Cohn

As an aside, the 2016 presidential election was held later in November (11/8), so although it's the same same date we're about 5 days closer to the election — PolitiTweet.org

Nate Cohn @Nate_Cohn

A historical note: 10/18 was more-or-less Clinton's October peak four years ago, as she topped out with a national lead of around 7 percentage points — PolitiTweet.org

Nate Cohn @Nate_Cohn

Yeah, and the Nevada voting situation--like every where--is quite a bit different. Way more mail voting. You can still vote in person. Who knows. https://t.co/hOckCkKhdp — PolitiTweet.org

Nate Cohn @Nate_Cohn

I should note, btw, that this thread is merely describing why advanced-vote based estimates have panned out well to date. There's no guarantee these conditions hold in the future, including this year. I don't *think* a D+4 electorate in NV will vote T, but we've seen it before — PolitiTweet.org

Nate Cohn @Nate_Cohn

I should note, btw, that this thread is merely describing why advanced-vote based estimates have panned out well to date. There's no guarantee these conditions hold in the future, including this year. I don't *think* a D+4 electorate in NV will vote T, but we've it before — PolitiTweet.org

Nate Cohn @Nate_Cohn

@ElectProject @doug_rivers yeah, though once you're at that point dif-in-dif is basically out the window and you're left waiting to see what the GOP does on election day — PolitiTweet.org

Nate Cohn @Nate_Cohn

@ElectProject @doug_rivers i think the all-mail states are probably workable, though even then i do wonder about the effect of the president politicizing the vote process. how many resistance-dems rushed their ballots out bc they've read about USPS delays? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@doug_rivers @ElectProject in terms of polling, i think the bias-variance tradeoff risks this year are really, really scary and i'd welcome any thoughts from you or anyone else on here — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Peoples_Pundit @jon_m_rob it's true, it wasn't clear--i'm just clarifying. and the cell skew among men has been clear for several years, not just in our polling but in the other polls that archive their raw data https://t.co/X33SsAa6yD — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Peoples_Pundit @jon_m_rob there's no temporal distribution in the gender breakdown of our respondents. as i said, we don't have any quotas on gender, so we don't 'fill up' on one group, like women, and start only calling men — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Peoples_Pundit @jon_m_rob no, 59/41 is the landline sample. 51/49 is the cell sample. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@ElectProject i'm deleting a tweet that i regret--it was rude. but, at the same time, i really don't see how we move toward better analysis with this kind of perspective on the way we thought about 2016 — PolitiTweet.org

Nate Cohn @Nate_Cohn

@ElectProject is this a joke? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Peoples_Pundit @jon_m_rob and in AZ, fwiw: https://t.co/oSU7dmMwCW — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Peoples_Pundit @jon_m_rob used to be true, but cells tilt male--something i know bc... we do poll? as a result, our raw sample over 16000 interviews, without quotas: https://t.co/4qUJfOyWR5 — PolitiTweet.org

Nate Cohn @Nate_Cohn

@RalstonReports @ForecasterEnten aw i was saying the concept was simple; we all appreciate your work — PolitiTweet.org

Nate Cohn @Nate_Cohn

Yeah, that's part of it. I'd add that it's also a really heterogeneous population, so varying underlying trends that might bias another state may be canceling out to an extent here https://t.co/BAWU0FB3YB — PolitiTweet.org

Nate Cohn @Nate_Cohn

There are very few places that meet both of these conditions. In some states, like CO, turnout is high enough and unbiased enough that you can guess the final electorate. But the unaffiliated vote has swung so overwhelmingly Dem that reg may not = vote anymore — PolitiTweet.org

Nate Cohn @Nate_Cohn

2: The NV early vote is both really large as a share of the electorate and relatively unbiased. So if the early vote is D+4 or whatever, the final electorate is usually about the same. Again, really simple, but totally untrue in many places. — PolitiTweet.org

Nate Cohn @Nate_Cohn

1: The relationship between party registration and presidential vote has stayed tight there. If the electorate is D+5, they often win by 5! Simple, but it's not true in a lot of places. Worse still, that relationship has changed a lot in many states: your old rules don't work — PolitiTweet.org

Nate Cohn @Nate_Cohn

As an aside, it's worth thinking about why the Ralston/NV-based predictions tend to hold up, while it can be embarrassing elsewhere. I think there are two basic reasons https://t.co/dlPNShttcc — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @SienaResearch: The New York Times / Siena College Battleground Polling so far. https://t.co/1esmlctzWe — PolitiTweet.org

Nate Cohn @Nate_Cohn

That's exactly the right feeling lol https://t.co/XaQ2OZePwc — PolitiTweet.org

Nate Cohn @Nate_Cohn

So on Election Night, if you see D+15 in a state's early vote, you should absolutely not be surprised if it comes in at Biden>30 or whatever. And you shouldn't be at all surprised when the GOP comes roaring back and closes it right back up — PolitiTweet.org

Nate Cohn @Nate_Cohn

Basically, the reg Dems/GOP/Ind voting early are all better for Biden than those who intend to vote on Election Day, and vice versa. If you prefer: the force drawing Dems than Reps to mail is also luring a relatively Dem set of Reps/INDs, and the most Dem set of Dems, and so on. — PolitiTweet.org

Nate Cohn @Nate_Cohn

This works both ways, to be clear: if a hypothetical early vote is D+15 by reg and the election day vote is R+15, then the early vote might be Biden+25 and the election day vote might be Trump+25 (just as an ex.) — PolitiTweet.org

Nate Cohn @Nate_Cohn

You'll see a lot of early voting statistics out there that cite the partisan makeup of early voters, like, say, a D+15 among absentees in Florida, based on registration. But remember this: the presidential preference of these voters is even more skewed than the party reg. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@jon_m_rob it's really one of the most astonishing threads i've ever seen — PolitiTweet.org

Nate Cohn @Nate_Cohn

Another few hundred dots left to go in Georgia, but thanks to 2x polls in GA/NC v. one poll in SC, this is already fairly proportionate to population already https://t.co/M0vPaam13i — PolitiTweet.org

Nate Cohn @Nate_Cohn

No, not really https://t.co/c6wi2NXASW — PolitiTweet.org