Nate Cohn

Nate Cohn @Nate_Cohn

There's really only one county where a decent amount of the vote has been counted, and that's Phillips County, where 'yes' is at 65% of the vote with nearly all precincts have been counted. It's just one data point, but it would be a modest underperformance for 'yes,' by my ests — PolitiTweet.org

Nate Cohn @Nate_Cohn

Democrats were about 50 percent likelier to vote early than Republicans, yielding a nearly even partisan split among early voters in a state where Republicans outnumber Democrats by nearly two-to-one. So no surprise that 'no' is off to a quick start. — PolitiTweet.org

Nate Cohn @Nate_Cohn

It's still very early in Kansas, where 'no'--which would protect abortion rights--is out to an early lead. So far, a disproportionate amount of the vote is from Democratic areas and nearly all of it is early vote, which is pretty Democratic-leaning as well — PolitiTweet.org

Nate Cohn @Nate_Cohn

@SeanTrende @jczive @TeSheaff libel (though in fairness, there's a shocking amount of false internet content, including on wikipedia, that erroneously says i was a parli debater) — PolitiTweet.org

Nate Cohn @Nate_Cohn

@DataProgress no tabs or composition of the electorate? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@davidshor @jon_m_rob @MattGrossmann @PoliticalKiwi that's definitely a serious problem with these analyses, including the one going around today, but i don't see why that supports any particular hypothesis here. both the ideology/education frames seem reasonable to me -- and you've advanced versions of both! — PolitiTweet.org

Nate Cohn @Nate_Cohn

@SpecialPuppy1 @jon_m_rob @FiveThirtyEight would have hurt in '20 unless it's applied inconsistently (only if it shows trump underperforming '16, for ex) — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @KevinQ: Our first job is for a deputy editor — someone who can edit and envision explanatory journalism, and data-driven and empirical… — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @KevinQ: We also have openings for a visual specialist — whether this speciality is in information graphics, data visualization, web dev… — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson from the standpoint of the regression, it is entirely possible that trump fares better among wwc if he never says a word about hispanics, muslims, blm, etc., sticks to china/trade and in doing so holds the low-RR WWC. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson it is very difficult to untangle this web. there's no equality among the measurements (maybe a well-calibrated xenophobia scale would have beaten RR). and regression doesn't really help. it can't even really rule out that racism hurt Trump with WWC — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson my second point was that there's a constellation of attitudes correlated with growing class polarization, just as many correlated with generational shift many of those attitudes are RR-adjacent (religion, anti-latino, anti-lgbt, anti-muslim, sexism, etc.). others are not. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson my point is that there's a constellation of attitudes correlate with growing class polarization, just like many issues correlated with generational change. many of those attitudes are RR-adjacent (religion, anti-latino, anti-lgbt, anti-muslim, sexism, etc.). others are not. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson yes, i too have been sent these documents (though vaverick/sides had it in their book first to be fair). by regression-based finding, i just mean that the core insight is that controlling for RR, education isn't so important. there are many charts — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson a tweet like 'it's mostly gay marriage, not age' or 'age is obfuscating' would be... dense. and that's leaving aside another issue: it would be dumb to conclude it's mostly gay marriage based on regression coefficients with asymmetric measurement — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson i would obviously have no problem with that characterization, but I don't think that's really comparable to your point. in this case, you've tweeted things like 'it's mostly RR, not education' or that education is 'obsfuscating' etc — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson it's totally fair to say, at the individual level, that attitudes about gay marriage were more important than age in '08. but that wouldn't justify downplaying the generational change frame in that election. why were the young people pro-gay marriage? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson it's totally fair, at the individual level, that attitudes about gay marriage were more predictive of vote in '08 than age. but if you want to downplay the generational change framing of that election, i think that would be a mistake. why were the young people pro-gay marriage? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson the context of this conversation, as far as i can tell, is that you tried to downplay the role of educational polarization on the grounds that RR > edu in a regression. the effect of that approach is that you lose underlying causes, since they're earlier in opinion formation — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson 'it's not generational change, it's about gay marriage' is just... missing something in '08, even though attitudes about gay marriage prevailed over age in a regression. 'it's not about industrial labor participation, it's about attitudes toward business' is similarly off — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson if you're focused on illuminated an individual's vote choice, sure. if you're talking about sociological, geographic or longitudinal analysis, probably not. those questions demand deeper insight into underlying causes. attitudes don't come from nowhere — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @davidshor @marcushjohnson demographics are at an earlier stage of opinion formation than attitudes, so there are hundreds of beliefs and attitudes with more individual-level value than education — PolitiTweet.org

Nate Cohn @Nate_Cohn

@davidshor @gelliottmorris @marcushjohnson yeah, the huge gap on the predictive value of RR/education in ecological v. individual regression really betrays why putting RR and education in competition like this is kind of dense — PolitiTweet.org

Nate Cohn @Nate_Cohn

The overrepresentation of white Biden voters in '20 is a more serious problem. It was not explained by turnout in the post-election analysis, which you'll get soon enough. It probably did reflect an underlying bias in response rates, which is much harder to address — PolitiTweet.org

Nate Cohn @Nate_Cohn

With that in mind, it did not surprise me--at all--that our polls showed typical Democratic strength among Latino voters and Latino states in '20. There was no methodological change — PolitiTweet.org

Nate Cohn @Nate_Cohn

It is true that our polling underestimated Democrats in some relatively Hispanic states/CDs in 2018, but in the post-election analysis virtually all of the effect was due to turnout/composition of the electorate. — PolitiTweet.org

Nate Cohn @Nate_Cohn

One dynamic in my 'replies' that's surprised me this AM: some followers showing a very good memory in recalling that our '18 polling underestimated Ds among Latinos, while worry about polls underestimated Trump among white voters in 20. The latter point is a lot more serious, IMO — PolitiTweet.org

Nate Cohn @Nate_Cohn

@dcg1114 VA/NJ kind of complicate that story, imo. but it's 51 to 31 among 18-29 — PolitiTweet.org

Nate Cohn @Nate_Cohn

@SpecialPuppy1 D10, R10 — PolitiTweet.org

Nate Cohn @Nate_Cohn

@dcg1114 10% 18-29; 32D, 32R — PolitiTweet.org