Nate Cohn

Nate Cohn @Nate_Cohn

The most important one--and it's hard to believe: the study does not appear to even consider the effect of weighting on education under any of the scenarios tested? If they did, it's not reported. So... that doesn't really offer much that's relevant to the question — PolitiTweet.org

Nate Cohn @Nate_Cohn

The most important one--and it's hard to believe:the study does not appear to even consider the effect of weighting on education under any of the scenarios tested? If they did, it's not reported. So... that is obviously an issue for the whole thing — PolitiTweet.org

Nate Cohn @Nate_Cohn

If the point, though, was to determine whether they needed to weight by education, then this study has a number of fairly serious issues — PolitiTweet.org

Nate Cohn @Nate_Cohn

Then, they have an extensive study of the effect of various sample selection choices on the result of the Kentucky governors race. They find that one of the five options tested did pretty well, which fair enough. — PolitiTweet.org

Nate Cohn @Nate_Cohn

in the article, the only argument *against* weighting on education is that it's hard to know the right target. unfortunately, this is 100% inapplicable to the Marist poll: they are an RDD survey of the general population, so they can use the census as they do for everything else — PolitiTweet.org

Nate Cohn @Nate_Cohn

This piece/study does not contest--and appears to concede--that education is correlated with both vote choice and nonresponse. if that's true, you weight on it. as far as i'm concerned, that's really all you need to know — PolitiTweet.org

Nate Cohn @Nate_Cohn

That said, this piece/study is attempting to justify why the NBC/Marist poll doesn't need to weight by education and I would have to say that this conclusion does not really follow from their findings — PolitiTweet.org

Nate Cohn @Nate_Cohn

the good thing: stratification and/or weighting on tightly defined regions correlated with urbanization/partisanship. it doesn't obviate the need to weight by education, but it's good and pollsters should do it. weighting by 'media market' or whatever isn't good enough — PolitiTweet.org

Nate Cohn @Nate_Cohn

The good thing here: stratification and/or weighting on tightly defined regions correlated with urbanization/partisanship is a good thing. it doesn't obviate the need to weight by education, but it's good and all pollsters should do it. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@arnonmishkin :( — PolitiTweet.org

Nate Cohn @Nate_Cohn

@arnonmishkin unfortunately you read it wrong (asked to double check) — PolitiTweet.org

Nate Cohn @Nate_Cohn

@arnonmishkin unfortunately you read it wrong — PolitiTweet.org

Nate Cohn @Nate_Cohn

I like all the parties involved, but if this is an attempt to justify not weighting by education, then this article/study does not seem to have made an effort https://t.co/qSJu680U4C — PolitiTweet.org

Nate Cohn @Nate_Cohn

But the idea that GOP losses might now well past '18 shouldn't be a surprise: Trump's approval rating today is a net-6 points lower than it was heading into the midterm. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Of course, the 'national trends may not manifest uniformly point' is particularly true by district. It wouldn't be hard to imagine, hypothetically, Biden winning an IN05 by 7 without flipping FL16. Real world election results always defy simple demographics — PolitiTweet.org

Nate Cohn @Nate_Cohn

Biden wins several of the Trump +10 or more districts that keep popping up in private polls, like IN05, MO02, TX21, etc. But he also wins districts that aren't getting much attention, perhaps because they're not overly heavy on college grads, like AZ06, OH10, MI06, FL16 — PolitiTweet.org

Nate Cohn @Nate_Cohn

The consequences for the GOP in the House might be equally alarming, as Biden carries more than 250 congressional districts in this scenario At this point, Biden gains among white voters really do reach, if barely, the 'busted gerrymander' point in OH/TX/MI. — PolitiTweet.org

Nate Cohn @Nate_Cohn

As an aside, the 2022 Senate map does not feature many competitive Democratic-held races, and if Dems could get to 53+ seats they'd have a very serious path to hold the Senate through 2024 and a hypothetical midterm wave — PolitiTweet.org

Nate Cohn @Nate_Cohn

Obviously it doesn't matter too much whether Biden flips AK or something. But it may matter a lot downballot: four of those states feature plausibly competitive Senate races, and extra Senate flips would matter a lot for Democrats. — PolitiTweet.org

Nate Cohn @Nate_Cohn

As cautioned earlier, individual states will defy national trends in significant ways, esp weird ones like AK/UT. But this creates risk for the GOP: it wouldn't take much more for a few to be really competitive, even if, for the same reasons, one or two might remain solid Trump — PolitiTweet.org

Nate Cohn @Nate_Cohn

All of this shouldn't be too surprising. More surprising is what comes next: AK Trump+4 UT Trump+5 SC Trump+7 MT Trump+7 KS Trump+8 MO/IN Trump+9 — PolitiTweet.org

Nate Cohn @Nate_Cohn

The next-tier of states also looks quite a bit like state polls, though the Democratic weakness among Hispanic voters in national polls (which I'm not sure I expect to last and may not materialize in TX) barely blocks blue TX ME-2 Biden+2 OH +2 IA +1 GA +1 TX Trump+1 — PolitiTweet.org

Nate Cohn @Nate_Cohn

By state, these estimates yield results that look a lot like recent polls (though keep in mind that, unlike this simulation, national shifts would undoubtedly manifest in non-uniform and unpredictable ways by state): MI Biden+10 WI +10 PA +9 FL +6 NC +6 AZ +4 — PolitiTweet.org

Nate Cohn @Nate_Cohn

Nationwide, this yields the following result: Biden+10, 54-44 Whites 46-52 ...col+ 59-39 ...no col 37-61 Black 90-9 Hispanic 63-35 plausible enough for illustrative purposes — PolitiTweet.org

Nate Cohn @Nate_Cohn

It's not easy to imagine the consequences of this kind of demographic shift, but they are potentially very significant. For illustrative purposes, consider a simple simulation: apply the shifts in national polls to our post '16 estimates of the result by group — PolitiTweet.org

Nate Cohn @Nate_Cohn

Trump's lead among white voters has all but vanished. Anything like it threatens longstanding GOP structural advantages https://t.co/2M0AowUu1r https://t.co/XnYkATKB58 — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @maggieNYT: This is unbelievable - Trump forced the RNC to find another place other than Charlotte NC, so they went to Fla. Now he's can… — PolitiTweet.org

Nate Cohn @Nate_Cohn

I could come up with 15 plausible hypotheses to explain the gap, some suggesting either side could be more or less correct. I'd have basically no evidence for any of them, so for now I'll just leave you my '19 thoughts on the state of online polling https://t.co/2h6LU3CXyN — PolitiTweet.org

Nate Cohn @Nate_Cohn

If we only had live interview polling, we'd probably say these last ten days of results among RVs--Q+15, ABC/Post+15, NBC/WSJ+11, Monmouth+13 PA, Q+13 FL--we might say that's Biden's best period yet — PolitiTweet.org

Nate Cohn @Nate_Cohn

It's been remarked on before, but recently there is a clear split between the live interview and online polling, with the live interview poling showing an unrelenting and even growing Biden lead while the online stuff shows Biden losing a bit of ground — PolitiTweet.org