Nate Cohn

Nate Cohn @Nate_Cohn

In our post-election analysis, turnout was the overwhelming reason why we missed in the Southwest, not support. Not only were there more Latino voters, but also a much younger and more Dem group. — PolitiTweet.org

Nate Cohn @Nate_Cohn

The undecided voters disapprove of Trump by a 2:1 margin, 60-30, so there's some upside for Biden there https://t.co/SPeWq0WAJr — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @MediumBuying: UPDATE: The Biden campaign is going up in the Dallas-Ft. Worth DMA for the final week https://t.co/cuu22Lxm2N — PolitiTweet.org

Nate Cohn @Nate_Cohn

And yes https://t.co/tPc2id9ab3 — PolitiTweet.org

Nate Cohn @Nate_Cohn

Take our polls of AZ and NV, for instance. In '18, we had McSally and Heller ahead--wrongly. I'd argue our AZ polls have been the very best for Dems of anyone so far this cycle. NV seems perfectly reasonable. So I don't think 18 necessarily predicts 20 — PolitiTweet.org

Nate Cohn @Nate_Cohn

The question, though, is why were we wrong? Was it because of sampling--we just can't reach the right voters? If so, then the pattern would repeat itself. But there are other possible causes that may not repeat itself--at least not to the same extent — PolitiTweet.org

Nate Cohn @Nate_Cohn

Well, we certainly did underestimate Hispanic turnout/vote in '18. And it's possible we'll do so again. Hispanic voters are challenging everywhere, but this is a state where there's no real record of competitive races, no party registration, etc https://t.co/i3hwt8JKS4 — PolitiTweet.org

Nate Cohn @Nate_Cohn

It's 5 percent undecided, which I do think is pretty reasonable. In addition, you've got over 3 percent saying Lib, 1 percent saying someone else, 1 percent saying they won't vote on the presidential ballot https://t.co/Dc0lugpXYk — PolitiTweet.org

Nate Cohn @Nate_Cohn

Small sample obviously, but Biden. Somewhat more robust: tie across the big fast growing suburb/exurb battlegrounds: Collin/Tarrant/Williamson/Denton/Hays. https://t.co/xpDChWAjHi — PolitiTweet.org

Nate Cohn @Nate_Cohn

Is it possible that Biden can keep drawing such a favorable group of voters to the polls, and upset the estimate here? I can't rule it out. It is also hard not to note that the president's biggest strength here--less educated nonwhite voters--is a very hard group to measure — PolitiTweet.org

Nate Cohn @Nate_Cohn

Turnout is obviously a key question in a state with no recent record of competitive contests. In early voting, Biden has amassed a 7 point lead among those who cast ballots by Friday, per L2 data. But Trump has a far larger lead among those who say they'll vote later. — PolitiTweet.org

Nate Cohn @Nate_Cohn

In the survey, white college grads in these competitive districts have swung (per their own self-report) from Trump+24 in '16 to Trump+2. But nonwhite voters statewide have swung toward Trump, modestly but significantly given they represent nearly half of voters — PolitiTweet.org

Nate Cohn @Nate_Cohn

The GOP still faces some pretty catastrophic risks downballot, as Biden leads by 5 across the competitive House districts (which were Trump+8 in 2016). There's barely any swing toward Biden in the rest of the state, where white no college, Black and Hispanic voters prevail. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Donald Trump still leads in Texas, according to the final Times/Siena poll of the state, as strength among nonwhite voters does just enough to overcome all out rebellion in the suburbs Trump 47, Biden 43 https://t.co/LP92cGZCgK — PolitiTweet.org

Nate Cohn @Nate_Cohn

@doug_rivers @gelliottmorris @BrendanNyhan given that state/demog correlated errors are pretty large, those aren't negligible negative correlations you're generating in this case! — PolitiTweet.org

Nate Cohn @Nate_Cohn

@doug_rivers @gelliottmorris @BrendanNyhan exactly--and which one will prevail depends on the size of the negative correlated, state errors v. the effective positively correlated national error, on these particular states (one of which is highly 'inelastic') — PolitiTweet.org

Nate Cohn @Nate_Cohn

@zjwhitman @gelliottmorris @BrendanNyhan only because i thought it was something we were sure to agree on, before we advanced into the more complicated settings — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @BrendanNyhan Are you saying you disagree with this?: if you run a simulation, stipulating zero national error, and state correlated error is simulated based on demographics, then there will probably be a negative relationship between MS and WA. — PolitiTweet.org

Nate Cohn @Nate_Cohn

33 percent do not have a validated record of voting in the 2016 election; 16 percent say they didn't vote (may just be overreporting, some could have voted in another state etc.) https://t.co/Rjye8pnpso — PolitiTweet.org

Nate Cohn @Nate_Cohn

GOP+4; Trump+10 on 2016 recalled vote https://t.co/RGri72xFCI — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @BrendanNyhan surely you'd agree that if we stipulate there's *no* national error, and WA is biased one way, that MS may need to be likelier than not to be biased the other way if you're using demographics as the basis for simulating state error? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @BrendanNyhan might have copy and pasted something wrong, but https://t.co/NlZg6K7IwF — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @BrendanNyhan and interestingly enough, it does seem like that's how it worked out in the economist 08-16 backruns — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @BrendanNyhan i'm not sure that's right — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @BrendanNyhan is it? seems to me like it very might well be true historically. don't you have backfit version of your model to evaluate whether it's so insane? — PolitiTweet.org

Nate Cohn @Nate_Cohn

TX is mainly pre-debate. All subsequent polls are entirely post-debate https://t.co/mIbSFnYrPt — PolitiTweet.org

Nate Cohn @Nate_Cohn

Today at 1PM: NYT/Siena in Texas Next up: Nevada, Michigan, North Carolina https://t.co/2KNyblIsDt — PolitiTweet.org

Nate Cohn @Nate_Cohn

Anyway, the polls could still be wrong. But unlike in 2016, where you could see the warning signs, I don't see much cause for concern at the moment — PolitiTweet.org

Nate Cohn @Nate_Cohn

I'm not very concerned about turnout? Or at least not any more than usual. In many ways, high turnout can make polling easier, not harder. But I do think it's easy to imagine issues in traditionally noncompetitive Sun Belt states, where high turnout could have big consequences — PolitiTweet.org

Nate Cohn @Nate_Cohn

I am still concerned about the white working class Midwest. We just don't know how deep the challenge is there. I am concerned about relatively Dem samples of Florida Latinos in statewide polls without a special design. — PolitiTweet.org