Nate Cohn

Nate Cohn @Nate_Cohn

That said, I do think that, on balance, the polls ought to be pretty unbiased! Now that's no guarantee that they will be. But I'd be sounding the alarm if I didn't think so, as I was for much of the last two years on education (almost everyone has now switched) — PolitiTweet.org

Nate Cohn @Nate_Cohn

I wouldn't say that pollsters are 'shaving down' a couple of points, per se! But pollsters have spent four years trying not to underestimate Trump, and that does make it easier to imagine that you overlook things the other way https://t.co/otTKXwRT6k — PolitiTweet.org

Nate Cohn @Nate_Cohn

Biden by 0.3 in the new AJC poll of Georgia, 46.5-46.2 Ossoff by 1, 46-45 https://t.co/i21Cmot4wo — PolitiTweet.org

Nate Cohn @Nate_Cohn

I've got good news for everyone https://t.co/nosaMooGWJ — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @bluestein: 🚨 🚨 The @ajc will publish the results of the latest poll of Georgia tomorrow at 6AM, with breakdowns of the presidential rac… — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Rasmussen_Poll ? https://t.co/3NMD2tSSX4 — PolitiTweet.org

Nate Cohn @Nate_Cohn

Interestingly, these were the first Trafalgar crosstabs that I can recall being published, too. Until now, they've shown like three not-very-useful bar charts on the composition of the electorate by race, age and CD. https://t.co/RuYiTWONRH — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @MediumBuying: FLASH: The Biden campaign is canceling TV ad schedules that had been booked in TEXAS (10/27-11/3 flight) They are still… — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @jmartNYT: Here we to... @KamalaHarris to TEXAS on Friday, per a Dem briefed — PolitiTweet.org

Nate Cohn @Nate_Cohn

Two traditionally GOP-leaning, IVR (automated phone polling) firms. A nice reminder that using IRV doesn't doesn't just let you automatically reach some pool of hidden Trump voters https://t.co/w2zj7JWrET — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @jmartNYT: And don’t miss where he’s going — not just Atl but also Warm Springs to summon FDR. — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @CNN: For months Dr. Jinendra Satiya has faced fears on the frontlines of a deadly pandemic, treating coronavirus patients. But now the… — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kabir_here post-debate? — PolitiTweet.org

Nate Cohn @Nate_Cohn

There's been research that COVID spikes lead to a bit of a shift toward Biden in the polls. With that in mind... https://t.co/bi3nsAGFKO — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @SienaResearch: Siena College/Syracuse.com NY-24 Poll: Katko 45% – Balter 45% – Williams 5% – Undecided 4% https://t.co/AywbM8zsBw — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @AdamBassWCCS: NEW: @JoeBiden will be traveling to GEORGIA on Tuesday. More details to come. — PolitiTweet.org

Nate Cohn @Nate_Cohn

And no one should be surprised by a wide Dem mail ballot edge in WI. It's the same story as Pennsylvania, except you know it there because of there's party registration. — PolitiTweet.org

Nate Cohn @Nate_Cohn

We've slowly been building up our own models in these states as we collect more data. Back in our first poll of WI-1 in '18, before we ever had any WI data, you can see how mediocre they were https://t.co/ItVaezHiTY — PolitiTweet.org

Nate Cohn @Nate_Cohn

Our voter file vendor (L2) classified these Biden 82, Trump 18 respondents as Dem 42, GOP 37. I'm not trying to be critical of either firm; there's just not much you can do to reliably identify partisan voters in a state without partisan data and limited geo/demo polarization — PolitiTweet.org

Nate Cohn @Nate_Cohn

A lot of people are asking about TargetSmart modeling, which shows a tight race in partisan balloting. The problem, as I've alluded to already in this thread, is that there's not really any good data (until mail voting lol) for identifying partisans in Wisconsin — PolitiTweet.org

Nate Cohn @Nate_Cohn

Because nearly as many Democrats voted as Republicans, so it's not very useful. Honestly, based on our data, I'd guess the 2016 primary electorate in WI voted Clinton, even though D>R ballot, with so many college grad GOPers voting Clinton later https://t.co/NX838KVA6i — PolitiTweet.org

Nate Cohn @Nate_Cohn

Hilariously one of our better ways to 'see' Republicans is whether you voted in the August 2012 primary. A real barnburner on the GOP side lol https://t.co/eB3Pdi1ewX — PolitiTweet.org

Nate Cohn @Nate_Cohn

The risk, interestingly, is that using this data could hurt Democrats in a Wisconsin poll. In general, partisans are most likely to respond to surveys. Here, we have a great way to see Dems! But we don't quite have a similar way to see Republicans, so they might remain too high — PolitiTweet.org

Nate Cohn @Nate_Cohn

It's potentially pretty useful data. Wisconsin is not a state with party registration or partisan primary vote, so all of this lopsided mail voting does make it easier to identify Democrats and Republicans... though doing so does come with risks. — PolitiTweet.org

Nate Cohn @Nate_Cohn

We've conducted four polls of Wisconsin over the last year, and we can go back and join them up to some of the absentee voting data there. It's a little out of date, but the folks who have already voted, per our data, backed Biden... 82-18 among decided major party voters. — PolitiTweet.org

Nate Cohn @Nate_Cohn

I think most polling supposes a high youth turnout. The catch, though, is that the RV pool is older than it was four years ago. The country is aging! Higher youth turnout may not get the youth share of the electorate much above where it was https://t.co/DNllzPgLeb — PolitiTweet.org

Nate Cohn @Nate_Cohn

@trevorjtweets @davidshor afaik, the critic uses the same approach but differs in execution — PolitiTweet.org

Nate Cohn @Nate_Cohn

@davidshor i may not have read closely enough, but it is not clear to me what kind of national error was supposed in those particular simulations — PolitiTweet.org

Nate Cohn @Nate_Cohn

@davidshor definitely, but conditional on the approach taken by 538/economist it is what it is — PolitiTweet.org

Nate Cohn @Nate_Cohn

@davidshor i'd also note, fwiw, that MS probably would have been an underpolled state in both '64 and '20, and therefore there'd be a lot of leaning back on the fundamentals. so i'm not sure that would be true in the forecast error — PolitiTweet.org